mirror of
https://github.com/Monadical-SAS/reflector.git
synced 2025-12-21 04:39:06 +00:00
remove conductor and add hatchet tests (no-mistakes)
This commit is contained in:
Igor Loskutov
@@ -1,617 +0,0 @@
|
|||||||
# Conductor Migration Requirements: Daily.co Multitrack Pipeline
|
|
||||||
|
|
||||||
## Executive Summary
|
|
||||||
|
|
||||||
Migrate the Daily.co multitrack diarization pipeline from a monolithic Celery task to a decomposed Conductor workflow, enabling visual progress tracking, granular retries, and operational observability.
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Business Value
|
|
||||||
|
|
||||||
### 1. Visibility: Where Are We Now? (UX, DevEx)
|
|
||||||
|
|
||||||
**Current State**: Users see only three states: `idle` → `processing` → `ended/error`. A 10-minute pipeline appears frozen with no feedback.
|
|
||||||
|
|
||||||
**Target State**: Real-time visibility into which step is executing:
|
|
||||||
- "Transcribing track 2 of 3"
|
|
||||||
- "Generating summary (step 8 of 9)"
|
|
||||||
- Visual DAG in admin UI showing completed/in-progress/pending nodes
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Reduced support tickets ("is it stuck?")
|
|
||||||
- Engineers can instantly identify bottlenecks
|
|
||||||
- Users have confidence the system is working
|
|
||||||
|
|
||||||
### 2. Progress Tracking: What's Left? (UX, DevEx)
|
|
||||||
|
|
||||||
**Current State**: No indication of remaining work. A failure at step 8 gives same error as failure at step 1.
|
|
||||||
|
|
||||||
**Target State**:
|
|
||||||
- Progress percentage based on completed steps
|
|
||||||
- Clear step enumeration (e.g., "Step 5/9: Transcription")
|
|
||||||
- Frontend receives structured progress events with step metadata
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Users can estimate completion time
|
|
||||||
- Frontend can render meaningful progress bars
|
|
||||||
- Error messages include context ("Failed during summary generation")
|
|
||||||
|
|
||||||
### 3. Audit Trail & Profiling (DevEx, Ops)
|
|
||||||
|
|
||||||
**Current State**: Logs scattered across Celery workers. No unified view of a single recording's journey. Resource consumption unknown per step.
|
|
||||||
|
|
||||||
**Target State**:
|
|
||||||
- Single workflow ID traces entire recording lifecycle
|
|
||||||
- Per-step execution times recorded
|
|
||||||
- Resource consumption (GPU seconds, LLM tokens) attributable to specific steps
|
|
||||||
- Conductor UI provides complete audit history
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Debugging: "Recording X failed at step Y after Z seconds"
|
|
||||||
- Cost attribution: "Transcription costs $X, summarization costs $Y"
|
|
||||||
- Performance optimization: identify slowest steps
|
|
||||||
|
|
||||||
### 4. Clear Event Dictionary (DevEx)
|
|
||||||
|
|
||||||
**Current State**: Frontend receives WebSocket events (`TRANSCRIPT`, `TOPIC`, `FINAL_TITLE`, etc.) but mapping to pipeline phases is implicit. Adding new events requires tracing through Python code.
|
|
||||||
|
|
||||||
**Target State**:
|
|
||||||
- Each Conductor task explicitly defines its output events
|
|
||||||
- Event schema documented alongside task definition
|
|
||||||
- Frontend developers can reference task→event mapping directly
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Faster frontend development
|
|
||||||
- Reduced miscommunication between backend/frontend teams
|
|
||||||
- Self-documenting pipeline
|
|
||||||
|
|
||||||
### 5. Restart Without Reprocessing (UX, DevEx)
|
|
||||||
|
|
||||||
**Current State**: Any failure restarts the entire pipeline. A timeout during summary generation re-runs transcription (wasting GPU costs).
|
|
||||||
|
|
||||||
**Target State**:
|
|
||||||
- Failures resume from last successful step
|
|
||||||
- Completed work is checkpointed (e.g., transcription results stored before summary)
|
|
||||||
- Manual retry triggers only failed step, not entire workflow
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Reduced GPU/LLM costs on retries
|
|
||||||
- Faster recovery from transient failures
|
|
||||||
- Users don't wait for re-transcription on summary failures
|
|
||||||
|
|
||||||
### 6. Per-Step Timeouts (UX, DevEx)
|
|
||||||
|
|
||||||
**Current State**: Single task timeout for entire pipeline. A hung GPU call blocks everything. Killing the task loses all progress.
|
|
||||||
|
|
||||||
**Target State**:
|
|
||||||
- Each step has independent timeout (e.g., transcription: 5min, LLM: 30s)
|
|
||||||
- Timeout kills only the hung step
|
|
||||||
- Pipeline can retry just that step or fail gracefully
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Faster detection of stuck external services
|
|
||||||
- Reduced blast radius from hung calls
|
|
||||||
- More granular SLAs per operation type
|
|
||||||
|
|
||||||
### 7. Native Retries with Backoff (DevEx, UX)
|
|
||||||
|
|
||||||
**Current State**: Celery retry logic is per-task, not per-external-call. Custom retry wrappers needed for each API call.
|
|
||||||
|
|
||||||
**Target State**:
|
|
||||||
- Conductor provides native retry policies per task
|
|
||||||
- Exponential backoff configured declaratively
|
|
||||||
- Retry state visible in UI (attempt 2/5)
|
|
||||||
|
|
||||||
**Business Impact**:
|
|
||||||
- Reduced boilerplate code
|
|
||||||
- Consistent retry behavior across all external calls
|
|
||||||
- Visibility into retry attempts for debugging
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Current Architecture
|
|
||||||
|
|
||||||
### Daily.co Multitrack Pipeline Flow
|
|
||||||
|
|
||||||
```
|
|
||||||
Daily webhook (recording.ready-to-download) Polling (every 3 min)
|
|
||||||
│ │
|
|
||||||
▼ ▼
|
|
||||||
_handle_recording_ready() poll_daily_recordings()
|
|
||||||
│ │
|
|
||||||
└──────────────┬─────────────────────────┘
|
|
||||||
▼
|
|
||||||
process_multitrack_recording.delay() ← Celery task #1
|
|
||||||
│
|
|
||||||
├── Daily API: GET /recordings/{id}
|
|
||||||
├── Daily API: GET /meetings/{mtgSessionId}/participants
|
|
||||||
├── DB: Create recording + transcript
|
|
||||||
│
|
|
||||||
▼
|
|
||||||
task_pipeline_multitrack_process.delay() ← Celery task #2 (MONOLITH)
|
|
||||||
│
|
|
||||||
│ ┌─────────────────────────────────────────────────┐
|
|
||||||
│ │ pipeline.process() - ALL PHASES INSIDE HERE │
|
|
||||||
│ │ │
|
|
||||||
│ │ Phase 2: Track Padding (N tracks, sequential) │
|
|
||||||
│ │ Phase 3: Mixdown → S3 upload │
|
|
||||||
│ │ Phase 4: Waveform generation │
|
|
||||||
│ │ Phase 5: Transcription (N GPU calls, serial!) │
|
|
||||||
│ │ Phase 6: Topic Detection (C LLM calls) │
|
|
||||||
│ │ Phase 7a: Title Generation (1 LLM call) │
|
|
||||||
│ │ Phase 7b: Summary Generation (2+2M LLM calls) │
|
|
||||||
│ │ Phase 8: Finalize status │
|
|
||||||
│ └─────────────────────────────────────────────────┘
|
|
||||||
│
|
|
||||||
▼
|
|
||||||
chain(cleanup → zulip → webhook).delay() ← Celery chain (3 tasks)
|
|
||||||
```
|
|
||||||
|
|
||||||
### Problem: Monolithic `pipeline.process()`
|
|
||||||
|
|
||||||
The heavy lifting happens inside a single Python function call. Celery only sees:
|
|
||||||
- Task started
|
|
||||||
- Task succeeded/failed
|
|
||||||
|
|
||||||
It cannot see or control the 8 internal phases.
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Target Architecture
|
|
||||||
|
|
||||||
### Decomposed Conductor Workflow
|
|
||||||
|
|
||||||
```
|
|
||||||
┌─────────────────────┐
|
|
||||||
│ get_recording │ ← Daily API
|
|
||||||
│ get_participants │
|
|
||||||
└──────────┬──────────┘
|
|
||||||
│
|
|
||||||
┌──────────────────┼──────────────────┐
|
|
||||||
▼ ▼ ▼
|
|
||||||
┌──────────┐ ┌──────────┐ ┌──────────┐
|
|
||||||
│ pad_tk_0 │ │ pad_tk_1 │ │ pad_tk_N │ ← FORK (parallel)
|
|
||||||
└────┬─────┘ └────┬─────┘ └────┬─────┘
|
|
||||||
└──────────────────┼──────────────────┘
|
|
||||||
▼
|
|
||||||
┌─────────────────────┐
|
|
||||||
│ mixdown_tracks │ ← PyAV → S3
|
|
||||||
└──────────┬──────────┘
|
|
||||||
│
|
|
||||||
┌──────────┴──────────┐
|
|
||||||
▼ ▼
|
|
||||||
┌───────────────┐ ┌───────────────┐
|
|
||||||
│generate_wave │ │ (continue) │ ← waveform parallel with transcription setup
|
|
||||||
└───────────────┘ └───────────────┘
|
|
||||||
│
|
|
||||||
┌──────────────────┼──────────────────┐
|
|
||||||
▼ ▼ ▼
|
|
||||||
┌────────────┐ ┌────────────┐ ┌────────────┐
|
|
||||||
│transcribe_0│ │transcribe_1│ │transcribe_N│ ← FORK (parallel GPU!)
|
|
||||||
└─────┬──────┘ └─────┬──────┘ └─────┬──────┘
|
|
||||||
└──────────────────┼──────────────────┘
|
|
||||||
▼
|
|
||||||
┌─────────────────────┐
|
|
||||||
│ merge_transcripts │
|
|
||||||
└──────────┬──────────┘
|
|
||||||
│
|
|
||||||
┌──────────┴──────────┐
|
|
||||||
▼ ▼
|
|
||||||
┌───────────────┐ ┌───────────────┐
|
|
||||||
│detect_topics │ │ (or) │ ← topic detection
|
|
||||||
└───────┬───────┘ └───────────────┘
|
|
||||||
│
|
|
||||||
┌──────────────┴──────────────┐
|
|
||||||
▼ ▼
|
|
||||||
┌─────────────┐ ┌─────────────┐
|
|
||||||
│generate_title│ │gen_summary │ ← FORK (parallel LLM)
|
|
||||||
└──────┬──────┘ └──────┬──────┘
|
|
||||||
└──────────────┬─────────────┘
|
|
||||||
▼
|
|
||||||
┌─────────────────────┐
|
|
||||||
│ finalize │
|
|
||||||
└──────────┬──────────┘
|
|
||||||
│
|
|
||||||
┌──────────────┼──────────────┐
|
|
||||||
▼ ▼ ▼
|
|
||||||
┌──────────┐ ┌──────────┐ ┌──────────┐
|
|
||||||
│ consent │──▶│ zulip │──▶│ webhook │ ← sequential chain
|
|
||||||
└──────────┘ └──────────┘ └──────────┘
|
|
||||||
```
|
|
||||||
|
|
||||||
### Key Improvements
|
|
||||||
|
|
||||||
| Aspect | Current (Celery) | Target (Conductor) |
|
|
||||||
|--------|------------------|-------------------|
|
|
||||||
| Transcription parallelism | Serial (N × 30s) | Parallel (max 30s) |
|
|
||||||
| Failure granularity | Restart all | Retry failed step only |
|
|
||||||
| Progress visibility | None | Per-step status in UI |
|
|
||||||
| Timeout control | Entire pipeline | Per-step timeouts |
|
|
||||||
| Audit trail | Scattered logs | Unified workflow history |
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Scope of Work
|
|
||||||
|
|
||||||
### Module 1: Conductor Infrastructure Setup
|
|
||||||
|
|
||||||
**Files to Create/Modify:**
|
|
||||||
- `docker-compose.yml` - Add Conductor server container
|
|
||||||
- `server/reflector/conductor/` - New module for Conductor client
|
|
||||||
- Environment configuration for Conductor URL
|
|
||||||
|
|
||||||
**Tasks:**
|
|
||||||
- [ ] Add `conductoross/conductor-standalone:3.15.0` to docker-compose
|
|
||||||
- [ ] Create Conductor client wrapper (Python `conductor-python` SDK)
|
|
||||||
- [ ] Configure health checks and service dependencies
|
|
||||||
- [ ] Document Conductor UI access (port 8127)
|
|
||||||
|
|
||||||
### Module 2: Task Decomposition - Worker Definitions
|
|
||||||
|
|
||||||
**Files to Create:**
|
|
||||||
- `server/reflector/conductor/workers/` directory with:
|
|
||||||
- `get_recording.py` - Daily API recording fetch
|
|
||||||
- `get_participants.py` - Daily API participant fetch
|
|
||||||
- `pad_track.py` - Single track padding (PyAV)
|
|
||||||
- `mixdown_tracks.py` - Multi-track mixdown
|
|
||||||
- `generate_waveform.py` - Waveform generation
|
|
||||||
- `transcribe_track.py` - Single track GPU transcription
|
|
||||||
- `merge_transcripts.py` - Combine transcriptions
|
|
||||||
- `detect_topics.py` - LLM topic detection
|
|
||||||
- `generate_title.py` - LLM title generation
|
|
||||||
- `generate_summary.py` - LLM summary generation
|
|
||||||
- `finalize.py` - Status update and cleanup
|
|
||||||
- `cleanup_consent.py` - Consent check
|
|
||||||
- `post_zulip.py` - Zulip notification
|
|
||||||
- `send_webhook.py` - External webhook
|
|
||||||
- `generate_dynamic_fork_tasks.py` - Helper for FORK_JOIN_DYNAMIC task generation
|
|
||||||
|
|
||||||
**Reference Files (Current Implementation):**
|
|
||||||
- `server/reflector/pipelines/main_multitrack_pipeline.py`
|
|
||||||
- `server/reflector/worker/process.py`
|
|
||||||
- `server/reflector/worker/webhook.py`
|
|
||||||
|
|
||||||
**Key Considerations:**
|
|
||||||
- Each worker receives input from previous step via Conductor
|
|
||||||
- Workers must be idempotent (same input → same output)
|
|
||||||
- State serialization between steps (JSON-compatible types)
|
|
||||||
|
|
||||||
### Module 3: Workflow Definition
|
|
||||||
|
|
||||||
**Files to Create:**
|
|
||||||
- `server/reflector/conductor/workflows/diarization_pipeline.json`
|
|
||||||
- `server/reflector/conductor/workflows/register.py` - Registration script
|
|
||||||
|
|
||||||
**Workflow Structure:**
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"name": "daily_diarization_pipeline",
|
|
||||||
"version": 1,
|
|
||||||
"tasks": [
|
|
||||||
{"name": "get_recording", "type": "SIMPLE"},
|
|
||||||
{"name": "get_participants", "type": "SIMPLE"},
|
|
||||||
{
|
|
||||||
"name": "fork_padding",
|
|
||||||
"type": "FORK_JOIN_DYNAMIC",
|
|
||||||
"dynamicForkTasksParam": "track_keys"
|
|
||||||
},
|
|
||||||
{"name": "mixdown_tracks", "type": "SIMPLE"},
|
|
||||||
{"name": "generate_waveform", "type": "SIMPLE"},
|
|
||||||
{
|
|
||||||
"name": "fork_transcription",
|
|
||||||
"type": "FORK_JOIN_DYNAMIC",
|
|
||||||
"dynamicForkTasksParam": "padded_urls"
|
|
||||||
},
|
|
||||||
{"name": "merge_transcripts", "type": "SIMPLE"},
|
|
||||||
{"name": "detect_topics", "type": "SIMPLE"},
|
|
||||||
{
|
|
||||||
"name": "fork_generation",
|
|
||||||
"type": "FORK_JOIN",
|
|
||||||
"forkTasks": [["generate_title"], ["generate_summary"]]
|
|
||||||
},
|
|
||||||
{"name": "finalize", "type": "SIMPLE"},
|
|
||||||
{"name": "cleanup_consent", "type": "SIMPLE"},
|
|
||||||
{"name": "post_zulip", "type": "SIMPLE"},
|
|
||||||
{"name": "send_webhook", "type": "SIMPLE"}
|
|
||||||
]
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
**Key Considerations:**
|
|
||||||
- Dynamic FORK for variable number of tracks (N)
|
|
||||||
- Timeout configuration per task type
|
|
||||||
- Retry policies with exponential backoff
|
|
||||||
|
|
||||||
### Module 4: Pipeline Trigger Migration
|
|
||||||
|
|
||||||
**Files to Modify:**
|
|
||||||
- `server/reflector/worker/process.py`
|
|
||||||
|
|
||||||
**Changes:**
|
|
||||||
- Replace `task_pipeline_multitrack_process.delay()` with Conductor workflow start
|
|
||||||
- Store workflow ID on Recording for status tracking
|
|
||||||
- Handle Conductor API errors
|
|
||||||
- Keep `process_multitrack_recording` as-is (creates DB entities before workflow)
|
|
||||||
|
|
||||||
**Note:** Both webhook AND polling entry points converge at `process_multitrack_recording`,
|
|
||||||
which then calls `task_pipeline_multitrack_process.delay()`. By modifying this single call site,
|
|
||||||
we capture both entry paths without duplicating integration logic.
|
|
||||||
|
|
||||||
### Module 5: Task Definition Registration
|
|
||||||
|
|
||||||
**Files to Create:**
|
|
||||||
- `server/reflector/conductor/tasks/definitions.py`
|
|
||||||
|
|
||||||
**Task Definitions with Timeouts:**
|
|
||||||
|
|
||||||
| Task | Timeout | Response Timeout | Retry Count |
|
|
||||||
|------|---------|------------------|-------------|
|
|
||||||
| get_recording | 60s | 30s | 3 |
|
|
||||||
| get_participants | 60s | 30s | 3 |
|
|
||||||
| pad_track | 300s | 120s | 3 |
|
|
||||||
| mixdown_tracks | 600s | 300s | 3 |
|
|
||||||
| generate_waveform | 120s | 60s | 3 |
|
|
||||||
| transcribe_track | 1800s | 900s | 3 |
|
|
||||||
| merge_transcripts | 60s | 30s | 3 |
|
|
||||||
| detect_topics | 300s | 120s | 3 |
|
|
||||||
| generate_title | 60s | 30s | 3 |
|
|
||||||
| generate_summary | 300s | 120s | 3 |
|
|
||||||
| finalize | 60s | 30s | 3 |
|
|
||||||
| cleanup_consent | 60s | 30s | 3 |
|
|
||||||
| post_zulip | 60s | 30s | 5 |
|
|
||||||
| send_webhook | 60s | 30s | 30 |
|
|
||||||
| generate_dynamic_fork_tasks | 30s | 15s | 3 |
|
|
||||||
|
|
||||||
### Module 6: Frontend Integration
|
|
||||||
|
|
||||||
**WebSocket Events (Already Defined):**
|
|
||||||
|
|
||||||
Events continue to be broadcast as today. No change to event structure.
|
|
||||||
|
|
||||||
| Event | Triggered By Task | Payload |
|
|
||||||
|-------|-------------------|---------|
|
|
||||||
| STATUS | finalize | `{value: "processing"\|"ended"\|"error"}` |
|
|
||||||
| DURATION | mixdown_tracks | `{duration: float}` |
|
|
||||||
| WAVEFORM | generate_waveform | `{waveform: float[]}` |
|
|
||||||
| TRANSCRIPT | merge_transcripts | `{text: string, translation: string\|null}` |
|
|
||||||
| TOPIC | detect_topics | `{id, title, summary, timestamp, duration}` |
|
|
||||||
| FINAL_TITLE | generate_title | `{title: string}` |
|
|
||||||
| FINAL_LONG_SUMMARY | generate_summary | `{long_summary: string}` |
|
|
||||||
| FINAL_SHORT_SUMMARY | generate_summary | `{short_summary: string}` |
|
|
||||||
|
|
||||||
**New: Progress Events**
|
|
||||||
|
|
||||||
Add new event type for granular progress:
|
|
||||||
|
|
||||||
```python
|
|
||||||
# PipelineProgressEvent
|
|
||||||
{
|
|
||||||
"event": "PIPELINE_PROGRESS",
|
|
||||||
"data": {
|
|
||||||
"workflow_id": str,
|
|
||||||
"current_step": str,
|
|
||||||
"step_index": int,
|
|
||||||
"total_steps": int,
|
|
||||||
"step_status": "pending" | "in_progress" | "completed" | "failed"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
### Module 7: State Management & Checkpointing
|
|
||||||
|
|
||||||
**Current State Storage:**
|
|
||||||
- `transcript.status` - High-level status
|
|
||||||
- `transcript.events[]` - Append-only event log
|
|
||||||
- `transcript.topics[]` - Topic results
|
|
||||||
- `transcript.title`, `transcript.long_summary`, etc.
|
|
||||||
|
|
||||||
**Conductor State Storage:**
|
|
||||||
- Workflow execution state in Conductor database
|
|
||||||
- Per-task input/output in Conductor
|
|
||||||
|
|
||||||
**Checkpointing Strategy:**
|
|
||||||
1. Each task reads required state from DB (not previous task output for large data)
|
|
||||||
2. Each task writes results to DB before returning
|
|
||||||
3. Task output contains references (IDs, URLs) not large payloads
|
|
||||||
4. On retry, task can check DB for existing results (idempotency)
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Data Flow Between Tasks
|
|
||||||
|
|
||||||
### Input/Output Contracts
|
|
||||||
|
|
||||||
```
|
|
||||||
get_recording
|
|
||||||
Input: { recording_id: string }
|
|
||||||
Output: { id, mtg_session_id, room_name, duration }
|
|
||||||
|
|
||||||
get_participants
|
|
||||||
Input: { mtg_session_id: string }
|
|
||||||
Output: { participants: [{participant_id, user_name}] }
|
|
||||||
|
|
||||||
pad_track
|
|
||||||
Input: { track_index: number, s3_key: string }
|
|
||||||
Output: { padded_url: string, size: number }
|
|
||||||
|
|
||||||
mixdown_tracks
|
|
||||||
Input: { padded_urls: string[] }
|
|
||||||
Output: { audio_key: string, duration: number }
|
|
||||||
|
|
||||||
generate_waveform
|
|
||||||
Input: { audio_key: string }
|
|
||||||
Output: { waveform: number[] }
|
|
||||||
|
|
||||||
transcribe_track
|
|
||||||
Input: { track_index: number, audio_url: string }
|
|
||||||
Output: { words: Word[] }
|
|
||||||
|
|
||||||
merge_transcripts
|
|
||||||
Input: { transcripts: Word[][] }
|
|
||||||
Output: { all_words: Word[], word_count: number }
|
|
||||||
|
|
||||||
detect_topics
|
|
||||||
Input: { words: Word[] }
|
|
||||||
Output: { topics: Topic[] }
|
|
||||||
|
|
||||||
generate_title
|
|
||||||
Input: { topics: Topic[] }
|
|
||||||
Output: { title: string }
|
|
||||||
|
|
||||||
generate_summary
|
|
||||||
Input: { words: Word[], topics: Topic[] }
|
|
||||||
Output: { summary: string, short_summary: string }
|
|
||||||
|
|
||||||
finalize
|
|
||||||
Input: { recording_id, title, summary, duration }
|
|
||||||
Output: { status: "COMPLETED" }
|
|
||||||
```
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## External API Calls Summary
|
|
||||||
|
|
||||||
### Per-Step External Dependencies
|
|
||||||
|
|
||||||
| Task | External Service | Calls | Notes |
|
|
||||||
|------|------------------|-------|-------|
|
|
||||||
| get_recording | Daily.co API | 1 | GET /recordings/{id} |
|
|
||||||
| get_participants | Daily.co API | 1 | GET /meetings/{id}/participants |
|
|
||||||
| pad_track | S3 | 2 | presign read + PUT padded |
|
|
||||||
| mixdown_tracks | S3 | 1 | PUT audio.mp3 |
|
|
||||||
| transcribe_track | Modal.com GPU | 1 | POST /transcriptions |
|
|
||||||
| detect_topics | LLM (OpenAI) | C | C = ceil(words/300) |
|
|
||||||
| generate_title | LLM (OpenAI) | 1 | - |
|
|
||||||
| generate_summary | LLM (OpenAI) | 2+2M | M = subjects (max 6) |
|
|
||||||
| post_zulip | Zulip API | 1 | POST or PATCH |
|
|
||||||
| send_webhook | External | 1 | Customer webhook URL |
|
|
||||||
|
|
||||||
### Cost Attribution Enabled
|
|
||||||
|
|
||||||
With decomposed tasks, costs can be attributed:
|
|
||||||
- **GPU costs**: Sum of `transcribe_track` durations
|
|
||||||
- **LLM costs**: Sum of `detect_topics` + `generate_title` + `generate_summary` token usage
|
|
||||||
- **S3 costs**: Bytes uploaded by `pad_track` + `mixdown_tracks`
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Idempotency Requirements
|
|
||||||
|
|
||||||
### By Task
|
|
||||||
|
|
||||||
| Task | Idempotent? | Strategy |
|
|
||||||
|------|-------------|----------|
|
|
||||||
| get_recording | ✅ | Read-only API call |
|
|
||||||
| get_participants | ✅ | Read-only API call |
|
|
||||||
| pad_track | ⚠️ | Overwrite same S3 key |
|
|
||||||
| mixdown_tracks | ⚠️ | Overwrite same S3 key |
|
|
||||||
| generate_waveform | ✅ | Deterministic from audio |
|
|
||||||
| transcribe_track | ❌ | Cache by hash(audio_url) |
|
|
||||||
| detect_topics | ❌ | Cache by hash(words) |
|
|
||||||
| generate_title | ❌ | Cache by hash(topic_titles) |
|
|
||||||
| generate_summary | ❌ | Cache by hash(words+topics) |
|
|
||||||
| finalize | ✅ | Upsert status |
|
|
||||||
| cleanup_consent | ✅ | Idempotent deletes |
|
|
||||||
| post_zulip | ⚠️ | Use message_id for updates |
|
|
||||||
| send_webhook | ⚠️ | Receiver's responsibility |
|
|
||||||
|
|
||||||
### Caching Strategy for LLM/GPU Calls
|
|
||||||
|
|
||||||
```python
|
|
||||||
class TaskCache:
|
|
||||||
async def get(self, input_hash: str) -> Optional[Output]: ...
|
|
||||||
async def set(self, input_hash: str, output: Output) -> None: ...
|
|
||||||
|
|
||||||
# Before calling external service:
|
|
||||||
cached = await cache.get(hash(input))
|
|
||||||
if cached:
|
|
||||||
return cached
|
|
||||||
|
|
||||||
result = await external_service.call(input)
|
|
||||||
await cache.set(hash(input), result)
|
|
||||||
return result
|
|
||||||
```
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Migration Strategy
|
|
||||||
|
|
||||||
### Phase 1: Infrastructure (No Behavior Change)
|
|
||||||
- Add Conductor container to docker-compose
|
|
||||||
- Create Conductor client library
|
|
||||||
- Verify Conductor UI accessible
|
|
||||||
|
|
||||||
### Phase 2: Parallel Implementation
|
|
||||||
- Implement all worker tasks
|
|
||||||
- Register workflow definition
|
|
||||||
- Test with synthetic recordings
|
|
||||||
|
|
||||||
### Phase 3: Shadow Mode
|
|
||||||
- Trigger both Celery and Conductor pipelines
|
|
||||||
- Compare results for consistency
|
|
||||||
- Monitor Conductor execution in UI
|
|
||||||
|
|
||||||
### Phase 4: Cutover
|
|
||||||
- Disable Celery pipeline trigger
|
|
||||||
- Enable Conductor-only execution
|
|
||||||
- Monitor error rates and performance
|
|
||||||
|
|
||||||
### Phase 5: Cleanup
|
|
||||||
- Remove Celery task definitions
|
|
||||||
- Remove old pipeline code
|
|
||||||
- Update documentation
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Risks & Mitigations
|
|
||||||
|
|
||||||
| Risk | Mitigation |
|
|
||||||
|------|------------|
|
|
||||||
| Conductor server downtime | Health checks, failover to Celery (Phase 3) |
|
|
||||||
| Worker serialization issues | Extensive testing with real data |
|
|
||||||
| Performance regression | Benchmark parallel vs serial transcription |
|
|
||||||
| Data loss on migration | Shadow mode comparison (Phase 3) |
|
|
||||||
| Learning curve for team | Documentation, Conductor UI training |
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Success Metrics
|
|
||||||
|
|
||||||
| Metric | Current | Target |
|
|
||||||
|--------|---------|--------|
|
|
||||||
| Pipeline visibility | 3 states | 14+ steps visible |
|
|
||||||
| Transcription latency (N tracks) | N × 30s | ~30s (parallel) |
|
|
||||||
| Retry granularity | Entire pipeline | Single step |
|
|
||||||
| Cost attribution | None | Per-step breakdown |
|
|
||||||
| Debug time for failures | ~30 min | ~5 min (UI trace) |
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Appendix: Conductor Mock Implementation
|
|
||||||
|
|
||||||
A working Python mock demonstrating the target workflow structure is available at:
|
|
||||||
`docs/conductor-pipeline-mock/`
|
|
||||||
|
|
||||||
To run:
|
|
||||||
```bash
|
|
||||||
cd docs/conductor-pipeline-mock
|
|
||||||
docker compose up --build
|
|
||||||
./test_workflow.sh
|
|
||||||
```
|
|
||||||
|
|
||||||
UI: http://localhost:8127
|
|
||||||
|
|
||||||
This mock validates:
|
|
||||||
- Workflow definition structure
|
|
||||||
- FORK_JOIN parallelism
|
|
||||||
- Worker task patterns
|
|
||||||
- Conductor SDK usage
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## References
|
|
||||||
|
|
||||||
- Diarization Pipeline Diagram: `DIARIZATION_PIPELINE_DIAGRAM.md`
|
|
||||||
- Current Celery Implementation: `server/reflector/pipelines/main_multitrack_pipeline.py`
|
|
||||||
- Conductor OSS Documentation: https://conductor-oss.github.io/conductor/
|
|
||||||
- Conductor Python SDK: https://github.com/conductor-sdk/conductor-python
|
|
||||||
125
TASKS.md
125
TASKS.md
@@ -1,6 +1,6 @@
|
|||||||
# Durable Workflow Migration Tasks
|
# Durable Workflow Migration Tasks
|
||||||
|
|
||||||
This document defines atomic, isolated work items for migrating the Daily.co multitrack diarization pipeline from Celery to durable workflow orchestration. Supports both **Conductor** and **Hatchet** via `DURABLE_WORKFLOW_PROVIDER` env var.
|
This document defines atomic, isolated work items for migrating the Daily.co multitrack diarization pipeline from Celery to durable workflow orchestration using **Hatchet**.
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
||||||
@@ -9,91 +9,46 @@ This document defines atomic, isolated work items for migrating the Daily.co mul
|
|||||||
```bash
|
```bash
|
||||||
# .env
|
# .env
|
||||||
DURABLE_WORKFLOW_PROVIDER=none # Celery only (default)
|
DURABLE_WORKFLOW_PROVIDER=none # Celery only (default)
|
||||||
DURABLE_WORKFLOW_PROVIDER=conductor # Use Conductor
|
|
||||||
DURABLE_WORKFLOW_PROVIDER=hatchet # Use Hatchet
|
DURABLE_WORKFLOW_PROVIDER=hatchet # Use Hatchet
|
||||||
DURABLE_WORKFLOW_SHADOW_MODE=true # Run both provider + Celery (for comparison)
|
DURABLE_WORKFLOW_SHADOW_MODE=true # Run both Hatchet + Celery (for comparison)
|
||||||
```
|
```
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
||||||
## Task Index
|
## Task Index
|
||||||
|
|
||||||
| ID | Title | Status | Conductor | Hatchet |
|
| ID | Title | Status |
|
||||||
|----|-------|--------|-----------|---------|
|
|----|-------|--------|
|
||||||
| INFRA-001 | Add container to docker-compose | Done | ✓ | ✓ |
|
| INFRA-001 | Add container to docker-compose | Done |
|
||||||
| INFRA-002 | Create Python client wrapper | Done | ✓ | ✓ |
|
| INFRA-002 | Create Python client wrapper | Done |
|
||||||
| INFRA-003 | Add environment configuration | Done | ✓ | ✓ |
|
| INFRA-003 | Add environment configuration | Done |
|
||||||
| TASK-001 | Create task definitions/workflow | Done | ✓ JSON | ✓ Python |
|
| TASK-001 | Create workflow definition | Done |
|
||||||
| TASK-002 | get_recording worker | Done | ✓ | ✓ |
|
| TASK-002 | get_recording task | Done |
|
||||||
| TASK-003 | get_participants worker | Done | ✓ | ✓ |
|
| TASK-003 | get_participants task | Done |
|
||||||
| TASK-004 | pad_track worker | Done | ✓ | ✓ |
|
| TASK-004 | pad_track task | Done |
|
||||||
| TASK-005 | mixdown_tracks worker | Done | ✓ | ✓ |
|
| TASK-005 | mixdown_tracks task | Done |
|
||||||
| TASK-006 | generate_waveform worker | Done | ✓ | ✓ |
|
| TASK-006 | generate_waveform task | Done |
|
||||||
| TASK-007 | transcribe_track worker | Done | ✓ | ✓ |
|
| TASK-007 | transcribe_track task | Done |
|
||||||
| TASK-008 | merge_transcripts worker | Done | ✓ | ✓ (in process_tracks) |
|
| TASK-008 | merge_transcripts task | Done (in process_tracks) |
|
||||||
| TASK-009 | detect_topics worker | Done | ✓ | ✓ |
|
| TASK-009 | detect_topics task | Done |
|
||||||
| TASK-010 | generate_title worker | Done | ✓ | ✓ |
|
| TASK-010 | generate_title task | Done |
|
||||||
| TASK-011 | generate_summary worker | Done | ✓ | ✓ |
|
| TASK-011 | generate_summary task | Done |
|
||||||
| TASK-012 | finalize worker | Done | ✓ | ✓ |
|
| TASK-012 | finalize task | Done |
|
||||||
| TASK-013 | cleanup_consent worker | Done | ✓ | ✓ |
|
| TASK-013 | cleanup_consent task | Done |
|
||||||
| TASK-014 | post_zulip worker | Done | ✓ | ✓ |
|
| TASK-014 | post_zulip task | Done |
|
||||||
| TASK-015 | send_webhook worker | Done | ✓ | ✓ |
|
| TASK-015 | send_webhook task | Done |
|
||||||
| EVENT-001 | Progress WebSocket events | Done | ✓ | ✓ |
|
| EVENT-001 | Progress WebSocket events | Done |
|
||||||
| INTEG-001 | Pipeline trigger integration | Done | ✓ | ✓ |
|
| INTEG-001 | Pipeline trigger integration | Done |
|
||||||
| SHADOW-001 | Shadow mode toggle | Done | ✓ | ✓ |
|
| SHADOW-001 | Shadow mode toggle | Done |
|
||||||
| TEST-001 | Integration tests | Pending | - | - |
|
| TEST-001 | Integration tests | Pending |
|
||||||
| TEST-002 | E2E workflow test | Pending | - | - |
|
| TEST-002 | E2E workflow test | Pending |
|
||||||
| CUTOVER-001 | Production cutover | Pending | - | - |
|
| CUTOVER-001 | Production cutover | Pending |
|
||||||
| CLEANUP-001 | Remove Celery code | Pending | - | - |
|
| CLEANUP-001 | Remove Celery code | Pending |
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Architecture Differences
|
|
||||||
|
|
||||||
| Aspect | Conductor | Hatchet |
|
|
||||||
|--------|-----------|---------|
|
|
||||||
| Worker model | Multiprocessing (fork) | Async (single process) |
|
|
||||||
| Task communication | REST polling | gRPC streaming |
|
|
||||||
| Workflow definition | JSON files | Python decorators |
|
|
||||||
| Child workflows | FORK_JOIN_DYNAMIC + JOIN task | `aio_run()` returns directly |
|
|
||||||
| Task definitions | Separate worker files | Embedded in workflow |
|
|
||||||
| Debug logging | Limited | Excellent with `HATCHET_DEBUG=true` |
|
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
||||||
## File Structure
|
## File Structure
|
||||||
|
|
||||||
### Conductor
|
|
||||||
```
|
|
||||||
server/reflector/conductor/
|
|
||||||
├── client.py # SDK wrapper
|
|
||||||
├── progress.py # WebSocket progress emission
|
|
||||||
├── run_workers.py # Worker startup
|
|
||||||
├── shadow_compare.py # Shadow mode comparison
|
|
||||||
├── tasks/
|
|
||||||
│ ├── definitions.py # Task definitions with timeouts
|
|
||||||
│ └── register.py # Registration script
|
|
||||||
├── workers/
|
|
||||||
│ ├── get_recording.py
|
|
||||||
│ ├── get_participants.py
|
|
||||||
│ ├── pad_track.py
|
|
||||||
│ ├── mixdown_tracks.py
|
|
||||||
│ ├── generate_waveform.py
|
|
||||||
│ ├── transcribe_track.py
|
|
||||||
│ ├── merge_transcripts.py
|
|
||||||
│ ├── detect_topics.py
|
|
||||||
│ ├── generate_title.py
|
|
||||||
│ ├── generate_summary.py
|
|
||||||
│ ├── finalize.py
|
|
||||||
│ ├── cleanup_consent.py
|
|
||||||
│ ├── post_zulip.py
|
|
||||||
│ ├── send_webhook.py
|
|
||||||
│ └── generate_dynamic_fork_tasks.py
|
|
||||||
└── workflows/
|
|
||||||
└── register.py
|
|
||||||
```
|
|
||||||
|
|
||||||
### Hatchet
|
|
||||||
```
|
```
|
||||||
server/reflector/hatchet/
|
server/reflector/hatchet/
|
||||||
├── client.py # SDK wrapper
|
├── client.py # SDK wrapper
|
||||||
@@ -109,9 +64,8 @@ server/reflector/hatchet/
|
|||||||
## Remaining Work
|
## Remaining Work
|
||||||
|
|
||||||
### TEST-001: Integration Tests
|
### TEST-001: Integration Tests
|
||||||
- [ ] Test each worker with mocked external services
|
- [ ] Test each task with mocked external services
|
||||||
- [ ] Test error handling and retries
|
- [ ] Test error handling and retries
|
||||||
- [ ] Test both Conductor and Hatchet paths
|
|
||||||
|
|
||||||
### TEST-002: E2E Workflow Test
|
### TEST-002: E2E Workflow Test
|
||||||
- [ ] Complete workflow run with real Daily.co recording
|
- [ ] Complete workflow run with real Daily.co recording
|
||||||
@@ -119,7 +73,7 @@ server/reflector/hatchet/
|
|||||||
- [ ] Performance comparison
|
- [ ] Performance comparison
|
||||||
|
|
||||||
### CUTOVER-001: Production Cutover
|
### CUTOVER-001: Production Cutover
|
||||||
- [ ] Deploy with `DURABLE_WORKFLOW_PROVIDER=conductor` or `hatchet`
|
- [ ] Deploy with `DURABLE_WORKFLOW_PROVIDER=hatchet`
|
||||||
- [ ] Monitor for failures
|
- [ ] Monitor for failures
|
||||||
- [ ] Compare results with shadow mode if needed
|
- [ ] Compare results with shadow mode if needed
|
||||||
|
|
||||||
@@ -132,30 +86,17 @@ server/reflector/hatchet/
|
|||||||
|
|
||||||
## Known Issues
|
## Known Issues
|
||||||
|
|
||||||
### Conductor
|
|
||||||
- See `CONDUCTOR_LLM_OBSERVATIONS.md` for debugging notes
|
|
||||||
- Ghost workers issue (multiple containers polling)
|
|
||||||
- Multiprocessing + AsyncIO conflicts
|
|
||||||
|
|
||||||
### Hatchet
|
### Hatchet
|
||||||
- See `HATCHET_LLM_OBSERVATIONS.md` for debugging notes
|
- See `HATCHET_LLM_OBSERVATIONS.md` for debugging notes
|
||||||
- SDK v1.21+ API changes (breaking)
|
- SDK v1.21+ API changes (breaking)
|
||||||
- JWT token Docker networking issues
|
- JWT token Docker networking issues
|
||||||
- Worker appears hung without debug mode
|
- Worker appears hung without debug mode
|
||||||
|
- Workflow replay is version-locked (use --force to run latest code)
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
||||||
## Quick Start
|
## Quick Start
|
||||||
|
|
||||||
### Conductor
|
|
||||||
```bash
|
|
||||||
# Start infrastructure
|
|
||||||
docker compose up -d conductor conductor-worker
|
|
||||||
|
|
||||||
# Register workflow
|
|
||||||
docker compose exec conductor-worker uv run python -m reflector.conductor.workflows.register
|
|
||||||
```
|
|
||||||
|
|
||||||
### Hatchet
|
### Hatchet
|
||||||
```bash
|
```bash
|
||||||
# Start infrastructure
|
# Start infrastructure
|
||||||
@@ -167,7 +108,7 @@ docker compose up -d hatchet hatchet-worker
|
|||||||
### Trigger Workflow
|
### Trigger Workflow
|
||||||
```bash
|
```bash
|
||||||
# Set provider in .env
|
# Set provider in .env
|
||||||
DURABLE_WORKFLOW_PROVIDER=hatchet # or conductor
|
DURABLE_WORKFLOW_PROVIDER=hatchet
|
||||||
|
|
||||||
# Process a Daily.co recording via webhook or API
|
# Process a Daily.co recording via webhook or API
|
||||||
# The pipeline trigger automatically uses the configured provider
|
# The pipeline trigger automatically uses the configured provider
|
||||||
|
|||||||
@@ -34,20 +34,6 @@ services:
|
|||||||
environment:
|
environment:
|
||||||
ENTRYPOINT: beat
|
ENTRYPOINT: beat
|
||||||
|
|
||||||
conductor-worker:
|
|
||||||
build:
|
|
||||||
context: server
|
|
||||||
volumes:
|
|
||||||
- ./server/:/app/
|
|
||||||
- /app/.venv
|
|
||||||
env_file:
|
|
||||||
- ./server/.env
|
|
||||||
environment:
|
|
||||||
ENTRYPOINT: conductor-worker
|
|
||||||
depends_on:
|
|
||||||
conductor:
|
|
||||||
condition: service_healthy
|
|
||||||
|
|
||||||
hatchet-worker:
|
hatchet-worker:
|
||||||
build:
|
build:
|
||||||
context: server
|
context: server
|
||||||
@@ -92,19 +78,6 @@ services:
|
|||||||
volumes:
|
volumes:
|
||||||
- ./data/postgres:/var/lib/postgresql/data
|
- ./data/postgres:/var/lib/postgresql/data
|
||||||
|
|
||||||
conductor:
|
|
||||||
image: conductoross/conductor-standalone:3.15.0
|
|
||||||
ports:
|
|
||||||
- "8180:8080"
|
|
||||||
- "5001:5000"
|
|
||||||
environment:
|
|
||||||
- conductor.db.type=memory
|
|
||||||
healthcheck:
|
|
||||||
test: ["CMD", "curl", "-f", "http://localhost:8080/health"]
|
|
||||||
interval: 30s
|
|
||||||
timeout: 10s
|
|
||||||
retries: 5
|
|
||||||
|
|
||||||
hatchet-postgres:
|
hatchet-postgres:
|
||||||
image: postgres:15.6
|
image: postgres:15.6
|
||||||
command: postgres -c 'max_connections=200'
|
command: postgres -c 'max_connections=200'
|
||||||
|
|||||||
@@ -1,345 +0,0 @@
|
|||||||
# Conductor OSS Migration - LLM Debugging Observations
|
|
||||||
|
|
||||||
This document captures hard-won debugging insights from migrating the multitrack diarization pipeline from Celery to Conductor OSS. These observations are particularly relevant for LLM assistants working on this codebase.
|
|
||||||
|
|
||||||
## Architecture Context
|
|
||||||
|
|
||||||
- **Conductor Python SDK** uses multiprocessing: 1 parent process spawns 15 `TaskRunner` subprocesses
|
|
||||||
- Each task type gets its own subprocess that polls Conductor server
|
|
||||||
- Workers are identified by container hostname (e.g., `595f5ddc9711`)
|
|
||||||
- Shadow mode (`CONDUCTOR_SHADOW_MODE=true`) runs both Celery and Conductor in parallel
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Challenge 1: Ghost Workers - Multiple Containers Polling Same Tasks
|
|
||||||
|
|
||||||
### Symptoms
|
|
||||||
- Tasks complete but with wrong/empty output
|
|
||||||
- Worker logs show no execution for a task that API shows as COMPLETED
|
|
||||||
- `workerId` in Conductor API doesn't match expected container
|
|
||||||
|
|
||||||
### Root Cause
|
|
||||||
Multiple containers may be running Conductor workers:
|
|
||||||
- `reflector-conductor-worker-1` (dedicated worker)
|
|
||||||
- `reflector-server-1` (if shadow mode enabled or worker code imported)
|
|
||||||
|
|
||||||
### Debugging Steps
|
|
||||||
```bash
|
|
||||||
# 1. Get the mystery worker ID from Conductor API
|
|
||||||
curl -s "http://localhost:8180/api/workflow/{id}" | jq '.tasks[] | {ref: .referenceTaskName, workerId}'
|
|
||||||
|
|
||||||
# 2. Find which container has that hostname
|
|
||||||
docker ps -a | grep {workerId}
|
|
||||||
# or
|
|
||||||
docker ps -a --format "{{.ID}} {{.Names}}" | grep {first-12-chars}
|
|
||||||
|
|
||||||
# 3. Check that container's code version
|
|
||||||
docker exec {container} cat /app/reflector/conductor/workers/{worker}.py | head -50
|
|
||||||
```
|
|
||||||
|
|
||||||
### Resolution
|
|
||||||
Restart ALL containers that might be polling Conductor tasks:
|
|
||||||
```bash
|
|
||||||
docker compose restart conductor-worker server
|
|
||||||
```
|
|
||||||
|
|
||||||
### Key Insight
|
|
||||||
**Always verify `workerId` matches your expected container.** In distributed worker setups, know ALL containers that poll for tasks.
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Challenge 2: Multiprocessing + AsyncIO + Database Conflicts
|
|
||||||
|
|
||||||
### Symptoms
|
|
||||||
```
|
|
||||||
InterfaceError: cannot perform operation: another operation is in progress
|
|
||||||
RuntimeError: Task <Task pending...> running at /app/.../worker.py
|
|
||||||
```
|
|
||||||
|
|
||||||
### Root Cause
|
|
||||||
Conductor Python SDK forks subprocesses. When subprocess calls `asyncio.run()`:
|
|
||||||
1. New event loop is created
|
|
||||||
2. But `get_database()` returns cached connection from parent process context
|
|
||||||
3. Parent's connection is incompatible with child's event loop
|
|
||||||
|
|
||||||
### Resolution
|
|
||||||
Reset context and create fresh connection in each subprocess:
|
|
||||||
```python
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
from reflector.db import _database_context
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
# Reset context var - don't inherit from parent
|
|
||||||
_database_context.set(None)
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
# ... rest of async code
|
|
||||||
```
|
|
||||||
|
|
||||||
### Key Insight
|
|
||||||
**Any singleton/cached resource (DB connections, S3 clients, HTTP sessions) must be recreated AFTER fork.** Never trust inherited state in multiprocessing workers.
|
|
||||||
|
|
||||||
### TODO: The Real Problem with get_database()
|
|
||||||
|
|
||||||
**Current solution is a hack.** The issue runs deeper than multiprocessing fork:
|
|
||||||
|
|
||||||
#### What's Actually Happening
|
|
||||||
1. Each Conductor subprocess calls `asyncio.run(_process())` repeatedly for each task
|
|
||||||
2. First `asyncio.run()`: creates DB connection, stores in ContextVar
|
|
||||||
3. First task completes, `asyncio.run()` exits, **event loop destroyed**
|
|
||||||
4. **But**: ContextVar still holds the connection reference (ContextVars persist across `asyncio.run()` calls)
|
|
||||||
5. Second `asyncio.run()`: Creates a **new event loop**
|
|
||||||
6. Code tries to use the **old connection** (from ContextVar) with the **new event loop**
|
|
||||||
7. Error: "another operation is in progress"
|
|
||||||
|
|
||||||
**Root issue**: `get_database()` as a global singleton is incompatible with repeated `asyncio.run()` calls in the same process.
|
|
||||||
|
|
||||||
#### Option 1: Explicit Connection Lifecycle (cleanest)
|
|
||||||
```python
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
# Don't use get_database() - create explicit connection
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
|
|
||||||
try:
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
# Problem: transcripts_controller.get_by_id() uses get_database() internally
|
|
||||||
# Would need to refactor controllers to accept db parameter
|
|
||||||
# e.g., await transcripts_controller.get_by_id(transcript_id, db=db)
|
|
||||||
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
```
|
|
||||||
|
|
||||||
**Pros**: Clean separation, explicit lifecycle
|
|
||||||
**Cons**: Requires refactoring all controller methods to accept `db` parameter
|
|
||||||
|
|
||||||
#### Option 2: Reset ContextVar Properly (pragmatic)
|
|
||||||
```python
|
|
||||||
async def _process():
|
|
||||||
from reflector.db import _database_context, get_database
|
|
||||||
|
|
||||||
# Ensure fresh connection per task
|
|
||||||
old_db = _database_context.get()
|
|
||||||
if old_db and old_db.is_connected:
|
|
||||||
await old_db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
|
|
||||||
# Now get_database() will create fresh connection
|
|
||||||
db = get_database()
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
try:
|
|
||||||
# ... work ...
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
```
|
|
||||||
|
|
||||||
**Pros**: Works with existing controller code
|
|
||||||
**Cons**: Still manipulating globals, cleanup needed in every worker
|
|
||||||
|
|
||||||
#### Option 3: Fix get_database() Itself (best long-term)
|
|
||||||
```python
|
|
||||||
# In reflector/db/__init__.py
|
|
||||||
def get_database() -> databases.Database:
|
|
||||||
"""Get database instance for current event loop"""
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
db = _database_context.get()
|
|
||||||
|
|
||||||
# Check if connection is valid for current event loop
|
|
||||||
if db is not None:
|
|
||||||
try:
|
|
||||||
loop = asyncio.get_running_loop()
|
|
||||||
# If connection's event loop differs, it's stale
|
|
||||||
if db._connection and hasattr(db._connection, '_loop'):
|
|
||||||
if db._connection._loop != loop:
|
|
||||||
# Stale connection from old event loop
|
|
||||||
db = None
|
|
||||||
except RuntimeError:
|
|
||||||
# No running loop
|
|
||||||
pass
|
|
||||||
|
|
||||||
if db is None:
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
|
|
||||||
return db
|
|
||||||
```
|
|
||||||
|
|
||||||
**Pros**: Fixes root cause, no changes needed in workers
|
|
||||||
**Cons**: Relies on implementation details of `databases` library
|
|
||||||
|
|
||||||
#### Recommendation
|
|
||||||
- **Short-term**: Option 2 (explicit cleanup in workers that need DB)
|
|
||||||
- **Long-term**: Option 1 (refactor to dependency injection) is the only architecturally clean solution
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Challenge 3: Type Mismatches Across Serialization Boundary
|
|
||||||
|
|
||||||
### Symptoms
|
|
||||||
```
|
|
||||||
ValidationError: 1 validation error for TranscriptTopic
|
|
||||||
transcript
|
|
||||||
Input should be a valid string [type=string_type, input_value={'translation': None, 'words': [...]}]
|
|
||||||
```
|
|
||||||
|
|
||||||
### Root Cause
|
|
||||||
Conductor JSON-serializes all task inputs/outputs. Complex Pydantic models get serialized to dicts:
|
|
||||||
- `TitleSummary.transcript: Transcript` becomes `{"translation": null, "words": [...]}`
|
|
||||||
- Next task expects `TranscriptTopic.transcript: str`
|
|
||||||
|
|
||||||
### Resolution
|
|
||||||
Explicitly reconstruct types when deserializing:
|
|
||||||
```python
|
|
||||||
from reflector.processors.types import TitleSummary, Transcript as TranscriptType, Word
|
|
||||||
|
|
||||||
def normalize_topic(t):
|
|
||||||
topic = dict(t)
|
|
||||||
transcript_data = topic.get("transcript")
|
|
||||||
if isinstance(transcript_data, dict):
|
|
||||||
words_list = transcript_data.get("words", [])
|
|
||||||
word_objects = [Word(**w) for w in words_list]
|
|
||||||
topic["transcript"] = TranscriptType(
|
|
||||||
words=word_objects,
|
|
||||||
translation=transcript_data.get("translation")
|
|
||||||
)
|
|
||||||
return topic
|
|
||||||
|
|
||||||
topic_objects = [TitleSummary(**normalize_topic(t)) for t in topics]
|
|
||||||
```
|
|
||||||
|
|
||||||
### Key Insight
|
|
||||||
**Conductor task I/O is always JSON.** Design workers to handle dict inputs and reconstruct domain objects explicitly.
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Challenge 4: Conductor Health Check Failures
|
|
||||||
|
|
||||||
### Symptoms
|
|
||||||
```
|
|
||||||
dependency failed to start: container reflector-conductor-1 is unhealthy
|
|
||||||
```
|
|
||||||
|
|
||||||
### Root Cause
|
|
||||||
Conductor OSS standalone container health endpoint can be slow/flaky, especially during startup or under load.
|
|
||||||
|
|
||||||
### Resolution
|
|
||||||
Bypass docker-compose health check dependency:
|
|
||||||
```bash
|
|
||||||
# Instead of: docker compose up -d conductor-worker
|
|
||||||
docker start reflector-conductor-worker-1
|
|
||||||
```
|
|
||||||
|
|
||||||
### Key Insight
|
|
||||||
For development, consider removing `depends_on.condition: service_healthy` or increasing health check timeout.
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Challenge 5: JOIN Task Output Format
|
|
||||||
|
|
||||||
### Symptoms
|
|
||||||
`merge_transcripts` receives data but outputs `word_count: 0`
|
|
||||||
|
|
||||||
### Root Cause
|
|
||||||
FORK_JOIN_DYNAMIC's JOIN task outputs a **dict keyed by task reference names**, not an array:
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"transcribe_track_0": {"words": [...], "track_index": 0},
|
|
||||||
"transcribe_track_1": {"words": [...], "track_index": 1}
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
### Resolution
|
|
||||||
Handle both dict and array inputs:
|
|
||||||
```python
|
|
||||||
transcripts = task.input_data.get("transcripts", [])
|
|
||||||
|
|
||||||
# Handle JOIN output (dict with task refs as keys)
|
|
||||||
if isinstance(transcripts, dict):
|
|
||||||
transcripts = list(transcripts.values())
|
|
||||||
|
|
||||||
for t in transcripts:
|
|
||||||
if isinstance(t, dict) and "words" in t:
|
|
||||||
all_words.extend(t["words"])
|
|
||||||
```
|
|
||||||
|
|
||||||
### Key Insight
|
|
||||||
**JOIN task output structure differs from FORK input.** Always log input types during debugging.
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Debugging Workflow
|
|
||||||
|
|
||||||
### 1. Add DEBUG Prints with Flush
|
|
||||||
Multiprocessing buffers stdout. Force immediate output:
|
|
||||||
```python
|
|
||||||
import sys
|
|
||||||
print("[DEBUG] worker entered", flush=True)
|
|
||||||
sys.stdout.flush()
|
|
||||||
```
|
|
||||||
|
|
||||||
### 2. Test Worker Functions Directly
|
|
||||||
Bypass Conductor entirely to verify logic:
|
|
||||||
```bash
|
|
||||||
docker compose exec conductor-worker uv run python -c "
|
|
||||||
from reflector.conductor.workers.merge_transcripts import merge_transcripts
|
|
||||||
from conductor.client.http.models import Task
|
|
||||||
|
|
||||||
mock_task = Task()
|
|
||||||
mock_task.input_data = {'transcripts': {...}, 'transcript_id': 'test'}
|
|
||||||
result = merge_transcripts(mock_task)
|
|
||||||
print(result.output_data)
|
|
||||||
"
|
|
||||||
```
|
|
||||||
|
|
||||||
### 3. Check Task Timing
|
|
||||||
Suspiciously fast completion (e.g., 10ms) indicates:
|
|
||||||
- Cached result from previous run
|
|
||||||
- Wrong worker processed it
|
|
||||||
- Task completed without actual execution
|
|
||||||
|
|
||||||
```bash
|
|
||||||
curl -s "http://localhost:8180/api/workflow/{id}" | \
|
|
||||||
jq '.tasks[] | {ref: .referenceTaskName, duration: (.endTime - .startTime)}'
|
|
||||||
```
|
|
||||||
|
|
||||||
### 4. Verify Container Code Version
|
|
||||||
```bash
|
|
||||||
docker compose exec conductor-worker cat /app/reflector/conductor/workers/{file}.py | head -50
|
|
||||||
```
|
|
||||||
|
|
||||||
### 5. Use Conductor Retry API
|
|
||||||
Retry from specific failed task without re-running entire workflow:
|
|
||||||
```bash
|
|
||||||
curl -X POST "http://localhost:8180/api/workflow/{id}/retry"
|
|
||||||
```
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Common Gotchas Summary
|
|
||||||
|
|
||||||
| Issue | Signal | Fix |
|
|
||||||
|-------|--------|-----|
|
|
||||||
| Wrong worker | `workerId` mismatch | Restart all worker containers |
|
|
||||||
| DB conflict | "another operation in progress" | Fresh DB connection per subprocess |
|
|
||||||
| Type mismatch | Pydantic validation error | Reconstruct objects from dicts |
|
|
||||||
| No logs | Task completes but no output | Check if different container processed |
|
|
||||||
| 0 results | JOIN output format | Convert dict.values() to list |
|
|
||||||
| Health check | Compose dependency fails | Use `docker start` directly |
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## Files Most Likely to Need Conductor-Specific Handling
|
|
||||||
|
|
||||||
- `server/reflector/conductor/workers/*.py` - All workers need multiprocessing-safe patterns
|
|
||||||
- `server/reflector/db/__init__.py` - Database singleton, needs context reset
|
|
||||||
- `server/reflector/conductor/workflows/*.json` - Workflow definitions, check input/output mappings
|
|
||||||
@@ -1,7 +1,7 @@
|
|||||||
"""add workflow_run_id to transcript
|
"""add workflow_run_id to transcript
|
||||||
|
|
||||||
Revision ID: 0f943fede0e0
|
Revision ID: 0f943fede0e0
|
||||||
Revises: a326252ac554
|
Revises: bbafedfa510c
|
||||||
Create Date: 2025-12-16 01:54:13.855106
|
Create Date: 2025-12-16 01:54:13.855106
|
||||||
|
|
||||||
"""
|
"""
|
||||||
@@ -13,7 +13,7 @@ from alembic import op
|
|||||||
|
|
||||||
# revision identifiers, used by Alembic.
|
# revision identifiers, used by Alembic.
|
||||||
revision: str = "0f943fede0e0"
|
revision: str = "0f943fede0e0"
|
||||||
down_revision: Union[str, None] = "a326252ac554"
|
down_revision: Union[str, None] = "bbafedfa510c"
|
||||||
branch_labels: Union[str, Sequence[str], None] = None
|
branch_labels: Union[str, Sequence[str], None] = None
|
||||||
depends_on: Union[str, Sequence[str], None] = None
|
depends_on: Union[str, Sequence[str], None] = None
|
||||||
|
|
||||||
|
|||||||
@@ -1,32 +0,0 @@
|
|||||||
"""add workflow_id to recording
|
|
||||||
|
|
||||||
Revision ID: a326252ac554
|
|
||||||
Revises: bbafedfa510c
|
|
||||||
Create Date: 2025-12-14 11:34:22.137910
|
|
||||||
|
|
||||||
"""
|
|
||||||
|
|
||||||
from typing import Sequence, Union
|
|
||||||
|
|
||||||
import sqlalchemy as sa
|
|
||||||
from alembic import op
|
|
||||||
|
|
||||||
# revision identifiers, used by Alembic.
|
|
||||||
revision: str = "a326252ac554"
|
|
||||||
down_revision: Union[str, None] = "bbafedfa510c"
|
|
||||||
branch_labels: Union[str, Sequence[str], None] = None
|
|
||||||
depends_on: Union[str, Sequence[str], None] = None
|
|
||||||
|
|
||||||
|
|
||||||
def upgrade() -> None:
|
|
||||||
with op.batch_alter_table("recording", schema=None) as batch_op:
|
|
||||||
batch_op.add_column(sa.Column("workflow_id", sa.String(), nullable=True))
|
|
||||||
batch_op.create_index(
|
|
||||||
"idx_recording_workflow_id", ["workflow_id"], unique=False
|
|
||||||
)
|
|
||||||
|
|
||||||
|
|
||||||
def downgrade() -> None:
|
|
||||||
with op.batch_alter_table("recording", schema=None) as batch_op:
|
|
||||||
batch_op.drop_index("idx_recording_workflow_id")
|
|
||||||
batch_op.drop_column("workflow_id")
|
|
||||||
@@ -39,7 +39,6 @@ dependencies = [
|
|||||||
"pytest-env>=1.1.5",
|
"pytest-env>=1.1.5",
|
||||||
"webvtt-py>=0.5.0",
|
"webvtt-py>=0.5.0",
|
||||||
"icalendar>=6.0.0",
|
"icalendar>=6.0.0",
|
||||||
"conductor-python>=1.2.3",
|
|
||||||
"hatchet-sdk>=0.47.0",
|
"hatchet-sdk>=0.47.0",
|
||||||
]
|
]
|
||||||
|
|
||||||
@@ -138,8 +137,6 @@ select = [
|
|||||||
"reflector/tools/**.py" = ["PLC0415"]
|
"reflector/tools/**.py" = ["PLC0415"]
|
||||||
"reflector/hatchet/run_workers.py" = ["PLC0415"]
|
"reflector/hatchet/run_workers.py" = ["PLC0415"]
|
||||||
"reflector/hatchet/workflows/**.py" = ["PLC0415"]
|
"reflector/hatchet/workflows/**.py" = ["PLC0415"]
|
||||||
"reflector/conductor/run_workers.py" = ["PLC0415"]
|
|
||||||
"reflector/conductor/workers/**.py" = ["PLC0415"]
|
|
||||||
"reflector/views/hatchet.py" = ["PLC0415"]
|
"reflector/views/hatchet.py" = ["PLC0415"]
|
||||||
"migrations/versions/**.py" = ["PLC0415"]
|
"migrations/versions/**.py" = ["PLC0415"]
|
||||||
"tests/**.py" = ["PLC0415"]
|
"tests/**.py" = ["PLC0415"]
|
||||||
|
|||||||
@@ -12,7 +12,6 @@ from reflector.events import subscribers_shutdown, subscribers_startup
|
|||||||
from reflector.logger import logger
|
from reflector.logger import logger
|
||||||
from reflector.metrics import metrics_init
|
from reflector.metrics import metrics_init
|
||||||
from reflector.settings import settings
|
from reflector.settings import settings
|
||||||
from reflector.views.conductor import router as conductor_router
|
|
||||||
from reflector.views.daily import router as daily_router
|
from reflector.views.daily import router as daily_router
|
||||||
from reflector.views.hatchet import router as hatchet_router
|
from reflector.views.hatchet import router as hatchet_router
|
||||||
from reflector.views.meetings import router as meetings_router
|
from reflector.views.meetings import router as meetings_router
|
||||||
@@ -100,7 +99,6 @@ app.include_router(user_ws_router, prefix="/v1")
|
|||||||
app.include_router(zulip_router, prefix="/v1")
|
app.include_router(zulip_router, prefix="/v1")
|
||||||
app.include_router(whereby_router, prefix="/v1")
|
app.include_router(whereby_router, prefix="/v1")
|
||||||
app.include_router(daily_router, prefix="/v1/daily")
|
app.include_router(daily_router, prefix="/v1/daily")
|
||||||
app.include_router(conductor_router, prefix="/v1")
|
|
||||||
app.include_router(hatchet_router, prefix="/v1")
|
app.include_router(hatchet_router, prefix="/v1")
|
||||||
add_pagination(app)
|
add_pagination(app)
|
||||||
|
|
||||||
|
|||||||
@@ -1,5 +0,0 @@
|
|||||||
"""Conductor workflow orchestration module."""
|
|
||||||
|
|
||||||
from reflector.conductor.client import ConductorClientManager
|
|
||||||
|
|
||||||
__all__ = ["ConductorClientManager"]
|
|
||||||
@@ -1,40 +0,0 @@
|
|||||||
"""Conductor Python client wrapper."""
|
|
||||||
|
|
||||||
from conductor.client.configuration.configuration import Configuration
|
|
||||||
from conductor.client.orkes_clients import OrkesClients
|
|
||||||
from conductor.client.workflow_client import WorkflowClient
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
class ConductorClientManager:
|
|
||||||
"""Singleton manager for Conductor client connections."""
|
|
||||||
|
|
||||||
_instance: OrkesClients | None = None
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def get_client(cls) -> WorkflowClient:
|
|
||||||
"""Get or create the workflow client."""
|
|
||||||
if cls._instance is None:
|
|
||||||
config = Configuration(
|
|
||||||
server_api_url=settings.CONDUCTOR_SERVER_URL,
|
|
||||||
debug=settings.CONDUCTOR_DEBUG,
|
|
||||||
)
|
|
||||||
cls._instance = OrkesClients(config)
|
|
||||||
return cls._instance.get_workflow_client()
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def start_workflow(cls, name: str, version: int, input_data: dict) -> str:
|
|
||||||
"""Start a workflow and return the workflow ID."""
|
|
||||||
client = cls.get_client()
|
|
||||||
return client.start_workflow_by_name(name, input_data, version=version)
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def get_workflow_status(cls, workflow_id: str) -> dict:
|
|
||||||
"""Get the current status of a workflow."""
|
|
||||||
client = cls.get_client()
|
|
||||||
return client.get_workflow(workflow_id, include_tasks=True)
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def reset(cls) -> None:
|
|
||||||
"""Reset the client instance (for testing)."""
|
|
||||||
cls._instance = None
|
|
||||||
@@ -1,103 +0,0 @@
|
|||||||
"""Progress event emission for Conductor workers."""
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
from typing import Literal
|
|
||||||
|
|
||||||
from reflector.db.transcripts import PipelineProgressData
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.ws_manager import get_ws_manager
|
|
||||||
|
|
||||||
# Step mapping for progress tracking
|
|
||||||
# Maps task names to their index in the pipeline
|
|
||||||
PIPELINE_STEPS = {
|
|
||||||
"get_recording": 1,
|
|
||||||
"get_participants": 2,
|
|
||||||
"pad_track": 3, # Fork tasks share same step
|
|
||||||
"mixdown_tracks": 4,
|
|
||||||
"generate_waveform": 5,
|
|
||||||
"transcribe_track": 6, # Fork tasks share same step
|
|
||||||
"merge_transcripts": 7,
|
|
||||||
"detect_topics": 8,
|
|
||||||
"generate_title": 9, # Fork tasks share same step
|
|
||||||
"generate_summary": 9, # Fork tasks share same step
|
|
||||||
"finalize": 10,
|
|
||||||
"cleanup_consent": 11,
|
|
||||||
"post_zulip": 12,
|
|
||||||
"send_webhook": 13,
|
|
||||||
}
|
|
||||||
|
|
||||||
TOTAL_STEPS = 13
|
|
||||||
|
|
||||||
|
|
||||||
async def _emit_progress_async(
|
|
||||||
transcript_id: str,
|
|
||||||
step: str,
|
|
||||||
status: Literal["pending", "in_progress", "completed", "failed"],
|
|
||||||
workflow_id: str | None = None,
|
|
||||||
) -> None:
|
|
||||||
"""Async implementation of progress emission."""
|
|
||||||
ws_manager = get_ws_manager()
|
|
||||||
step_index = PIPELINE_STEPS.get(step, 0)
|
|
||||||
|
|
||||||
data = PipelineProgressData(
|
|
||||||
workflow_id=workflow_id,
|
|
||||||
current_step=step,
|
|
||||||
step_index=step_index,
|
|
||||||
total_steps=TOTAL_STEPS,
|
|
||||||
step_status=status,
|
|
||||||
)
|
|
||||||
|
|
||||||
await ws_manager.send_json(
|
|
||||||
room_id=f"ts:{transcript_id}",
|
|
||||||
message={
|
|
||||||
"event": "PIPELINE_PROGRESS",
|
|
||||||
"data": data.model_dump(),
|
|
||||||
},
|
|
||||||
)
|
|
||||||
|
|
||||||
logger.debug(
|
|
||||||
"[Progress] Emitted",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
step=step,
|
|
||||||
status=status,
|
|
||||||
step_index=step_index,
|
|
||||||
)
|
|
||||||
|
|
||||||
|
|
||||||
def emit_progress(
|
|
||||||
transcript_id: str,
|
|
||||||
step: str,
|
|
||||||
status: Literal["pending", "in_progress", "completed", "failed"],
|
|
||||||
workflow_id: str | None = None,
|
|
||||||
) -> None:
|
|
||||||
"""Emit a pipeline progress event (sync wrapper for Conductor workers).
|
|
||||||
|
|
||||||
Args:
|
|
||||||
transcript_id: The transcript ID to emit progress for
|
|
||||||
step: The current step name (e.g., "transcribe_track")
|
|
||||||
status: The step status
|
|
||||||
workflow_id: Optional workflow ID
|
|
||||||
"""
|
|
||||||
try:
|
|
||||||
# Get or create event loop for sync context
|
|
||||||
try:
|
|
||||||
loop = asyncio.get_running_loop()
|
|
||||||
except RuntimeError:
|
|
||||||
loop = None
|
|
||||||
|
|
||||||
if loop is not None and loop.is_running():
|
|
||||||
# Already in async context, schedule the coroutine
|
|
||||||
asyncio.create_task(
|
|
||||||
_emit_progress_async(transcript_id, step, status, workflow_id)
|
|
||||||
)
|
|
||||||
else:
|
|
||||||
# Not in async context, run synchronously
|
|
||||||
asyncio.run(_emit_progress_async(transcript_id, step, status, workflow_id))
|
|
||||||
except Exception as e:
|
|
||||||
# Progress emission should never break the pipeline
|
|
||||||
logger.warning(
|
|
||||||
"[Progress] Failed to emit progress event",
|
|
||||||
error=str(e),
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
step=step,
|
|
||||||
)
|
|
||||||
@@ -1,58 +0,0 @@
|
|||||||
"""
|
|
||||||
Run Conductor workers for the diarization pipeline.
|
|
||||||
|
|
||||||
Usage:
|
|
||||||
uv run -m reflector.conductor.run_workers
|
|
||||||
|
|
||||||
# Or via docker:
|
|
||||||
docker compose exec server uv run -m reflector.conductor.run_workers
|
|
||||||
"""
|
|
||||||
|
|
||||||
import signal
|
|
||||||
import sys
|
|
||||||
import time
|
|
||||||
|
|
||||||
from conductor.client.automator.task_handler import TaskHandler
|
|
||||||
from conductor.client.configuration.configuration import Configuration
|
|
||||||
from reflector.conductor import workers # noqa: F401 - registers workers via decorators
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
def main() -> None:
|
|
||||||
"""Start Conductor worker polling."""
|
|
||||||
if not settings.CONDUCTOR_ENABLED:
|
|
||||||
logger.error("CONDUCTOR_ENABLED is False, not starting workers")
|
|
||||||
sys.exit(1)
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"Starting Conductor workers",
|
|
||||||
server_url=settings.CONDUCTOR_SERVER_URL,
|
|
||||||
)
|
|
||||||
|
|
||||||
config = Configuration(
|
|
||||||
server_api_url=settings.CONDUCTOR_SERVER_URL,
|
|
||||||
debug=settings.CONDUCTOR_DEBUG,
|
|
||||||
)
|
|
||||||
|
|
||||||
task_handler = TaskHandler(configuration=config)
|
|
||||||
|
|
||||||
# Handle graceful shutdown
|
|
||||||
def shutdown_handler(signum: int, frame) -> None:
|
|
||||||
logger.info("Received shutdown signal, stopping workers...")
|
|
||||||
task_handler.stop_processes()
|
|
||||||
sys.exit(0)
|
|
||||||
|
|
||||||
signal.signal(signal.SIGINT, shutdown_handler)
|
|
||||||
signal.signal(signal.SIGTERM, shutdown_handler)
|
|
||||||
|
|
||||||
logger.info("Starting task polling...")
|
|
||||||
task_handler.start_processes()
|
|
||||||
|
|
||||||
# Keep alive
|
|
||||||
while True:
|
|
||||||
time.sleep(1)
|
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
|
||||||
main()
|
|
||||||
@@ -1,207 +0,0 @@
|
|||||||
"""Shadow mode comparison for Celery vs Conductor pipeline results."""
|
|
||||||
|
|
||||||
from dataclasses import dataclass
|
|
||||||
from typing import Any
|
|
||||||
|
|
||||||
from reflector.conductor.client import ConductorClientManager
|
|
||||||
from reflector.db.transcripts import Transcript, transcripts_controller
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@dataclass
|
|
||||||
class FieldDifference:
|
|
||||||
"""A difference between Celery and Conductor field values."""
|
|
||||||
|
|
||||||
field: str
|
|
||||||
celery_value: Any
|
|
||||||
conductor_value: Any
|
|
||||||
|
|
||||||
|
|
||||||
@dataclass
|
|
||||||
class ComparisonResult:
|
|
||||||
"""Result of comparing Celery and Conductor outputs."""
|
|
||||||
|
|
||||||
match: bool
|
|
||||||
differences: list[FieldDifference]
|
|
||||||
celery_status: str
|
|
||||||
conductor_status: str
|
|
||||||
error: str | None = None
|
|
||||||
|
|
||||||
|
|
||||||
async def compare_content_results(
|
|
||||||
recording_id: str, workflow_id: str
|
|
||||||
) -> ComparisonResult:
|
|
||||||
"""
|
|
||||||
Compare content results from Celery and Conductor pipelines.
|
|
||||||
|
|
||||||
Args:
|
|
||||||
recording_id: Recording ID to look up Celery transcript
|
|
||||||
workflow_id: Conductor workflow ID to get workflow output
|
|
||||||
|
|
||||||
Returns:
|
|
||||||
ComparisonResult with match status and any differences
|
|
||||||
"""
|
|
||||||
try:
|
|
||||||
# Get Celery result from DB
|
|
||||||
celery_transcript = await transcripts_controller.get_by_recording_id(
|
|
||||||
recording_id
|
|
||||||
)
|
|
||||||
if not celery_transcript:
|
|
||||||
return ComparisonResult(
|
|
||||||
match=False,
|
|
||||||
differences=[],
|
|
||||||
celery_status="not_found",
|
|
||||||
conductor_status="unknown",
|
|
||||||
error=f"No transcript found for recording_id={recording_id}",
|
|
||||||
)
|
|
||||||
|
|
||||||
# Get Conductor workflow status
|
|
||||||
workflow_status = ConductorClientManager.get_workflow_status(workflow_id)
|
|
||||||
conductor_status = workflow_status.status if workflow_status else "unknown"
|
|
||||||
|
|
||||||
# If workflow not completed, can't compare
|
|
||||||
if conductor_status != "COMPLETED":
|
|
||||||
return ComparisonResult(
|
|
||||||
match=False,
|
|
||||||
differences=[],
|
|
||||||
celery_status=celery_transcript.status,
|
|
||||||
conductor_status=conductor_status,
|
|
||||||
error=f"Conductor workflow not completed: {conductor_status}",
|
|
||||||
)
|
|
||||||
|
|
||||||
# Extract output from workflow
|
|
||||||
workflow_output = (
|
|
||||||
workflow_status.output if hasattr(workflow_status, "output") else {}
|
|
||||||
)
|
|
||||||
|
|
||||||
differences = _compare_fields(celery_transcript, workflow_output)
|
|
||||||
|
|
||||||
result = ComparisonResult(
|
|
||||||
match=len(differences) == 0,
|
|
||||||
differences=differences,
|
|
||||||
celery_status=celery_transcript.status,
|
|
||||||
conductor_status=conductor_status,
|
|
||||||
)
|
|
||||||
|
|
||||||
# Log comparison result
|
|
||||||
if result.match:
|
|
||||||
logger.info(
|
|
||||||
"Shadow mode comparison: MATCH",
|
|
||||||
recording_id=recording_id,
|
|
||||||
workflow_id=workflow_id,
|
|
||||||
)
|
|
||||||
else:
|
|
||||||
logger.warning(
|
|
||||||
"Shadow mode comparison: MISMATCH",
|
|
||||||
recording_id=recording_id,
|
|
||||||
workflow_id=workflow_id,
|
|
||||||
differences=[
|
|
||||||
{
|
|
||||||
"field": d.field,
|
|
||||||
"celery": d.celery_value,
|
|
||||||
"conductor": d.conductor_value,
|
|
||||||
}
|
|
||||||
for d in differences
|
|
||||||
],
|
|
||||||
)
|
|
||||||
|
|
||||||
return result
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error(
|
|
||||||
"Shadow mode comparison failed",
|
|
||||||
recording_id=recording_id,
|
|
||||||
workflow_id=workflow_id,
|
|
||||||
error=str(e),
|
|
||||||
exc_info=True,
|
|
||||||
)
|
|
||||||
return ComparisonResult(
|
|
||||||
match=False,
|
|
||||||
differences=[],
|
|
||||||
celery_status="unknown",
|
|
||||||
conductor_status="unknown",
|
|
||||||
error=str(e),
|
|
||||||
)
|
|
||||||
|
|
||||||
|
|
||||||
def _compare_fields(
|
|
||||||
celery_transcript: Transcript, workflow_output: dict
|
|
||||||
) -> list[FieldDifference]:
|
|
||||||
"""Compare specific content fields between Celery and Conductor."""
|
|
||||||
differences = []
|
|
||||||
|
|
||||||
# Compare title
|
|
||||||
conductor_title = workflow_output.get("title")
|
|
||||||
if celery_transcript.title != conductor_title:
|
|
||||||
differences.append(
|
|
||||||
FieldDifference(
|
|
||||||
field="title",
|
|
||||||
celery_value=celery_transcript.title,
|
|
||||||
conductor_value=conductor_title,
|
|
||||||
)
|
|
||||||
)
|
|
||||||
|
|
||||||
# Compare short_summary
|
|
||||||
conductor_short_summary = workflow_output.get("short_summary")
|
|
||||||
if celery_transcript.short_summary != conductor_short_summary:
|
|
||||||
differences.append(
|
|
||||||
FieldDifference(
|
|
||||||
field="short_summary",
|
|
||||||
celery_value=celery_transcript.short_summary,
|
|
||||||
conductor_value=conductor_short_summary,
|
|
||||||
)
|
|
||||||
)
|
|
||||||
|
|
||||||
# Compare long_summary
|
|
||||||
conductor_long_summary = workflow_output.get("summary")
|
|
||||||
if celery_transcript.long_summary != conductor_long_summary:
|
|
||||||
differences.append(
|
|
||||||
FieldDifference(
|
|
||||||
field="long_summary",
|
|
||||||
celery_value=celery_transcript.long_summary,
|
|
||||||
conductor_value=conductor_long_summary,
|
|
||||||
)
|
|
||||||
)
|
|
||||||
|
|
||||||
# Compare topic count
|
|
||||||
celery_topics = celery_transcript.topics or []
|
|
||||||
conductor_topics = workflow_output.get("topics", [])
|
|
||||||
if len(celery_topics) != len(conductor_topics):
|
|
||||||
differences.append(
|
|
||||||
FieldDifference(
|
|
||||||
field="topic_count",
|
|
||||||
celery_value=len(celery_topics),
|
|
||||||
conductor_value=len(conductor_topics),
|
|
||||||
)
|
|
||||||
)
|
|
||||||
|
|
||||||
# Compare word count from events
|
|
||||||
celery_events = celery_transcript.events or {}
|
|
||||||
celery_words = (
|
|
||||||
celery_events.get("words", []) if isinstance(celery_events, dict) else []
|
|
||||||
)
|
|
||||||
conductor_words = workflow_output.get("all_words", [])
|
|
||||||
if len(celery_words) != len(conductor_words):
|
|
||||||
differences.append(
|
|
||||||
FieldDifference(
|
|
||||||
field="word_count",
|
|
||||||
celery_value=len(celery_words),
|
|
||||||
conductor_value=len(conductor_words),
|
|
||||||
)
|
|
||||||
)
|
|
||||||
|
|
||||||
# Compare duration
|
|
||||||
conductor_duration = workflow_output.get("duration")
|
|
||||||
if (
|
|
||||||
conductor_duration is not None
|
|
||||||
and celery_transcript.duration != conductor_duration
|
|
||||||
):
|
|
||||||
differences.append(
|
|
||||||
FieldDifference(
|
|
||||||
field="duration",
|
|
||||||
celery_value=celery_transcript.duration,
|
|
||||||
conductor_value=conductor_duration,
|
|
||||||
)
|
|
||||||
)
|
|
||||||
|
|
||||||
return differences
|
|
||||||
@@ -1,6 +0,0 @@
|
|||||||
"""Conductor task definitions module."""
|
|
||||||
|
|
||||||
from reflector.conductor.tasks.definitions import TASK_DEFINITIONS
|
|
||||||
from reflector.conductor.tasks.register import register_task_definitions
|
|
||||||
|
|
||||||
__all__ = ["TASK_DEFINITIONS", "register_task_definitions"]
|
|
||||||
@@ -1,161 +0,0 @@
|
|||||||
"""Task definitions for Conductor workflow orchestration.
|
|
||||||
|
|
||||||
Timeout reference (from CONDUCTOR_MIGRATION_REQUIREMENTS.md):
|
|
||||||
| Task | Timeout (s) | Response Timeout (s) | Retry Count |
|
|
||||||
|-------------------|-------------|----------------------|-------------|
|
|
||||||
| get_recording | 60 | 30 | 3 |
|
|
||||||
| get_participants | 60 | 30 | 3 |
|
|
||||||
| pad_track | 300 | 120 | 3 |
|
|
||||||
| mixdown_tracks | 600 | 300 | 3 |
|
|
||||||
| generate_waveform | 120 | 60 | 3 |
|
|
||||||
| transcribe_track | 1800 | 900 | 3 |
|
|
||||||
| merge_transcripts | 60 | 30 | 3 |
|
|
||||||
| detect_topics | 300 | 120 | 3 |
|
|
||||||
| generate_title | 60 | 30 | 3 |
|
|
||||||
| generate_summary | 300 | 120 | 3 |
|
|
||||||
| finalize | 60 | 30 | 3 |
|
|
||||||
| cleanup_consent | 60 | 30 | 3 |
|
|
||||||
| post_zulip | 60 | 30 | 5 |
|
|
||||||
| send_webhook | 60 | 30 | 30 |
|
|
||||||
"""
|
|
||||||
|
|
||||||
OWNER_EMAIL = "reflector@example.com"
|
|
||||||
|
|
||||||
TASK_DEFINITIONS = [
|
|
||||||
{
|
|
||||||
"name": "get_recording",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["recording_id"],
|
|
||||||
"outputKeys": ["id", "mtg_session_id", "room_name", "duration"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "get_participants",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["mtg_session_id"],
|
|
||||||
"outputKeys": ["participants"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "pad_track",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 300,
|
|
||||||
"responseTimeoutSeconds": 120,
|
|
||||||
"inputKeys": ["track_index", "s3_key", "bucket_name", "transcript_id"],
|
|
||||||
"outputKeys": ["padded_url", "size", "track_index"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "mixdown_tracks",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 600,
|
|
||||||
"responseTimeoutSeconds": 300,
|
|
||||||
"inputKeys": ["padded_urls", "transcript_id"],
|
|
||||||
"outputKeys": ["audio_key", "duration", "size"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_waveform",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 120,
|
|
||||||
"responseTimeoutSeconds": 60,
|
|
||||||
"inputKeys": ["audio_key", "transcript_id"],
|
|
||||||
"outputKeys": ["waveform"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "transcribe_track",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 1800,
|
|
||||||
"responseTimeoutSeconds": 900,
|
|
||||||
"inputKeys": ["track_index", "audio_url", "language"],
|
|
||||||
"outputKeys": ["words", "track_index"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "merge_transcripts",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["transcripts", "transcript_id"],
|
|
||||||
"outputKeys": ["all_words", "word_count"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "detect_topics",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 300,
|
|
||||||
"responseTimeoutSeconds": 120,
|
|
||||||
"inputKeys": ["words", "transcript_id", "target_language"],
|
|
||||||
"outputKeys": ["topics"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_title",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["topics", "transcript_id"],
|
|
||||||
"outputKeys": ["title"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_summary",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 300,
|
|
||||||
"responseTimeoutSeconds": 120,
|
|
||||||
"inputKeys": ["words", "topics", "transcript_id"],
|
|
||||||
"outputKeys": ["summary", "short_summary"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "finalize",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["transcript_id", "title", "summary", "short_summary", "duration"],
|
|
||||||
"outputKeys": ["status"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "cleanup_consent",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["transcript_id"],
|
|
||||||
"outputKeys": ["audio_deleted", "reason"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "post_zulip",
|
|
||||||
"retryCount": 5,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["transcript_id"],
|
|
||||||
"outputKeys": ["message_id"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "send_webhook",
|
|
||||||
"retryCount": 30,
|
|
||||||
"timeoutSeconds": 60,
|
|
||||||
"responseTimeoutSeconds": 30,
|
|
||||||
"inputKeys": ["transcript_id", "room_id"],
|
|
||||||
"outputKeys": ["sent", "status_code"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_dynamic_fork_tasks",
|
|
||||||
"retryCount": 3,
|
|
||||||
"timeoutSeconds": 30,
|
|
||||||
"responseTimeoutSeconds": 15,
|
|
||||||
"inputKeys": ["tracks", "task_type", "transcript_id", "bucket_name"],
|
|
||||||
"outputKeys": ["tasks", "inputs"],
|
|
||||||
"ownerEmail": OWNER_EMAIL,
|
|
||||||
"description": "Helper task to generate dynamic fork structure for variable track counts",
|
|
||||||
},
|
|
||||||
]
|
|
||||||
@@ -1,60 +0,0 @@
|
|||||||
"""Register task definitions with Conductor server."""
|
|
||||||
|
|
||||||
import httpx
|
|
||||||
|
|
||||||
from reflector.conductor.tasks.definitions import TASK_DEFINITIONS
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
def register_task_definitions() -> None:
|
|
||||||
"""Register all task definitions with Conductor server.
|
|
||||||
|
|
||||||
Raises:
|
|
||||||
httpx.HTTPStatusError: If registration fails.
|
|
||||||
"""
|
|
||||||
base_url = settings.CONDUCTOR_SERVER_URL.rstrip("/")
|
|
||||||
url = f"{base_url}/metadata/taskdefs"
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"Registering task definitions",
|
|
||||||
count=len(TASK_DEFINITIONS),
|
|
||||||
url=url,
|
|
||||||
)
|
|
||||||
|
|
||||||
with httpx.Client(timeout=30.0) as client:
|
|
||||||
resp = client.post(
|
|
||||||
url,
|
|
||||||
json=TASK_DEFINITIONS,
|
|
||||||
headers={"Content-Type": "application/json"},
|
|
||||||
)
|
|
||||||
resp.raise_for_status()
|
|
||||||
|
|
||||||
logger.info("Task definitions registered successfully")
|
|
||||||
|
|
||||||
|
|
||||||
async def register_task_definitions_async() -> None:
|
|
||||||
"""Async version of register_task_definitions."""
|
|
||||||
base_url = settings.CONDUCTOR_SERVER_URL.rstrip("/")
|
|
||||||
url = f"{base_url}/metadata/taskdefs"
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"Registering task definitions",
|
|
||||||
count=len(TASK_DEFINITIONS),
|
|
||||||
url=url,
|
|
||||||
)
|
|
||||||
|
|
||||||
async with httpx.AsyncClient(timeout=30.0) as client:
|
|
||||||
resp = await client.post(
|
|
||||||
url,
|
|
||||||
json=TASK_DEFINITIONS,
|
|
||||||
headers={"Content-Type": "application/json"},
|
|
||||||
)
|
|
||||||
resp.raise_for_status()
|
|
||||||
|
|
||||||
logger.info("Task definitions registered successfully")
|
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
|
||||||
register_task_definitions()
|
|
||||||
print(f"Registered {len(TASK_DEFINITIONS)} task definitions")
|
|
||||||
@@ -1,37 +0,0 @@
|
|||||||
"""Conductor workers for the diarization pipeline."""
|
|
||||||
|
|
||||||
from reflector.conductor.workers.cleanup_consent import cleanup_consent
|
|
||||||
from reflector.conductor.workers.detect_topics import detect_topics
|
|
||||||
from reflector.conductor.workers.finalize import finalize
|
|
||||||
from reflector.conductor.workers.generate_dynamic_fork_tasks import (
|
|
||||||
generate_dynamic_fork_tasks,
|
|
||||||
)
|
|
||||||
from reflector.conductor.workers.generate_summary import generate_summary
|
|
||||||
from reflector.conductor.workers.generate_title import generate_title
|
|
||||||
from reflector.conductor.workers.generate_waveform import generate_waveform
|
|
||||||
from reflector.conductor.workers.get_participants import get_participants
|
|
||||||
from reflector.conductor.workers.get_recording import get_recording
|
|
||||||
from reflector.conductor.workers.merge_transcripts import merge_transcripts
|
|
||||||
from reflector.conductor.workers.mixdown_tracks import mixdown_tracks
|
|
||||||
from reflector.conductor.workers.pad_track import pad_track
|
|
||||||
from reflector.conductor.workers.post_zulip import post_zulip
|
|
||||||
from reflector.conductor.workers.send_webhook import send_webhook
|
|
||||||
from reflector.conductor.workers.transcribe_track import transcribe_track
|
|
||||||
|
|
||||||
__all__ = [
|
|
||||||
"get_recording",
|
|
||||||
"get_participants",
|
|
||||||
"pad_track",
|
|
||||||
"mixdown_tracks",
|
|
||||||
"generate_waveform",
|
|
||||||
"transcribe_track",
|
|
||||||
"merge_transcripts",
|
|
||||||
"detect_topics",
|
|
||||||
"generate_title",
|
|
||||||
"generate_summary",
|
|
||||||
"finalize",
|
|
||||||
"cleanup_consent",
|
|
||||||
"post_zulip",
|
|
||||||
"send_webhook",
|
|
||||||
"generate_dynamic_fork_tasks",
|
|
||||||
]
|
|
||||||
@@ -1,126 +0,0 @@
|
|||||||
"""Conductor worker: cleanup_consent - Check consent and delete audio if denied."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="cleanup_consent")
|
|
||||||
def cleanup_consent(task: Task) -> TaskResult:
|
|
||||||
"""Check participant consent and delete audio if denied.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
audio_deleted: bool - Whether audio was deleted
|
|
||||||
reason: str | None - Reason for deletion
|
|
||||||
"""
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info("[Worker] cleanup_consent", transcript_id=transcript_id)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "cleanup_consent", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
|
|
||||||
from reflector.db import _database_context
|
|
||||||
from reflector.db.transcripts import transcripts_controller
|
|
||||||
from reflector.settings import settings
|
|
||||||
from reflector.storage import get_transcripts_storage
|
|
||||||
|
|
||||||
# Create fresh database connection for subprocess (not shared from parent)
|
|
||||||
_database_context.set(None)
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
try:
|
|
||||||
transcript = await transcripts_controller.get_by_id(transcript_id)
|
|
||||||
if transcript is None:
|
|
||||||
raise ValueError(f"Transcript {transcript_id} not found in database")
|
|
||||||
|
|
||||||
# Check if any participant denied consent
|
|
||||||
# This mirrors the logic from main_live_pipeline.task_cleanup_consent
|
|
||||||
audio_deleted = False
|
|
||||||
reason = None
|
|
||||||
|
|
||||||
if transcript.participants:
|
|
||||||
for p in transcript.participants:
|
|
||||||
if hasattr(p, "consent") and p.consent == "denied":
|
|
||||||
audio_deleted = True
|
|
||||||
reason = f"Participant {p.name or p.id} denied consent"
|
|
||||||
break
|
|
||||||
|
|
||||||
if audio_deleted:
|
|
||||||
storage = get_transcripts_storage()
|
|
||||||
audio_key = f"{transcript_id}/audio.mp3"
|
|
||||||
try:
|
|
||||||
await storage.delete_file(audio_key)
|
|
||||||
await transcripts_controller.update(
|
|
||||||
transcript, {"audio_deleted": True}
|
|
||||||
)
|
|
||||||
logger.info(
|
|
||||||
"[Worker] cleanup_consent: audio deleted",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
reason=reason,
|
|
||||||
)
|
|
||||||
except Exception as e:
|
|
||||||
logger.warning(
|
|
||||||
"[Worker] cleanup_consent: failed to delete audio",
|
|
||||||
error=str(e),
|
|
||||||
)
|
|
||||||
|
|
||||||
return audio_deleted, reason
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
|
|
||||||
try:
|
|
||||||
audio_deleted, reason = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"audio_deleted": audio_deleted,
|
|
||||||
"reason": reason,
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] cleanup_consent complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
audio_deleted=audio_deleted,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "cleanup_consent", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] cleanup_consent failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "cleanup_consent", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,93 +0,0 @@
|
|||||||
"""Conductor worker: detect_topics - Detect topics using LLM."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="detect_topics")
|
|
||||||
def detect_topics(task: Task) -> TaskResult:
|
|
||||||
"""Detect topics using LLM.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
words: list[dict] - Transcribed words
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
target_language: str - Target language code (default: "en")
|
|
||||||
|
|
||||||
Output:
|
|
||||||
topics: list[dict] - Detected topics
|
|
||||||
"""
|
|
||||||
words = task.input_data.get("words", [])
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
target_language = task.input_data.get("target_language", "en")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] detect_topics",
|
|
||||||
word_count=len(words),
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "detect_topics", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
from reflector.pipelines import topic_processing
|
|
||||||
from reflector.processors.types import Transcript as TranscriptType
|
|
||||||
from reflector.processors.types import Word
|
|
||||||
|
|
||||||
# Convert word dicts to Word objects
|
|
||||||
word_objects = [Word(**w) for w in words]
|
|
||||||
transcript = TranscriptType(words=word_objects)
|
|
||||||
|
|
||||||
empty_pipeline = topic_processing.EmptyPipeline(logger=logger)
|
|
||||||
|
|
||||||
async def noop_callback(t):
|
|
||||||
pass
|
|
||||||
|
|
||||||
topics = await topic_processing.detect_topics(
|
|
||||||
transcript,
|
|
||||||
target_language,
|
|
||||||
on_topic_callback=noop_callback,
|
|
||||||
empty_pipeline=empty_pipeline,
|
|
||||||
)
|
|
||||||
|
|
||||||
return [t.model_dump() for t in topics]
|
|
||||||
|
|
||||||
try:
|
|
||||||
topics = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"topics": topics}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] detect_topics complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
topic_count=len(topics),
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "detect_topics", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] detect_topics failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "detect_topics", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,111 +0,0 @@
|
|||||||
"""Conductor worker: finalize - Finalize transcript status and update database."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="finalize")
|
|
||||||
def finalize(task: Task) -> TaskResult:
|
|
||||||
"""Finalize the transcript status and update the database.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
title: str - Generated title
|
|
||||||
summary: str - Long summary
|
|
||||||
short_summary: str - Short summary
|
|
||||||
duration: float - Audio duration
|
|
||||||
|
|
||||||
Output:
|
|
||||||
status: str - "COMPLETED"
|
|
||||||
"""
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
title = task.input_data.get("title", "")
|
|
||||||
summary = task.input_data.get("summary", "")
|
|
||||||
short_summary = task.input_data.get("short_summary", "")
|
|
||||||
duration = task.input_data.get("duration", 0)
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] finalize",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
title=title,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "finalize", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
|
|
||||||
from reflector.db import _database_context
|
|
||||||
from reflector.db.transcripts import transcripts_controller
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
# Create fresh database connection for subprocess (not shared from parent)
|
|
||||||
_database_context.set(None)
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
try:
|
|
||||||
transcript = await transcripts_controller.get_by_id(transcript_id)
|
|
||||||
if transcript is None:
|
|
||||||
raise ValueError(f"Transcript {transcript_id} not found in database")
|
|
||||||
|
|
||||||
await transcripts_controller.update(
|
|
||||||
transcript,
|
|
||||||
{
|
|
||||||
"status": "ended",
|
|
||||||
"title": title,
|
|
||||||
"long_summary": summary,
|
|
||||||
"short_summary": short_summary,
|
|
||||||
"duration": duration,
|
|
||||||
},
|
|
||||||
)
|
|
||||||
return True
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
|
|
||||||
try:
|
|
||||||
asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"status": "COMPLETED"}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] finalize complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "finalize", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] finalize failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "finalize", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,110 +0,0 @@
|
|||||||
"""Conductor worker: generate_dynamic_fork_tasks - Helper for FORK_JOIN_DYNAMIC."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="generate_dynamic_fork_tasks")
|
|
||||||
def generate_dynamic_fork_tasks(task: Task) -> TaskResult:
|
|
||||||
"""Generate dynamic fork task structure for variable track counts.
|
|
||||||
|
|
||||||
This helper task generates the task definitions and inputs needed for
|
|
||||||
FORK_JOIN_DYNAMIC to process N tracks in parallel.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
tracks: list[dict] - List of track info with s3_key
|
|
||||||
task_type: str - Either "pad_track" or "transcribe_track"
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
bucket_name: str - S3 bucket name (for pad_track)
|
|
||||||
padded_urls: list[dict] - Padded track outputs (for transcribe_track)
|
|
||||||
|
|
||||||
Output:
|
|
||||||
tasks: list[dict] - Task definitions for dynamic fork
|
|
||||||
inputs: dict - Input parameters keyed by task reference name
|
|
||||||
"""
|
|
||||||
tracks = task.input_data.get("tracks", [])
|
|
||||||
task_type = task.input_data.get("task_type")
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
bucket_name = task.input_data.get("bucket_name")
|
|
||||||
padded_urls = task.input_data.get("padded_urls", {})
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_dynamic_fork_tasks",
|
|
||||||
task_type=task_type,
|
|
||||||
track_count=len(tracks),
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not tracks or not task_type:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing tracks or task_type"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
try:
|
|
||||||
tasks = []
|
|
||||||
inputs = {}
|
|
||||||
|
|
||||||
for idx, track in enumerate(tracks):
|
|
||||||
ref_name = f"{task_type}_{idx}"
|
|
||||||
|
|
||||||
# Task definition
|
|
||||||
tasks.append(
|
|
||||||
{
|
|
||||||
"name": task_type,
|
|
||||||
"taskReferenceName": ref_name,
|
|
||||||
"type": "SIMPLE",
|
|
||||||
}
|
|
||||||
)
|
|
||||||
|
|
||||||
# Task input based on type
|
|
||||||
if task_type == "pad_track":
|
|
||||||
inputs[ref_name] = {
|
|
||||||
"track_index": idx,
|
|
||||||
"s3_key": track.get("s3_key"),
|
|
||||||
"bucket_name": bucket_name,
|
|
||||||
"transcript_id": transcript_id,
|
|
||||||
}
|
|
||||||
elif task_type == "transcribe_track":
|
|
||||||
# Get padded URL from previous fork join output
|
|
||||||
padded_url = None
|
|
||||||
if isinstance(padded_urls, dict):
|
|
||||||
# Try to get from join output structure
|
|
||||||
pad_ref = f"pad_track_{idx}"
|
|
||||||
if pad_ref in padded_urls:
|
|
||||||
padded_url = padded_urls[pad_ref].get("padded_url")
|
|
||||||
elif "padded_url" in padded_urls:
|
|
||||||
# Single track case
|
|
||||||
padded_url = padded_urls.get("padded_url")
|
|
||||||
|
|
||||||
inputs[ref_name] = {
|
|
||||||
"track_index": idx,
|
|
||||||
"audio_url": padded_url,
|
|
||||||
"language": "en",
|
|
||||||
"transcript_id": transcript_id,
|
|
||||||
}
|
|
||||||
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"tasks": tasks,
|
|
||||||
"inputs": inputs,
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_dynamic_fork_tasks complete",
|
|
||||||
task_type=task_type,
|
|
||||||
task_count=len(tasks),
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] generate_dynamic_fork_tasks failed", error=str(e))
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,150 +0,0 @@
|
|||||||
"""Conductor worker: generate_summary - Generate meeting summaries using LLM."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="generate_summary")
|
|
||||||
def generate_summary(task: Task) -> TaskResult:
|
|
||||||
"""Generate long and short summaries from topics and words using LLM.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
words: list[dict] - Transcribed words
|
|
||||||
topics: list[dict] - Detected topics
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
summary: str - Long summary
|
|
||||||
short_summary: str - Short summary
|
|
||||||
"""
|
|
||||||
words = task.input_data.get("words", [])
|
|
||||||
topics = task.input_data.get("topics", [])
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_summary",
|
|
||||||
word_count=len(words),
|
|
||||||
topic_count=len(topics),
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_summary", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
|
|
||||||
from reflector.db import _database_context
|
|
||||||
from reflector.db.transcripts import transcripts_controller
|
|
||||||
from reflector.pipelines import topic_processing
|
|
||||||
from reflector.processors.types import TitleSummary, Word
|
|
||||||
from reflector.processors.types import Transcript as TranscriptType
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
# Create fresh database connection for subprocess (not shared from parent)
|
|
||||||
# Reset context var to ensure we get a fresh connection
|
|
||||||
_database_context.set(None)
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
try:
|
|
||||||
# detect_topics returns TitleSummary objects (with transcript: Transcript)
|
|
||||||
# When serialized, transcript becomes {translation, words} dict
|
|
||||||
# We need to reconstruct TitleSummary objects with proper Transcript
|
|
||||||
def normalize_topic(t):
|
|
||||||
topic = dict(t)
|
|
||||||
transcript_data = topic.get("transcript")
|
|
||||||
if isinstance(transcript_data, dict):
|
|
||||||
# Reconstruct Transcript object from serialized dict
|
|
||||||
words_list = transcript_data.get("words", [])
|
|
||||||
word_objects = [
|
|
||||||
Word(**w) if isinstance(w, dict) else w for w in words_list
|
|
||||||
]
|
|
||||||
topic["transcript"] = TranscriptType(
|
|
||||||
words=word_objects,
|
|
||||||
translation=transcript_data.get("translation"),
|
|
||||||
)
|
|
||||||
elif transcript_data is None:
|
|
||||||
topic["transcript"] = TranscriptType(words=[])
|
|
||||||
return topic
|
|
||||||
|
|
||||||
topic_objects = [TitleSummary(**normalize_topic(t)) for t in topics]
|
|
||||||
empty_pipeline = topic_processing.EmptyPipeline(logger=logger)
|
|
||||||
|
|
||||||
transcript = await transcripts_controller.get_by_id(transcript_id)
|
|
||||||
|
|
||||||
long_summary = ""
|
|
||||||
short_summary = ""
|
|
||||||
|
|
||||||
async def on_long(s):
|
|
||||||
nonlocal long_summary
|
|
||||||
# s is FinalLongSummary object
|
|
||||||
long_summary = s.long_summary if hasattr(s, "long_summary") else str(s)
|
|
||||||
|
|
||||||
async def on_short(s):
|
|
||||||
nonlocal short_summary
|
|
||||||
# s is FinalShortSummary object
|
|
||||||
short_summary = (
|
|
||||||
s.short_summary if hasattr(s, "short_summary") else str(s)
|
|
||||||
)
|
|
||||||
|
|
||||||
await topic_processing.generate_summaries(
|
|
||||||
topic_objects,
|
|
||||||
transcript,
|
|
||||||
on_long_summary_callback=on_long,
|
|
||||||
on_short_summary_callback=on_short,
|
|
||||||
empty_pipeline=empty_pipeline,
|
|
||||||
logger=logger,
|
|
||||||
)
|
|
||||||
|
|
||||||
return long_summary, short_summary
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
|
|
||||||
try:
|
|
||||||
summary, short_summary = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"summary": summary,
|
|
||||||
"short_summary": short_summary,
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_summary complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
summary_len=len(summary) if summary else 0,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id,
|
|
||||||
"generate_summary",
|
|
||||||
"completed",
|
|
||||||
task.workflow_instance_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] generate_summary failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_summary", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,111 +0,0 @@
|
|||||||
"""Conductor worker: generate_title - Generate meeting title using LLM."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="generate_title")
|
|
||||||
def generate_title(task: Task) -> TaskResult:
|
|
||||||
"""Generate meeting title from detected topics using LLM.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
topics: list[dict] - Detected topics
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
title: str - Generated title
|
|
||||||
"""
|
|
||||||
topics = task.input_data.get("topics", [])
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_title",
|
|
||||||
topic_count=len(topics),
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_title", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not topics:
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"title": "Untitled Meeting"}
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
from reflector.pipelines import topic_processing
|
|
||||||
from reflector.processors.types import TitleSummary, Word
|
|
||||||
from reflector.processors.types import Transcript as TranscriptType
|
|
||||||
|
|
||||||
# detect_topics returns TitleSummary objects (with transcript: Transcript)
|
|
||||||
# When serialized, transcript becomes {translation, words} dict
|
|
||||||
# We need to reconstruct TitleSummary objects with proper Transcript
|
|
||||||
def normalize_topic(t):
|
|
||||||
topic = dict(t)
|
|
||||||
transcript_data = topic.get("transcript")
|
|
||||||
if isinstance(transcript_data, dict):
|
|
||||||
# Reconstruct Transcript object from serialized dict
|
|
||||||
words_list = transcript_data.get("words", [])
|
|
||||||
word_objects = [
|
|
||||||
Word(**w) if isinstance(w, dict) else w for w in words_list
|
|
||||||
]
|
|
||||||
topic["transcript"] = TranscriptType(
|
|
||||||
words=word_objects, translation=transcript_data.get("translation")
|
|
||||||
)
|
|
||||||
elif transcript_data is None:
|
|
||||||
topic["transcript"] = TranscriptType(words=[])
|
|
||||||
return topic
|
|
||||||
|
|
||||||
topic_objects = [TitleSummary(**normalize_topic(t)) for t in topics]
|
|
||||||
empty_pipeline = topic_processing.EmptyPipeline(logger=logger)
|
|
||||||
|
|
||||||
async def noop_callback(t):
|
|
||||||
pass
|
|
||||||
|
|
||||||
title = await topic_processing.generate_title(
|
|
||||||
topic_objects,
|
|
||||||
on_title_callback=noop_callback,
|
|
||||||
empty_pipeline=empty_pipeline,
|
|
||||||
logger=logger,
|
|
||||||
)
|
|
||||||
return title
|
|
||||||
|
|
||||||
try:
|
|
||||||
title = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"title": title}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_title complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
title=title,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_title", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] generate_title failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_title", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,106 +0,0 @@
|
|||||||
"""Conductor worker: generate_waveform - Generate waveform visualization data."""
|
|
||||||
|
|
||||||
import tempfile
|
|
||||||
from pathlib import Path
|
|
||||||
|
|
||||||
import httpx
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.storage import get_transcripts_storage
|
|
||||||
from reflector.utils.audio_waveform import get_audio_waveform
|
|
||||||
|
|
||||||
PRESIGNED_URL_EXPIRATION_SECONDS = 7200
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="generate_waveform")
|
|
||||||
def generate_waveform(task: Task) -> TaskResult:
|
|
||||||
"""Generate waveform visualization data from mixed audio.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
audio_key: str - S3 key of the audio file
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
waveform: list[float] - Waveform peaks array
|
|
||||||
"""
|
|
||||||
audio_key = task.input_data.get("audio_key")
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_waveform", audio_key=audio_key, transcript_id=transcript_id
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_waveform", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not audio_key or not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing audio_key or transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
storage = get_transcripts_storage()
|
|
||||||
audio_url = await storage.get_file_url(
|
|
||||||
audio_key,
|
|
||||||
operation="get_object",
|
|
||||||
expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
|
|
||||||
)
|
|
||||||
|
|
||||||
# Download audio to temp file
|
|
||||||
with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as tmp:
|
|
||||||
tmp_path = Path(tmp.name)
|
|
||||||
async with httpx.AsyncClient() as client:
|
|
||||||
resp = await client.get(audio_url)
|
|
||||||
resp.raise_for_status()
|
|
||||||
tmp.write(resp.content)
|
|
||||||
|
|
||||||
try:
|
|
||||||
waveform = get_audio_waveform(tmp_path, segments_count=255)
|
|
||||||
finally:
|
|
||||||
tmp_path.unlink(missing_ok=True)
|
|
||||||
|
|
||||||
return waveform
|
|
||||||
|
|
||||||
try:
|
|
||||||
waveform = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"waveform": waveform}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] generate_waveform complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
peaks_count=len(waveform) if waveform else 0,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id,
|
|
||||||
"generate_waveform",
|
|
||||||
"completed",
|
|
||||||
task.workflow_instance_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] generate_waveform failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "generate_waveform", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,96 +0,0 @@
|
|||||||
"""Conductor worker: get_participants - Fetch meeting participants from Daily.co API."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.dailyco_api.client import DailyApiClient
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="get_participants")
|
|
||||||
def get_participants(task: Task) -> TaskResult:
|
|
||||||
"""Fetch meeting participants from Daily.co API.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
mtg_session_id: str - Daily.co meeting session identifier
|
|
||||||
transcript_id: str - Transcript ID for progress tracking
|
|
||||||
|
|
||||||
Output:
|
|
||||||
participants: list[dict] - List of participant info
|
|
||||||
- participant_id: str
|
|
||||||
- user_name: str | None
|
|
||||||
- user_id: str | None
|
|
||||||
"""
|
|
||||||
mtg_session_id = task.input_data.get("mtg_session_id")
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info("[Worker] get_participants", mtg_session_id=mtg_session_id)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "get_participants", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not mtg_session_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing mtg_session_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
if not settings.DAILY_API_KEY:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "DAILY_API_KEY not configured"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _fetch():
|
|
||||||
async with DailyApiClient(api_key=settings.DAILY_API_KEY) as client:
|
|
||||||
return await client.get_meeting_participants(mtg_session_id)
|
|
||||||
|
|
||||||
try:
|
|
||||||
response = asyncio.run(_fetch())
|
|
||||||
|
|
||||||
participants = [
|
|
||||||
{
|
|
||||||
"participant_id": p.participant_id,
|
|
||||||
"user_name": p.user_name,
|
|
||||||
"user_id": p.user_id,
|
|
||||||
}
|
|
||||||
for p in response.data
|
|
||||||
]
|
|
||||||
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"participants": participants}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] get_participants complete",
|
|
||||||
mtg_session_id=mtg_session_id,
|
|
||||||
count=len(participants),
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id,
|
|
||||||
"get_participants",
|
|
||||||
"completed",
|
|
||||||
task.workflow_instance_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] get_participants failed", error=str(e))
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "get_participants", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,90 +0,0 @@
|
|||||||
"""Conductor worker: get_recording - Fetch recording metadata from Daily.co API."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.dailyco_api.client import DailyApiClient
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="get_recording")
|
|
||||||
def get_recording(task: Task) -> TaskResult:
|
|
||||||
"""Fetch recording metadata from Daily.co API.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
recording_id: str - Daily.co recording identifier
|
|
||||||
transcript_id: str - Transcript ID for progress tracking
|
|
||||||
|
|
||||||
Output:
|
|
||||||
id: str - Recording ID
|
|
||||||
mtg_session_id: str - Meeting session ID
|
|
||||||
room_name: str - Room name
|
|
||||||
duration: int - Recording duration in seconds
|
|
||||||
"""
|
|
||||||
recording_id = task.input_data.get("recording_id")
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info("[Worker] get_recording", recording_id=recording_id)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "get_recording", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not recording_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing recording_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
if not settings.DAILY_API_KEY:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "DAILY_API_KEY not configured"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _fetch():
|
|
||||||
async with DailyApiClient(api_key=settings.DAILY_API_KEY) as client:
|
|
||||||
return await client.get_recording(recording_id)
|
|
||||||
|
|
||||||
try:
|
|
||||||
recording = asyncio.run(_fetch())
|
|
||||||
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"id": recording.id,
|
|
||||||
"mtg_session_id": recording.mtgSessionId,
|
|
||||||
"room_name": recording.room_name,
|
|
||||||
"duration": recording.duration,
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] get_recording complete",
|
|
||||||
recording_id=recording_id,
|
|
||||||
room_name=recording.room_name,
|
|
||||||
duration=recording.duration,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "get_recording", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] get_recording failed", error=str(e))
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "get_recording", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,89 +0,0 @@
|
|||||||
"""Conductor worker: merge_transcripts - Merge multiple track transcriptions."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="merge_transcripts")
|
|
||||||
def merge_transcripts(task: Task) -> TaskResult:
|
|
||||||
"""Merge multiple track transcriptions into single timeline sorted by timestamp.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
transcripts: list[dict] - List of transcription results with words
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
all_words: list[dict] - Merged and sorted words
|
|
||||||
word_count: int - Total word count
|
|
||||||
"""
|
|
||||||
transcripts = task.input_data.get("transcripts", [])
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] merge_transcripts",
|
|
||||||
transcript_count=len(transcripts)
|
|
||||||
if isinstance(transcripts, (list, dict))
|
|
||||||
else 0,
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "merge_transcripts", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
try:
|
|
||||||
all_words = []
|
|
||||||
|
|
||||||
# Handle JOIN output (dict with task refs as keys)
|
|
||||||
if isinstance(transcripts, dict):
|
|
||||||
transcripts = list(transcripts.values())
|
|
||||||
|
|
||||||
for t in transcripts:
|
|
||||||
if isinstance(t, list):
|
|
||||||
all_words.extend(t)
|
|
||||||
elif isinstance(t, dict) and "words" in t:
|
|
||||||
all_words.extend(t["words"])
|
|
||||||
|
|
||||||
# Sort by start timestamp
|
|
||||||
all_words.sort(key=lambda w: w.get("start", 0))
|
|
||||||
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"all_words": all_words,
|
|
||||||
"word_count": len(all_words),
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] merge_transcripts complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
word_count=len(all_words),
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id,
|
|
||||||
"merge_transcripts",
|
|
||||||
"completed",
|
|
||||||
task.workflow_instance_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] merge_transcripts failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "merge_transcripts", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,278 +0,0 @@
|
|||||||
"""Conductor worker: mixdown_tracks - Mix multiple audio tracks into single file.
|
|
||||||
|
|
||||||
Builds PyAV filter graph with amix filter to combine N padded tracks into
|
|
||||||
a single stereo MP3 file.
|
|
||||||
"""
|
|
||||||
|
|
||||||
import tempfile
|
|
||||||
from fractions import Fraction
|
|
||||||
from pathlib import Path
|
|
||||||
|
|
||||||
import av
|
|
||||||
from av.audio.resampler import AudioResampler
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.storage import get_transcripts_storage
|
|
||||||
|
|
||||||
PRESIGNED_URL_EXPIRATION_SECONDS = 7200
|
|
||||||
MP3_BITRATE = 192000
|
|
||||||
|
|
||||||
|
|
||||||
def _build_mixdown_filter_graph(containers: list, target_sample_rate: int):
|
|
||||||
"""Build PyAV filter graph: N abuffer -> amix -> aformat -> sink.
|
|
||||||
|
|
||||||
Args:
|
|
||||||
containers: List of PyAV containers for input tracks
|
|
||||||
target_sample_rate: Output sample rate
|
|
||||||
|
|
||||||
Returns:
|
|
||||||
Tuple of (graph, inputs list, sink)
|
|
||||||
"""
|
|
||||||
graph = av.filter.Graph()
|
|
||||||
inputs = []
|
|
||||||
|
|
||||||
for idx in range(len(containers)):
|
|
||||||
args = (
|
|
||||||
f"time_base=1/{target_sample_rate}:"
|
|
||||||
f"sample_rate={target_sample_rate}:"
|
|
||||||
f"sample_fmt=s32:"
|
|
||||||
f"channel_layout=stereo"
|
|
||||||
)
|
|
||||||
in_ctx = graph.add("abuffer", args=args, name=f"in{idx}")
|
|
||||||
inputs.append(in_ctx)
|
|
||||||
|
|
||||||
# amix with normalize=0 to prevent volume reduction
|
|
||||||
mixer = graph.add("amix", args=f"inputs={len(containers)}:normalize=0", name="mix")
|
|
||||||
fmt = graph.add(
|
|
||||||
"aformat",
|
|
||||||
args=f"sample_fmts=s16:channel_layouts=stereo:sample_rates={target_sample_rate}",
|
|
||||||
name="fmt",
|
|
||||||
)
|
|
||||||
sink = graph.add("abuffersink", name="out")
|
|
||||||
|
|
||||||
for idx, in_ctx in enumerate(inputs):
|
|
||||||
in_ctx.link_to(mixer, 0, idx)
|
|
||||||
mixer.link_to(fmt)
|
|
||||||
fmt.link_to(sink)
|
|
||||||
graph.configure()
|
|
||||||
|
|
||||||
return graph, inputs, sink
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="mixdown_tracks")
|
|
||||||
def mixdown_tracks(task: Task) -> TaskResult:
|
|
||||||
"""Mix multiple audio tracks into single stereo file.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
padded_urls: list[str] - Presigned URLs of padded tracks
|
|
||||||
transcript_id: str - Transcript ID for storage path
|
|
||||||
|
|
||||||
Output:
|
|
||||||
audio_key: str - S3 key of mixed audio file
|
|
||||||
duration: float - Audio duration in seconds
|
|
||||||
size: int - File size in bytes
|
|
||||||
"""
|
|
||||||
padded_urls = task.input_data.get("padded_urls", [])
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] mixdown_tracks",
|
|
||||||
track_count=len(padded_urls),
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "mixdown_tracks", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not padded_urls or not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing padded_urls or transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
storage = get_transcripts_storage()
|
|
||||||
|
|
||||||
# Determine target sample rate from first track
|
|
||||||
target_sample_rate = None
|
|
||||||
for url in padded_urls:
|
|
||||||
if not url:
|
|
||||||
continue
|
|
||||||
try:
|
|
||||||
with av.open(url) as container:
|
|
||||||
for frame in container.decode(audio=0):
|
|
||||||
target_sample_rate = frame.sample_rate
|
|
||||||
break
|
|
||||||
except Exception:
|
|
||||||
continue
|
|
||||||
if target_sample_rate:
|
|
||||||
break
|
|
||||||
|
|
||||||
if not target_sample_rate:
|
|
||||||
raise Exception("Mixdown failed: No decodable audio frames in any track")
|
|
||||||
|
|
||||||
# Open all containers with reconnect options for S3 streaming
|
|
||||||
containers = []
|
|
||||||
valid_urls = [url for url in padded_urls if url]
|
|
||||||
|
|
||||||
for url in valid_urls:
|
|
||||||
try:
|
|
||||||
c = av.open(
|
|
||||||
url,
|
|
||||||
options={
|
|
||||||
"reconnect": "1",
|
|
||||||
"reconnect_streamed": "1",
|
|
||||||
"reconnect_delay_max": "5",
|
|
||||||
},
|
|
||||||
)
|
|
||||||
containers.append(c)
|
|
||||||
except Exception as e:
|
|
||||||
logger.warning(
|
|
||||||
"Mixdown: failed to open container", url=url[:50], error=str(e)
|
|
||||||
)
|
|
||||||
|
|
||||||
if not containers:
|
|
||||||
raise Exception("Mixdown failed: Could not open any track containers")
|
|
||||||
|
|
||||||
try:
|
|
||||||
# Build filter graph
|
|
||||||
graph, inputs, sink = _build_mixdown_filter_graph(
|
|
||||||
containers, target_sample_rate
|
|
||||||
)
|
|
||||||
|
|
||||||
# Create temp file for output
|
|
||||||
with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_file:
|
|
||||||
temp_path = temp_file.name
|
|
||||||
|
|
||||||
try:
|
|
||||||
# Open output container for MP3
|
|
||||||
with av.open(temp_path, "w", format="mp3") as out_container:
|
|
||||||
out_stream = out_container.add_stream(
|
|
||||||
"libmp3lame", rate=target_sample_rate
|
|
||||||
)
|
|
||||||
out_stream.bit_rate = MP3_BITRATE
|
|
||||||
|
|
||||||
decoders = [c.decode(audio=0) for c in containers]
|
|
||||||
active = [True] * len(decoders)
|
|
||||||
resamplers = [
|
|
||||||
AudioResampler(
|
|
||||||
format="s32", layout="stereo", rate=target_sample_rate
|
|
||||||
)
|
|
||||||
for _ in decoders
|
|
||||||
]
|
|
||||||
|
|
||||||
duration_samples = 0
|
|
||||||
|
|
||||||
while any(active):
|
|
||||||
for i, (dec, is_active) in enumerate(zip(decoders, active)):
|
|
||||||
if not is_active:
|
|
||||||
continue
|
|
||||||
try:
|
|
||||||
frame = next(dec)
|
|
||||||
except StopIteration:
|
|
||||||
active[i] = False
|
|
||||||
inputs[i].push(None)
|
|
||||||
continue
|
|
||||||
|
|
||||||
if frame.sample_rate != target_sample_rate:
|
|
||||||
continue
|
|
||||||
|
|
||||||
out_frames = resamplers[i].resample(frame) or []
|
|
||||||
for rf in out_frames:
|
|
||||||
rf.sample_rate = target_sample_rate
|
|
||||||
rf.time_base = Fraction(1, target_sample_rate)
|
|
||||||
inputs[i].push(rf)
|
|
||||||
|
|
||||||
while True:
|
|
||||||
try:
|
|
||||||
mixed = sink.pull()
|
|
||||||
except Exception:
|
|
||||||
break
|
|
||||||
mixed.sample_rate = target_sample_rate
|
|
||||||
mixed.time_base = Fraction(1, target_sample_rate)
|
|
||||||
duration_samples += mixed.samples
|
|
||||||
for packet in out_stream.encode(mixed):
|
|
||||||
out_container.mux(packet)
|
|
||||||
|
|
||||||
# Flush remaining
|
|
||||||
while True:
|
|
||||||
try:
|
|
||||||
mixed = sink.pull()
|
|
||||||
except Exception:
|
|
||||||
break
|
|
||||||
mixed.sample_rate = target_sample_rate
|
|
||||||
mixed.time_base = Fraction(1, target_sample_rate)
|
|
||||||
duration_samples += mixed.samples
|
|
||||||
for packet in out_stream.encode(mixed):
|
|
||||||
out_container.mux(packet)
|
|
||||||
|
|
||||||
for packet in out_stream.encode(None):
|
|
||||||
out_container.mux(packet)
|
|
||||||
|
|
||||||
# Get file size and duration
|
|
||||||
file_size = Path(temp_path).stat().st_size
|
|
||||||
duration = (
|
|
||||||
duration_samples / target_sample_rate if target_sample_rate else 0
|
|
||||||
)
|
|
||||||
|
|
||||||
# Upload to S3
|
|
||||||
storage_path = f"{transcript_id}/audio.mp3"
|
|
||||||
with open(temp_path, "rb") as mp3_file:
|
|
||||||
await storage.put_file(storage_path, mp3_file)
|
|
||||||
|
|
||||||
finally:
|
|
||||||
Path(temp_path).unlink(missing_ok=True)
|
|
||||||
|
|
||||||
finally:
|
|
||||||
for c in containers:
|
|
||||||
try:
|
|
||||||
c.close()
|
|
||||||
except Exception:
|
|
||||||
pass
|
|
||||||
|
|
||||||
return {
|
|
||||||
"audio_key": storage_path,
|
|
||||||
"duration": duration,
|
|
||||||
"size": file_size,
|
|
||||||
}
|
|
||||||
|
|
||||||
try:
|
|
||||||
result = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = result
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] mixdown_tracks complete",
|
|
||||||
audio_key=result["audio_key"],
|
|
||||||
duration=result["duration"],
|
|
||||||
size=result["size"],
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "mixdown_tracks", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] mixdown_tracks failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "mixdown_tracks", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,322 +0,0 @@
|
|||||||
"""Conductor worker: pad_track - Pad audio track with silence for alignment.
|
|
||||||
|
|
||||||
This worker extracts stream.start_time from WebM container metadata and applies
|
|
||||||
silence padding using PyAV filter graph (adelay). The padded audio is uploaded
|
|
||||||
to S3 and a presigned URL is returned.
|
|
||||||
"""
|
|
||||||
|
|
||||||
import math
|
|
||||||
import tempfile
|
|
||||||
from fractions import Fraction
|
|
||||||
from pathlib import Path
|
|
||||||
|
|
||||||
import av
|
|
||||||
from av.audio.resampler import AudioResampler
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
# Audio constants matching existing pipeline
|
|
||||||
OPUS_STANDARD_SAMPLE_RATE = 48000
|
|
||||||
OPUS_DEFAULT_BIT_RATE = 64000
|
|
||||||
PRESIGNED_URL_EXPIRATION_SECONDS = 7200
|
|
||||||
|
|
||||||
|
|
||||||
def _extract_stream_start_time_from_container(container, track_idx: int) -> float:
|
|
||||||
"""Extract meeting-relative start time from WebM stream metadata.
|
|
||||||
|
|
||||||
Uses PyAV to read stream.start_time from WebM container.
|
|
||||||
More accurate than filename timestamps by ~209ms due to network/encoding delays.
|
|
||||||
|
|
||||||
Args:
|
|
||||||
container: PyAV container object
|
|
||||||
track_idx: Track index for logging
|
|
||||||
|
|
||||||
Returns:
|
|
||||||
Start time in seconds (0.0 if not found)
|
|
||||||
"""
|
|
||||||
start_time_seconds = 0.0
|
|
||||||
try:
|
|
||||||
audio_streams = [s for s in container.streams if s.type == "audio"]
|
|
||||||
stream = audio_streams[0] if audio_streams else container.streams[0]
|
|
||||||
|
|
||||||
# 1) Try stream-level start_time (most reliable for Daily.co tracks)
|
|
||||||
if stream.start_time is not None and stream.time_base is not None:
|
|
||||||
start_time_seconds = float(stream.start_time * stream.time_base)
|
|
||||||
|
|
||||||
# 2) Fallback to container-level start_time (in av.time_base units)
|
|
||||||
if (start_time_seconds <= 0) and (container.start_time is not None):
|
|
||||||
start_time_seconds = float(container.start_time * av.time_base)
|
|
||||||
|
|
||||||
# 3) Fallback to first packet DTS in stream.time_base
|
|
||||||
if start_time_seconds <= 0:
|
|
||||||
for packet in container.demux(stream):
|
|
||||||
if packet.dts is not None:
|
|
||||||
start_time_seconds = float(packet.dts * stream.time_base)
|
|
||||||
break
|
|
||||||
except Exception as e:
|
|
||||||
logger.warning(
|
|
||||||
"PyAV metadata read failed; assuming 0 start_time",
|
|
||||||
track_idx=track_idx,
|
|
||||||
error=str(e),
|
|
||||||
)
|
|
||||||
start_time_seconds = 0.0
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
f"Track {track_idx} stream metadata: start_time={start_time_seconds:.3f}s",
|
|
||||||
track_idx=track_idx,
|
|
||||||
)
|
|
||||||
return start_time_seconds
|
|
||||||
|
|
||||||
|
|
||||||
def _apply_audio_padding_to_file(
|
|
||||||
in_container,
|
|
||||||
output_path: str,
|
|
||||||
start_time_seconds: float,
|
|
||||||
track_idx: int,
|
|
||||||
) -> None:
|
|
||||||
"""Apply silence padding to audio track using PyAV filter graph.
|
|
||||||
|
|
||||||
Filter chain: abuffer -> aresample -> adelay -> abuffersink
|
|
||||||
|
|
||||||
Args:
|
|
||||||
in_container: PyAV input container
|
|
||||||
output_path: Path to write padded output
|
|
||||||
start_time_seconds: Amount of silence to prepend
|
|
||||||
track_idx: Track index for logging
|
|
||||||
"""
|
|
||||||
delay_ms = math.floor(start_time_seconds * 1000)
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
f"Padding track {track_idx} with {delay_ms}ms delay using PyAV",
|
|
||||||
track_idx=track_idx,
|
|
||||||
delay_ms=delay_ms,
|
|
||||||
)
|
|
||||||
|
|
||||||
with av.open(output_path, "w", format="webm") as out_container:
|
|
||||||
in_stream = next((s for s in in_container.streams if s.type == "audio"), None)
|
|
||||||
if in_stream is None:
|
|
||||||
raise Exception("No audio stream in input")
|
|
||||||
|
|
||||||
out_stream = out_container.add_stream("libopus", rate=OPUS_STANDARD_SAMPLE_RATE)
|
|
||||||
out_stream.bit_rate = OPUS_DEFAULT_BIT_RATE
|
|
||||||
graph = av.filter.Graph()
|
|
||||||
|
|
||||||
abuf_args = (
|
|
||||||
f"time_base=1/{OPUS_STANDARD_SAMPLE_RATE}:"
|
|
||||||
f"sample_rate={OPUS_STANDARD_SAMPLE_RATE}:"
|
|
||||||
f"sample_fmt=s16:"
|
|
||||||
f"channel_layout=stereo"
|
|
||||||
)
|
|
||||||
src = graph.add("abuffer", args=abuf_args, name="src")
|
|
||||||
aresample_f = graph.add("aresample", args="async=1", name="ares")
|
|
||||||
# adelay requires one delay value per channel separated by '|'
|
|
||||||
delays_arg = f"{delay_ms}|{delay_ms}"
|
|
||||||
adelay_f = graph.add("adelay", args=f"delays={delays_arg}:all=1", name="delay")
|
|
||||||
sink = graph.add("abuffersink", name="sink")
|
|
||||||
|
|
||||||
src.link_to(aresample_f)
|
|
||||||
aresample_f.link_to(adelay_f)
|
|
||||||
adelay_f.link_to(sink)
|
|
||||||
graph.configure()
|
|
||||||
|
|
||||||
resampler = AudioResampler(
|
|
||||||
format="s16", layout="stereo", rate=OPUS_STANDARD_SAMPLE_RATE
|
|
||||||
)
|
|
||||||
|
|
||||||
# Decode -> resample -> push through graph -> encode Opus
|
|
||||||
for frame in in_container.decode(in_stream):
|
|
||||||
out_frames = resampler.resample(frame) or []
|
|
||||||
for rframe in out_frames:
|
|
||||||
rframe.sample_rate = OPUS_STANDARD_SAMPLE_RATE
|
|
||||||
rframe.time_base = Fraction(1, OPUS_STANDARD_SAMPLE_RATE)
|
|
||||||
src.push(rframe)
|
|
||||||
|
|
||||||
while True:
|
|
||||||
try:
|
|
||||||
f_out = sink.pull()
|
|
||||||
except Exception:
|
|
||||||
break
|
|
||||||
f_out.sample_rate = OPUS_STANDARD_SAMPLE_RATE
|
|
||||||
f_out.time_base = Fraction(1, OPUS_STANDARD_SAMPLE_RATE)
|
|
||||||
for packet in out_stream.encode(f_out):
|
|
||||||
out_container.mux(packet)
|
|
||||||
|
|
||||||
# Flush remaining frames
|
|
||||||
src.push(None)
|
|
||||||
while True:
|
|
||||||
try:
|
|
||||||
f_out = sink.pull()
|
|
||||||
except Exception:
|
|
||||||
break
|
|
||||||
f_out.sample_rate = OPUS_STANDARD_SAMPLE_RATE
|
|
||||||
f_out.time_base = Fraction(1, OPUS_STANDARD_SAMPLE_RATE)
|
|
||||||
for packet in out_stream.encode(f_out):
|
|
||||||
out_container.mux(packet)
|
|
||||||
|
|
||||||
for packet in out_stream.encode(None):
|
|
||||||
out_container.mux(packet)
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="pad_track")
|
|
||||||
def pad_track(task: Task) -> TaskResult:
|
|
||||||
"""Pad audio track with silence for alignment.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
track_index: int - Index of the track
|
|
||||||
s3_key: str - S3 key of the source audio file
|
|
||||||
bucket_name: str - S3 bucket name
|
|
||||||
transcript_id: str - Transcript ID for storage path
|
|
||||||
|
|
||||||
Output:
|
|
||||||
padded_url: str - Presigned URL of padded track
|
|
||||||
size: int - File size in bytes
|
|
||||||
track_index: int - Track index (echoed back)
|
|
||||||
"""
|
|
||||||
track_index = task.input_data.get("track_index", 0)
|
|
||||||
s3_key = task.input_data.get("s3_key")
|
|
||||||
bucket_name = task.input_data.get("bucket_name")
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] pad_track",
|
|
||||||
track_index=track_index,
|
|
||||||
s3_key=s3_key,
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "pad_track", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not s3_key or not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing s3_key or transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
# Create fresh storage instance to avoid aioboto3 fork issues
|
|
||||||
from reflector.settings import settings
|
|
||||||
from reflector.storage.storage_aws import AwsStorage
|
|
||||||
|
|
||||||
storage = AwsStorage(
|
|
||||||
aws_bucket_name=settings.TRANSCRIPT_STORAGE_AWS_BUCKET_NAME,
|
|
||||||
aws_region=settings.TRANSCRIPT_STORAGE_AWS_REGION,
|
|
||||||
aws_access_key_id=settings.TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID,
|
|
||||||
aws_secret_access_key=settings.TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY,
|
|
||||||
)
|
|
||||||
|
|
||||||
# Get presigned URL for source file
|
|
||||||
source_url = await storage.get_file_url(
|
|
||||||
s3_key,
|
|
||||||
operation="get_object",
|
|
||||||
expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
|
|
||||||
bucket=bucket_name,
|
|
||||||
)
|
|
||||||
|
|
||||||
# Open container and extract start time
|
|
||||||
with av.open(source_url) as in_container:
|
|
||||||
start_time_seconds = _extract_stream_start_time_from_container(
|
|
||||||
in_container, track_index
|
|
||||||
)
|
|
||||||
|
|
||||||
# If no padding needed, return original URL
|
|
||||||
if start_time_seconds <= 0:
|
|
||||||
logger.info(
|
|
||||||
f"Track {track_index} requires no padding",
|
|
||||||
track_index=track_index,
|
|
||||||
)
|
|
||||||
return {
|
|
||||||
"padded_url": source_url,
|
|
||||||
"size": 0,
|
|
||||||
"track_index": track_index,
|
|
||||||
}
|
|
||||||
|
|
||||||
# Create temp file for padded output
|
|
||||||
with tempfile.NamedTemporaryFile(suffix=".webm", delete=False) as temp_file:
|
|
||||||
temp_path = temp_file.name
|
|
||||||
|
|
||||||
try:
|
|
||||||
_apply_audio_padding_to_file(
|
|
||||||
in_container, temp_path, start_time_seconds, track_index
|
|
||||||
)
|
|
||||||
|
|
||||||
# Get file size
|
|
||||||
file_size = Path(temp_path).stat().st_size
|
|
||||||
|
|
||||||
# Upload using storage layer (use separate path in shadow mode to avoid conflicts)
|
|
||||||
storage_path = f"file_pipeline_conductor/{transcript_id}/tracks/padded_{track_index}.webm"
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
f"About to upload padded track",
|
|
||||||
key=storage_path,
|
|
||||||
size=file_size,
|
|
||||||
)
|
|
||||||
|
|
||||||
with open(temp_path, "rb") as padded_file:
|
|
||||||
upload_result = await storage.put_file(storage_path, padded_file)
|
|
||||||
logger.info(
|
|
||||||
f"storage.put_file returned",
|
|
||||||
result=str(upload_result),
|
|
||||||
)
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
f"Uploaded padded track to S3",
|
|
||||||
key=storage_path,
|
|
||||||
size=file_size,
|
|
||||||
)
|
|
||||||
|
|
||||||
finally:
|
|
||||||
Path(temp_path).unlink(missing_ok=True)
|
|
||||||
|
|
||||||
# Get presigned URL for padded file
|
|
||||||
padded_url = await storage.get_file_url(
|
|
||||||
storage_path,
|
|
||||||
operation="get_object",
|
|
||||||
expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
|
|
||||||
)
|
|
||||||
|
|
||||||
return {
|
|
||||||
"padded_url": padded_url,
|
|
||||||
"size": file_size,
|
|
||||||
"track_index": track_index,
|
|
||||||
}
|
|
||||||
|
|
||||||
try:
|
|
||||||
result = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = result
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] pad_track complete",
|
|
||||||
track_index=track_index,
|
|
||||||
padded_url=result["padded_url"][:50] + "...",
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "pad_track", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] pad_track failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "pad_track", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,101 +0,0 @@
|
|||||||
"""Conductor worker: post_zulip - Post or update Zulip message with transcript summary."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="post_zulip")
|
|
||||||
def post_zulip(task: Task) -> TaskResult:
|
|
||||||
"""Post or update a Zulip message with the transcript summary.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
message_id: str | None - Zulip message ID
|
|
||||||
"""
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info("[Worker] post_zulip", transcript_id=transcript_id)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "post_zulip", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
# Check if Zulip is configured
|
|
||||||
if not settings.ZULIP_REALM or not settings.ZULIP_API_KEY:
|
|
||||||
logger.info("[Worker] post_zulip: Zulip not configured, skipping")
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {"message_id": None}
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
|
|
||||||
from reflector.db import _database_context
|
|
||||||
from reflector.db.transcripts import transcripts_controller
|
|
||||||
from reflector.settings import settings as app_settings
|
|
||||||
from reflector.zulip import post_transcript_to_zulip
|
|
||||||
|
|
||||||
# Create fresh database connection for subprocess (not shared from parent)
|
|
||||||
_database_context.set(None)
|
|
||||||
db = databases.Database(app_settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
try:
|
|
||||||
transcript = await transcripts_controller.get_by_id(transcript_id)
|
|
||||||
if transcript is None:
|
|
||||||
raise ValueError(f"Transcript {transcript_id} not found in database")
|
|
||||||
message_id = await post_transcript_to_zulip(transcript)
|
|
||||||
return message_id
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
|
|
||||||
try:
|
|
||||||
message_id = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"message_id": str(message_id) if message_id else None
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] post_zulip complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
message_id=message_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "post_zulip", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] post_zulip failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "post_zulip", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,115 +0,0 @@
|
|||||||
"""Conductor worker: send_webhook - Send transcript completion webhook."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="send_webhook")
|
|
||||||
def send_webhook(task: Task) -> TaskResult:
|
|
||||||
"""Send the transcript completion webhook to the configured URL.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
transcript_id: str - Transcript ID
|
|
||||||
room_id: str - Room ID
|
|
||||||
|
|
||||||
Output:
|
|
||||||
sent: bool - Whether webhook was sent
|
|
||||||
status_code: int | None - HTTP status code
|
|
||||||
"""
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
room_id = task.input_data.get("room_id")
|
|
||||||
|
|
||||||
logger.info("[Worker] send_webhook", transcript_id=transcript_id, room_id=room_id)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "send_webhook", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not transcript_id:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing transcript_id"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
import databases
|
|
||||||
|
|
||||||
from reflector.db import _database_context
|
|
||||||
from reflector.db.rooms import rooms_controller
|
|
||||||
from reflector.db.transcripts import transcripts_controller
|
|
||||||
from reflector.settings import settings
|
|
||||||
from reflector.worker.webhook import send_transcript_webhook
|
|
||||||
|
|
||||||
# Create fresh database connection for subprocess (not shared from parent)
|
|
||||||
_database_context.set(None)
|
|
||||||
db = databases.Database(settings.DATABASE_URL)
|
|
||||||
_database_context.set(db)
|
|
||||||
await db.connect()
|
|
||||||
|
|
||||||
try:
|
|
||||||
transcript = await transcripts_controller.get_by_id(transcript_id)
|
|
||||||
if transcript is None:
|
|
||||||
raise ValueError(f"Transcript {transcript_id} not found in database")
|
|
||||||
|
|
||||||
# Get room for webhook URL
|
|
||||||
room = None
|
|
||||||
if room_id:
|
|
||||||
try:
|
|
||||||
room = await rooms_controller.get_by_id(room_id)
|
|
||||||
except Exception:
|
|
||||||
pass
|
|
||||||
|
|
||||||
if not room or not room.webhook_url:
|
|
||||||
logger.info(
|
|
||||||
"[Worker] send_webhook: No webhook URL configured",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
)
|
|
||||||
return False, None
|
|
||||||
|
|
||||||
status_code = await send_transcript_webhook(transcript, room)
|
|
||||||
return True, status_code
|
|
||||||
finally:
|
|
||||||
await db.disconnect()
|
|
||||||
_database_context.set(None)
|
|
||||||
|
|
||||||
try:
|
|
||||||
sent, status_code = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"sent": sent,
|
|
||||||
"status_code": status_code,
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] send_webhook complete",
|
|
||||||
transcript_id=transcript_id,
|
|
||||||
sent=sent,
|
|
||||||
status_code=status_code,
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "send_webhook", "completed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] send_webhook failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "send_webhook", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,96 +0,0 @@
|
|||||||
"""Conductor worker: transcribe_track - Transcribe audio track using GPU service."""
|
|
||||||
|
|
||||||
from conductor.client.http.models import Task, TaskResult
|
|
||||||
from conductor.client.http.models.task_result_status import TaskResultStatus
|
|
||||||
from conductor.client.worker.worker_task import worker_task
|
|
||||||
from reflector.conductor.progress import emit_progress
|
|
||||||
from reflector.logger import logger
|
|
||||||
|
|
||||||
|
|
||||||
@worker_task(task_definition_name="transcribe_track")
|
|
||||||
def transcribe_track(task: Task) -> TaskResult:
|
|
||||||
"""Transcribe audio track using GPU (Modal.com) or local Whisper.
|
|
||||||
|
|
||||||
Input:
|
|
||||||
track_index: int - Index of the track
|
|
||||||
audio_url: str - Presigned URL of the audio file
|
|
||||||
language: str - Language code (default: "en")
|
|
||||||
transcript_id: str - Transcript ID for progress tracking
|
|
||||||
|
|
||||||
Output:
|
|
||||||
words: list[dict] - List of transcribed words with timestamps and speaker
|
|
||||||
track_index: int - Track index (echoed back)
|
|
||||||
"""
|
|
||||||
track_index = task.input_data.get("track_index", 0)
|
|
||||||
audio_url = task.input_data.get("audio_url")
|
|
||||||
language = task.input_data.get("language", "en")
|
|
||||||
transcript_id = task.input_data.get("transcript_id")
|
|
||||||
|
|
||||||
logger.info("[Worker] transcribe_track", track_index=track_index, language=language)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "transcribe_track", "in_progress", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
task_result = TaskResult(
|
|
||||||
task_id=task.task_id,
|
|
||||||
workflow_instance_id=task.workflow_instance_id,
|
|
||||||
worker_id=task.worker_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
if not audio_url:
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = "Missing audio_url"
|
|
||||||
return task_result
|
|
||||||
|
|
||||||
import asyncio
|
|
||||||
|
|
||||||
async def _process():
|
|
||||||
from reflector.pipelines.transcription_helpers import (
|
|
||||||
transcribe_file_with_processor,
|
|
||||||
)
|
|
||||||
|
|
||||||
transcript = await transcribe_file_with_processor(audio_url, language)
|
|
||||||
|
|
||||||
# Tag all words with speaker index
|
|
||||||
words = []
|
|
||||||
for word in transcript.words:
|
|
||||||
word_dict = word.model_dump()
|
|
||||||
word_dict["speaker"] = track_index
|
|
||||||
words.append(word_dict)
|
|
||||||
|
|
||||||
return words
|
|
||||||
|
|
||||||
try:
|
|
||||||
words = asyncio.run(_process())
|
|
||||||
task_result.status = TaskResultStatus.COMPLETED
|
|
||||||
task_result.output_data = {
|
|
||||||
"words": words,
|
|
||||||
"track_index": track_index,
|
|
||||||
}
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"[Worker] transcribe_track complete",
|
|
||||||
track_index=track_index,
|
|
||||||
word_count=len(words),
|
|
||||||
)
|
|
||||||
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id,
|
|
||||||
"transcribe_track",
|
|
||||||
"completed",
|
|
||||||
task.workflow_instance_id,
|
|
||||||
)
|
|
||||||
|
|
||||||
except Exception as e:
|
|
||||||
logger.error("[Worker] transcribe_track failed", error=str(e), exc_info=True)
|
|
||||||
task_result.status = TaskResultStatus.FAILED
|
|
||||||
task_result.reason_for_incompletion = str(e)
|
|
||||||
if transcript_id:
|
|
||||||
emit_progress(
|
|
||||||
transcript_id, "transcribe_track", "failed", task.workflow_instance_id
|
|
||||||
)
|
|
||||||
|
|
||||||
return task_result
|
|
||||||
@@ -1,205 +0,0 @@
|
|||||||
{
|
|
||||||
"name": "diarization_pipeline",
|
|
||||||
"description": "Reflector multitrack diarization pipeline",
|
|
||||||
"version": 1,
|
|
||||||
"schemaVersion": 2,
|
|
||||||
"inputParameters": [
|
|
||||||
"recording_id",
|
|
||||||
"room_name",
|
|
||||||
"tracks",
|
|
||||||
"bucket_name",
|
|
||||||
"transcript_id",
|
|
||||||
"room_id"
|
|
||||||
],
|
|
||||||
"tasks": [
|
|
||||||
{
|
|
||||||
"name": "get_recording",
|
|
||||||
"taskReferenceName": "get_recording",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"recording_id": "${workflow.input.recording_id}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "get_participants",
|
|
||||||
"taskReferenceName": "get_participants",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"mtg_session_id": "${get_recording.output.mtg_session_id}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_dynamic_fork_tasks",
|
|
||||||
"taskReferenceName": "generate_padding_tasks",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"tracks": "${workflow.input.tracks}",
|
|
||||||
"task_type": "pad_track",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}",
|
|
||||||
"bucket_name": "${workflow.input.bucket_name}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "fork_track_padding",
|
|
||||||
"taskReferenceName": "fork_track_padding",
|
|
||||||
"type": "FORK_JOIN_DYNAMIC",
|
|
||||||
"inputParameters": {
|
|
||||||
"dynamicTasks": "${generate_padding_tasks.output.tasks}",
|
|
||||||
"dynamicTasksInput": "${generate_padding_tasks.output.inputs}"
|
|
||||||
},
|
|
||||||
"dynamicForkTasksParam": "dynamicTasks",
|
|
||||||
"dynamicForkTasksInputParamName": "dynamicTasksInput"
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "join_padding",
|
|
||||||
"taskReferenceName": "join_padding",
|
|
||||||
"type": "JOIN"
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "mixdown_tracks",
|
|
||||||
"taskReferenceName": "mixdown_tracks",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"padded_urls": "${join_padding.output..padded_url}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_waveform",
|
|
||||||
"taskReferenceName": "generate_waveform",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"audio_key": "${mixdown_tracks.output.audio_key}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "generate_dynamic_fork_tasks",
|
|
||||||
"taskReferenceName": "generate_transcription_tasks",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"tracks": "${workflow.input.tracks}",
|
|
||||||
"task_type": "transcribe_track",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}",
|
|
||||||
"padded_urls": "${join_padding.output}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "fork_transcription",
|
|
||||||
"taskReferenceName": "fork_transcription",
|
|
||||||
"type": "FORK_JOIN_DYNAMIC",
|
|
||||||
"inputParameters": {
|
|
||||||
"dynamicTasks": "${generate_transcription_tasks.output.tasks}",
|
|
||||||
"dynamicTasksInput": "${generate_transcription_tasks.output.inputs}"
|
|
||||||
},
|
|
||||||
"dynamicForkTasksParam": "dynamicTasks",
|
|
||||||
"dynamicForkTasksInputParamName": "dynamicTasksInput"
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "join_transcription",
|
|
||||||
"taskReferenceName": "join_transcription",
|
|
||||||
"type": "JOIN"
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "merge_transcripts",
|
|
||||||
"taskReferenceName": "merge_transcripts",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"transcripts": "${join_transcription.output}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "detect_topics",
|
|
||||||
"taskReferenceName": "detect_topics",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"words": "${merge_transcripts.output.all_words}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}",
|
|
||||||
"target_language": "en"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "fork_generation",
|
|
||||||
"taskReferenceName": "fork_generation",
|
|
||||||
"type": "FORK_JOIN",
|
|
||||||
"forkTasks": [
|
|
||||||
[
|
|
||||||
{
|
|
||||||
"name": "generate_title",
|
|
||||||
"taskReferenceName": "generate_title",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"topics": "${detect_topics.output.topics}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
],
|
|
||||||
[
|
|
||||||
{
|
|
||||||
"name": "generate_summary",
|
|
||||||
"taskReferenceName": "generate_summary",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"words": "${merge_transcripts.output.all_words}",
|
|
||||||
"topics": "${detect_topics.output.topics}",
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
]
|
|
||||||
]
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "join_generation",
|
|
||||||
"taskReferenceName": "join_generation",
|
|
||||||
"type": "JOIN",
|
|
||||||
"joinOn": ["generate_title", "generate_summary"]
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "finalize",
|
|
||||||
"taskReferenceName": "finalize",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}",
|
|
||||||
"title": "${generate_title.output.title}",
|
|
||||||
"summary": "${generate_summary.output.summary}",
|
|
||||||
"short_summary": "${generate_summary.output.short_summary}",
|
|
||||||
"duration": "${mixdown_tracks.output.duration}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "cleanup_consent",
|
|
||||||
"taskReferenceName": "cleanup_consent",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "post_zulip",
|
|
||||||
"taskReferenceName": "post_zulip",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}"
|
|
||||||
}
|
|
||||||
},
|
|
||||||
{
|
|
||||||
"name": "send_webhook",
|
|
||||||
"taskReferenceName": "send_webhook",
|
|
||||||
"type": "SIMPLE",
|
|
||||||
"inputParameters": {
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}",
|
|
||||||
"room_id": "${workflow.input.room_id}"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
],
|
|
||||||
"outputParameters": {
|
|
||||||
"transcript_id": "${workflow.input.transcript_id}",
|
|
||||||
"title": "${generate_title.output.title}",
|
|
||||||
"summary": "${generate_summary.output.summary}",
|
|
||||||
"duration": "${mixdown_tracks.output.duration}",
|
|
||||||
"word_count": "${merge_transcripts.output.word_count}"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
@@ -1,74 +0,0 @@
|
|||||||
"""Register workflow definition with Conductor server."""
|
|
||||||
|
|
||||||
import json
|
|
||||||
from pathlib import Path
|
|
||||||
|
|
||||||
import httpx
|
|
||||||
|
|
||||||
from reflector.logger import logger
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
|
|
||||||
def register_workflow() -> None:
|
|
||||||
"""Register the diarization pipeline workflow with Conductor server.
|
|
||||||
|
|
||||||
Raises:
|
|
||||||
httpx.HTTPStatusError: If registration fails.
|
|
||||||
"""
|
|
||||||
workflow_path = Path(__file__).parent / "diarization_pipeline.json"
|
|
||||||
|
|
||||||
with open(workflow_path) as f:
|
|
||||||
workflow = json.load(f)
|
|
||||||
|
|
||||||
base_url = settings.CONDUCTOR_SERVER_URL.rstrip("/")
|
|
||||||
url = f"{base_url}/metadata/workflow"
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"Registering workflow",
|
|
||||||
name=workflow["name"],
|
|
||||||
version=workflow["version"],
|
|
||||||
url=url,
|
|
||||||
)
|
|
||||||
|
|
||||||
with httpx.Client(timeout=30.0) as client:
|
|
||||||
resp = client.put(
|
|
||||||
url,
|
|
||||||
json=[workflow],
|
|
||||||
headers={"Content-Type": "application/json"},
|
|
||||||
)
|
|
||||||
resp.raise_for_status()
|
|
||||||
|
|
||||||
logger.info("Workflow registered successfully", name=workflow["name"])
|
|
||||||
|
|
||||||
|
|
||||||
async def register_workflow_async() -> None:
|
|
||||||
"""Async version of register_workflow."""
|
|
||||||
workflow_path = Path(__file__).parent / "diarization_pipeline.json"
|
|
||||||
|
|
||||||
with open(workflow_path) as f:
|
|
||||||
workflow = json.load(f)
|
|
||||||
|
|
||||||
base_url = settings.CONDUCTOR_SERVER_URL.rstrip("/")
|
|
||||||
url = f"{base_url}/metadata/workflow"
|
|
||||||
|
|
||||||
logger.info(
|
|
||||||
"Registering workflow",
|
|
||||||
name=workflow["name"],
|
|
||||||
version=workflow["version"],
|
|
||||||
url=url,
|
|
||||||
)
|
|
||||||
|
|
||||||
async with httpx.AsyncClient(timeout=30.0) as client:
|
|
||||||
resp = await client.put(
|
|
||||||
url,
|
|
||||||
json=[workflow],
|
|
||||||
headers={"Content-Type": "application/json"},
|
|
||||||
)
|
|
||||||
resp.raise_for_status()
|
|
||||||
|
|
||||||
logger.info("Workflow registered successfully", name=workflow["name"])
|
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
|
||||||
register_workflow()
|
|
||||||
print("Workflow registration complete!")
|
|
||||||
@@ -22,9 +22,7 @@ recordings = sa.Table(
|
|||||||
),
|
),
|
||||||
sa.Column("meeting_id", sa.String),
|
sa.Column("meeting_id", sa.String),
|
||||||
sa.Column("track_keys", sa.JSON, nullable=True),
|
sa.Column("track_keys", sa.JSON, nullable=True),
|
||||||
sa.Column("workflow_id", sa.String, nullable=True),
|
|
||||||
sa.Index("idx_recording_meeting_id", "meeting_id"),
|
sa.Index("idx_recording_meeting_id", "meeting_id"),
|
||||||
sa.Index("idx_recording_workflow_id", "workflow_id"),
|
|
||||||
)
|
)
|
||||||
|
|
||||||
|
|
||||||
@@ -40,8 +38,6 @@ class Recording(BaseModel):
|
|||||||
# track_keys can be empty list [] if recording finished but no audio was captured (silence/muted)
|
# track_keys can be empty list [] if recording finished but no audio was captured (silence/muted)
|
||||||
# None means not a multitrack recording, [] means multitrack with no tracks
|
# None means not a multitrack recording, [] means multitrack with no tracks
|
||||||
track_keys: list[str] | None = None
|
track_keys: list[str] | None = None
|
||||||
# Conductor workflow ID for tracking pipeline execution
|
|
||||||
workflow_id: str | None = None
|
|
||||||
|
|
||||||
@property
|
@property
|
||||||
def is_multitrack(self) -> bool:
|
def is_multitrack(self) -> bool:
|
||||||
|
|||||||
@@ -12,7 +12,6 @@ from typing import Literal, Union, assert_never
|
|||||||
import celery
|
import celery
|
||||||
from celery.result import AsyncResult
|
from celery.result import AsyncResult
|
||||||
|
|
||||||
from reflector.conductor.client import ConductorClientManager
|
|
||||||
from reflector.db.recordings import recordings_controller
|
from reflector.db.recordings import recordings_controller
|
||||||
from reflector.db.transcripts import Transcript
|
from reflector.db.transcripts import Transcript
|
||||||
from reflector.hatchet.client import HatchetClientManager
|
from reflector.hatchet.client import HatchetClientManager
|
||||||
@@ -263,26 +262,6 @@ def dispatch_transcript_processing(
|
|||||||
logger.info("Hatchet workflow dispatched", workflow_id=workflow_id)
|
logger.info("Hatchet workflow dispatched", workflow_id=workflow_id)
|
||||||
durable_started = True
|
durable_started = True
|
||||||
|
|
||||||
elif settings.CONDUCTOR_ENABLED:
|
|
||||||
workflow_id = ConductorClientManager.start_workflow(
|
|
||||||
name="diarization_pipeline",
|
|
||||||
version=1,
|
|
||||||
input_data={
|
|
||||||
"recording_id": config.recording_id,
|
|
||||||
"room_name": None, # Not available in reprocess path
|
|
||||||
"tracks": [{"s3_key": k} for k in config.track_keys],
|
|
||||||
"bucket_name": config.bucket_name,
|
|
||||||
"transcript_id": config.transcript_id,
|
|
||||||
"room_id": config.room_id,
|
|
||||||
},
|
|
||||||
)
|
|
||||||
logger.info(
|
|
||||||
"Started Conductor workflow (reprocess)",
|
|
||||||
workflow_id=workflow_id,
|
|
||||||
transcript_id=config.transcript_id,
|
|
||||||
)
|
|
||||||
durable_started = True
|
|
||||||
|
|
||||||
# If durable workflow started and not in shadow mode, skip Celery
|
# If durable workflow started and not in shadow mode, skip Celery
|
||||||
if durable_started and not settings.DURABLE_WORKFLOW_SHADOW_MODE:
|
if durable_started and not settings.DURABLE_WORKFLOW_SHADOW_MODE:
|
||||||
return None
|
return None
|
||||||
|
|||||||
@@ -151,33 +151,19 @@ class Settings(BaseSettings):
|
|||||||
ZULIP_BOT_EMAIL: str | None = None
|
ZULIP_BOT_EMAIL: str | None = None
|
||||||
|
|
||||||
# Durable workflow orchestration
|
# Durable workflow orchestration
|
||||||
# Provider: "hatchet" or "conductor" (or "none" to disable)
|
# Provider: "hatchet" (or "none" to disable)
|
||||||
DURABLE_WORKFLOW_PROVIDER: str = "none"
|
DURABLE_WORKFLOW_PROVIDER: str = "none"
|
||||||
DURABLE_WORKFLOW_SHADOW_MODE: bool = False # Run both provider + Celery
|
DURABLE_WORKFLOW_SHADOW_MODE: bool = False # Run both provider + Celery
|
||||||
|
|
||||||
# Conductor workflow orchestration
|
|
||||||
CONDUCTOR_SERVER_URL: str = "http://conductor:8080/api"
|
|
||||||
CONDUCTOR_DEBUG: bool = False
|
|
||||||
|
|
||||||
# Hatchet workflow orchestration
|
# Hatchet workflow orchestration
|
||||||
HATCHET_CLIENT_TOKEN: str | None = None
|
HATCHET_CLIENT_TOKEN: str | None = None
|
||||||
HATCHET_CLIENT_TLS_STRATEGY: str = "none" # none, tls, mtls
|
HATCHET_CLIENT_TLS_STRATEGY: str = "none" # none, tls, mtls
|
||||||
HATCHET_DEBUG: bool = False
|
HATCHET_DEBUG: bool = False
|
||||||
|
|
||||||
@property
|
|
||||||
def CONDUCTOR_ENABLED(self) -> bool:
|
|
||||||
"""Legacy compatibility: True if Conductor is the active provider."""
|
|
||||||
return self.DURABLE_WORKFLOW_PROVIDER == "conductor"
|
|
||||||
|
|
||||||
@property
|
@property
|
||||||
def HATCHET_ENABLED(self) -> bool:
|
def HATCHET_ENABLED(self) -> bool:
|
||||||
"""True if Hatchet is the active provider."""
|
"""True if Hatchet is the active provider."""
|
||||||
return self.DURABLE_WORKFLOW_PROVIDER == "hatchet"
|
return self.DURABLE_WORKFLOW_PROVIDER == "hatchet"
|
||||||
|
|
||||||
@property
|
|
||||||
def CONDUCTOR_SHADOW_MODE(self) -> bool:
|
|
||||||
"""Legacy compatibility for shadow mode."""
|
|
||||||
return self.DURABLE_WORKFLOW_SHADOW_MODE and self.CONDUCTOR_ENABLED
|
|
||||||
|
|
||||||
|
|
||||||
settings = Settings()
|
settings = Settings()
|
||||||
|
|||||||
@@ -1,45 +0,0 @@
|
|||||||
"""Conductor health and status endpoints."""
|
|
||||||
|
|
||||||
import httpx
|
|
||||||
from fastapi import APIRouter
|
|
||||||
|
|
||||||
from reflector.settings import settings
|
|
||||||
|
|
||||||
router = APIRouter(prefix="/conductor", tags=["conductor"])
|
|
||||||
|
|
||||||
|
|
||||||
@router.get("/health")
|
|
||||||
async def conductor_health():
|
|
||||||
"""Check Conductor server connectivity and status."""
|
|
||||||
if not settings.CONDUCTOR_ENABLED:
|
|
||||||
return {"status": "disabled", "connected": False}
|
|
||||||
|
|
||||||
# Extract base URL (remove /api suffix for health check)
|
|
||||||
base_url = settings.CONDUCTOR_SERVER_URL.rstrip("/api").rstrip("/")
|
|
||||||
health_url = f"{base_url}/health"
|
|
||||||
|
|
||||||
try:
|
|
||||||
async with httpx.AsyncClient(timeout=5.0) as client:
|
|
||||||
resp = await client.get(health_url)
|
|
||||||
if resp.status_code == 200:
|
|
||||||
return {"status": "healthy", "connected": True}
|
|
||||||
else:
|
|
||||||
return {
|
|
||||||
"status": "unhealthy",
|
|
||||||
"connected": True,
|
|
||||||
"error": f"Health check returned {resp.status_code}",
|
|
||||||
}
|
|
||||||
except httpx.TimeoutException:
|
|
||||||
return {
|
|
||||||
"status": "unhealthy",
|
|
||||||
"connected": False,
|
|
||||||
"error": "Connection timeout",
|
|
||||||
}
|
|
||||||
except httpx.ConnectError as e:
|
|
||||||
return {
|
|
||||||
"status": "unhealthy",
|
|
||||||
"connected": False,
|
|
||||||
"error": f"Connection failed: {e}",
|
|
||||||
}
|
|
||||||
except Exception as e:
|
|
||||||
return {"status": "unhealthy", "connected": False, "error": str(e)}
|
|
||||||
@@ -286,7 +286,7 @@ async def _process_multitrack_recording_inner(
|
|||||||
room_id=room.id,
|
room_id=room.id,
|
||||||
)
|
)
|
||||||
|
|
||||||
# Start durable workflow if enabled (Hatchet or Conductor)
|
# Start durable workflow if enabled (Hatchet)
|
||||||
durable_started = False
|
durable_started = False
|
||||||
|
|
||||||
if settings.HATCHET_ENABLED:
|
if settings.HATCHET_ENABLED:
|
||||||
@@ -309,33 +309,10 @@ async def _process_multitrack_recording_inner(
|
|||||||
transcript_id=transcript.id,
|
transcript_id=transcript.id,
|
||||||
)
|
)
|
||||||
|
|
||||||
# Store workflow_id on recording for status tracking
|
# Store workflow_run_id on transcript for replay/resume
|
||||||
await recordings_controller.update(recording, {"workflow_id": workflow_id})
|
await transcripts_controller.update(
|
||||||
durable_started = True
|
transcript, {"workflow_run_id": workflow_id}
|
||||||
|
|
||||||
elif settings.CONDUCTOR_ENABLED:
|
|
||||||
from reflector.conductor.client import ConductorClientManager # noqa: PLC0415
|
|
||||||
|
|
||||||
workflow_id = ConductorClientManager.start_workflow(
|
|
||||||
name="diarization_pipeline",
|
|
||||||
version=1,
|
|
||||||
input_data={
|
|
||||||
"recording_id": recording_id,
|
|
||||||
"room_name": daily_room_name,
|
|
||||||
"tracks": [{"s3_key": k} for k in filter_cam_audio_tracks(track_keys)],
|
|
||||||
"bucket_name": bucket_name,
|
|
||||||
"transcript_id": transcript.id,
|
|
||||||
"room_id": room.id,
|
|
||||||
},
|
|
||||||
)
|
)
|
||||||
logger.info(
|
|
||||||
"Started Conductor workflow",
|
|
||||||
workflow_id=workflow_id,
|
|
||||||
transcript_id=transcript.id,
|
|
||||||
)
|
|
||||||
|
|
||||||
# Store workflow_id on recording for status tracking
|
|
||||||
await recordings_controller.update(recording, {"workflow_id": workflow_id})
|
|
||||||
durable_started = True
|
durable_started = True
|
||||||
|
|
||||||
# If durable workflow started and not in shadow mode, skip Celery
|
# If durable workflow started and not in shadow mode, skip Celery
|
||||||
|
|||||||
@@ -7,8 +7,6 @@ elif [ "${ENTRYPOINT}" = "worker" ]; then
|
|||||||
uv run celery -A reflector.worker.app worker --loglevel=info
|
uv run celery -A reflector.worker.app worker --loglevel=info
|
||||||
elif [ "${ENTRYPOINT}" = "beat" ]; then
|
elif [ "${ENTRYPOINT}" = "beat" ]; then
|
||||||
uv run celery -A reflector.worker.app beat --loglevel=info
|
uv run celery -A reflector.worker.app beat --loglevel=info
|
||||||
elif [ "${ENTRYPOINT}" = "conductor-worker" ]; then
|
|
||||||
uv run python -m reflector.conductor.run_workers
|
|
||||||
elif [ "${ENTRYPOINT}" = "hatchet-worker" ]; then
|
elif [ "${ENTRYPOINT}" = "hatchet-worker" ]; then
|
||||||
uv run python -m reflector.hatchet.run_workers
|
uv run python -m reflector.hatchet.run_workers
|
||||||
else
|
else
|
||||||
|
|||||||
59
server/tests/test_hatchet_client.py
Normal file
59
server/tests/test_hatchet_client.py
Normal file
@@ -0,0 +1,59 @@
|
|||||||
|
"""
|
||||||
|
Tests for HatchetClientManager error handling and validation.
|
||||||
|
|
||||||
|
Only tests that catch real bugs - not mock verification tests.
|
||||||
|
"""
|
||||||
|
|
||||||
|
from unittest.mock import AsyncMock, MagicMock, patch
|
||||||
|
|
||||||
|
import pytest
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_client_can_replay_handles_exception():
|
||||||
|
"""Test can_replay returns False when status check fails.
|
||||||
|
|
||||||
|
Useful: Ensures network/API errors don't crash the system and
|
||||||
|
gracefully allow reprocessing when workflow state is unknown.
|
||||||
|
"""
|
||||||
|
from reflector.hatchet.client import HatchetClientManager
|
||||||
|
|
||||||
|
HatchetClientManager._instance = None
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_CLIENT_TOKEN = "test-token"
|
||||||
|
mock_settings.HATCHET_DEBUG = False
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.Hatchet") as mock_hatchet_class:
|
||||||
|
mock_client = MagicMock()
|
||||||
|
mock_hatchet_class.return_value = mock_client
|
||||||
|
|
||||||
|
mock_client.runs.aio_get_status = AsyncMock(
|
||||||
|
side_effect=Exception("Network error")
|
||||||
|
)
|
||||||
|
|
||||||
|
can_replay = await HatchetClientManager.can_replay("workflow-123")
|
||||||
|
|
||||||
|
# Should return False on error (workflow might be gone)
|
||||||
|
assert can_replay is False
|
||||||
|
|
||||||
|
HatchetClientManager._instance = None
|
||||||
|
|
||||||
|
|
||||||
|
def test_hatchet_client_raises_without_token():
|
||||||
|
"""Test that get_client raises ValueError without token.
|
||||||
|
|
||||||
|
Useful: Catches if someone removes the token validation,
|
||||||
|
which would cause cryptic errors later.
|
||||||
|
"""
|
||||||
|
from reflector.hatchet.client import HatchetClientManager
|
||||||
|
|
||||||
|
HatchetClientManager._instance = None
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_CLIENT_TOKEN = None
|
||||||
|
|
||||||
|
with pytest.raises(ValueError, match="HATCHET_CLIENT_TOKEN must be set"):
|
||||||
|
HatchetClientManager.get_client()
|
||||||
|
|
||||||
|
HatchetClientManager._instance = None
|
||||||
376
server/tests/test_hatchet_dispatch.py
Normal file
376
server/tests/test_hatchet_dispatch.py
Normal file
@@ -0,0 +1,376 @@
|
|||||||
|
"""
|
||||||
|
Tests for Hatchet workflow dispatch and routing logic.
|
||||||
|
|
||||||
|
These tests verify:
|
||||||
|
1. Routing to Hatchet when HATCHET_ENABLED=True
|
||||||
|
2. Replay logic for failed workflows
|
||||||
|
3. Force flag to cancel and restart
|
||||||
|
4. Validation prevents concurrent workflows
|
||||||
|
"""
|
||||||
|
|
||||||
|
from unittest.mock import AsyncMock, patch
|
||||||
|
|
||||||
|
import pytest
|
||||||
|
|
||||||
|
from reflector.db.transcripts import Transcript
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_blocks_running_workflow():
|
||||||
|
"""Test that validation blocks reprocessing when workflow is running."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationAlreadyScheduled,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="processing",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id="running-workflow-123",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = True
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.HatchetClientManager") as mock_hatchet:
|
||||||
|
mock_hatchet.get_workflow_run_status = AsyncMock(return_value="RUNNING")
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
assert isinstance(result, ValidationAlreadyScheduled)
|
||||||
|
assert "running" in result.detail.lower()
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_blocks_queued_workflow():
|
||||||
|
"""Test that validation blocks reprocessing when workflow is queued."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationAlreadyScheduled,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="processing",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id="queued-workflow-123",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = True
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.HatchetClientManager") as mock_hatchet:
|
||||||
|
mock_hatchet.get_workflow_run_status = AsyncMock(return_value="QUEUED")
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
assert isinstance(result, ValidationAlreadyScheduled)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_allows_failed_workflow():
|
||||||
|
"""Test that validation allows reprocessing when workflow has failed."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationOk,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="error",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id="failed-workflow-123",
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = True
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.HatchetClientManager") as mock_hatchet:
|
||||||
|
mock_hatchet.get_workflow_run_status = AsyncMock(return_value="FAILED")
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
assert isinstance(result, ValidationOk)
|
||||||
|
assert result.transcript_id == "test-transcript-id"
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_allows_completed_workflow():
|
||||||
|
"""Test that validation allows reprocessing when workflow has completed."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationOk,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="ended",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id="completed-workflow-123",
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = True
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.HatchetClientManager") as mock_hatchet:
|
||||||
|
mock_hatchet.get_workflow_run_status = AsyncMock(return_value="COMPLETED")
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
assert isinstance(result, ValidationOk)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_allows_when_status_check_fails():
|
||||||
|
"""Test that validation allows reprocessing when status check fails (workflow might be gone)."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationOk,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="error",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id="old-workflow-123",
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = True
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.HatchetClientManager") as mock_hatchet:
|
||||||
|
# Status check fails (workflow might be deleted)
|
||||||
|
mock_hatchet.get_workflow_run_status = AsyncMock(
|
||||||
|
side_effect=Exception("Workflow not found")
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
# Should allow processing when we can't get status
|
||||||
|
assert isinstance(result, ValidationOk)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_skipped_when_no_workflow_id():
|
||||||
|
"""Test that Hatchet validation is skipped when transcript has no workflow_run_id."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationOk,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="uploaded",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id=None, # No workflow yet
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = True
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.client.HatchetClientManager") as mock_hatchet:
|
||||||
|
# Should not be called
|
||||||
|
mock_hatchet.get_workflow_run_status = AsyncMock()
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
# Should not check Hatchet status
|
||||||
|
mock_hatchet.get_workflow_run_status.assert_not_called()
|
||||||
|
assert isinstance(result, ValidationOk)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_hatchet_validation_skipped_when_disabled():
|
||||||
|
"""Test that Hatchet validation is skipped when HATCHET_ENABLED is False."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationOk,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="uploaded",
|
||||||
|
source_kind="room",
|
||||||
|
workflow_run_id="some-workflow-123",
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch("reflector.services.transcript_process.settings") as mock_settings:
|
||||||
|
mock_settings.HATCHET_ENABLED = False # Hatchet disabled
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.task_is_scheduled_or_active"
|
||||||
|
) as mock_celery_check:
|
||||||
|
mock_celery_check.return_value = False
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
# Should not check Hatchet at all
|
||||||
|
assert isinstance(result, ValidationOk)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_validation_locked_transcript():
|
||||||
|
"""Test that validation rejects locked transcripts."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationLocked,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="ended",
|
||||||
|
source_kind="room",
|
||||||
|
locked=True,
|
||||||
|
)
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
assert isinstance(result, ValidationLocked)
|
||||||
|
assert "locked" in result.detail.lower()
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_validation_idle_transcript():
|
||||||
|
"""Test that validation rejects idle transcripts (not ready)."""
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
ValidationNotReady,
|
||||||
|
validate_transcript_for_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_transcript = Transcript(
|
||||||
|
id="test-transcript-id",
|
||||||
|
name="Test",
|
||||||
|
status="idle",
|
||||||
|
source_kind="room",
|
||||||
|
)
|
||||||
|
|
||||||
|
result = await validate_transcript_for_processing(mock_transcript)
|
||||||
|
|
||||||
|
assert isinstance(result, ValidationNotReady)
|
||||||
|
assert "not ready" in result.detail.lower()
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_prepare_multitrack_config():
|
||||||
|
"""Test preparing multitrack processing config."""
|
||||||
|
from reflector.db.recordings import Recording
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
MultitrackProcessingConfig,
|
||||||
|
ValidationOk,
|
||||||
|
prepare_transcript_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
validation = ValidationOk(
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
transcript_id="test-transcript-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_recording = Recording(
|
||||||
|
id="test-recording-id",
|
||||||
|
bucket_name="test-bucket",
|
||||||
|
object_key="recordings/test",
|
||||||
|
recorded_at="2024-01-01T00:00:00Z",
|
||||||
|
track_keys=["track1.webm", "track2.webm"],
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.recordings_controller"
|
||||||
|
) as mock_rc:
|
||||||
|
mock_rc.get_by_id = AsyncMock(return_value=mock_recording)
|
||||||
|
|
||||||
|
result = await prepare_transcript_processing(validation, room_id="test-room")
|
||||||
|
|
||||||
|
assert isinstance(result, MultitrackProcessingConfig)
|
||||||
|
assert result.bucket_name == "test-bucket"
|
||||||
|
assert result.track_keys == ["track1.webm", "track2.webm"]
|
||||||
|
assert result.transcript_id == "test-transcript-id"
|
||||||
|
assert result.room_id == "test-room"
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.usefixtures("setup_database")
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_prepare_file_config():
|
||||||
|
"""Test preparing file processing config (no track keys)."""
|
||||||
|
from reflector.db.recordings import Recording
|
||||||
|
from reflector.services.transcript_process import (
|
||||||
|
FileProcessingConfig,
|
||||||
|
ValidationOk,
|
||||||
|
prepare_transcript_processing,
|
||||||
|
)
|
||||||
|
|
||||||
|
validation = ValidationOk(
|
||||||
|
recording_id="test-recording-id",
|
||||||
|
transcript_id="test-transcript-id",
|
||||||
|
)
|
||||||
|
|
||||||
|
mock_recording = Recording(
|
||||||
|
id="test-recording-id",
|
||||||
|
bucket_name="test-bucket",
|
||||||
|
object_key="recordings/test.mp4",
|
||||||
|
recorded_at="2024-01-01T00:00:00Z",
|
||||||
|
track_keys=None, # No track keys = file pipeline
|
||||||
|
)
|
||||||
|
|
||||||
|
with patch(
|
||||||
|
"reflector.services.transcript_process.recordings_controller"
|
||||||
|
) as mock_rc:
|
||||||
|
mock_rc.get_by_id = AsyncMock(return_value=mock_recording)
|
||||||
|
|
||||||
|
result = await prepare_transcript_processing(validation)
|
||||||
|
|
||||||
|
assert isinstance(result, FileProcessingConfig)
|
||||||
|
assert result.transcript_id == "test-transcript-id"
|
||||||
62
server/tests/test_hatchet_progress.py
Normal file
62
server/tests/test_hatchet_progress.py
Normal file
@@ -0,0 +1,62 @@
|
|||||||
|
"""
|
||||||
|
Tests for Hatchet progress emission.
|
||||||
|
|
||||||
|
Only tests that catch real bugs - error handling and step completeness.
|
||||||
|
"""
|
||||||
|
|
||||||
|
from unittest.mock import AsyncMock, MagicMock, patch
|
||||||
|
|
||||||
|
import pytest
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_emit_progress_async_handles_exception():
|
||||||
|
"""Test that emit_progress_async catches exceptions gracefully.
|
||||||
|
|
||||||
|
Critical: Progress emission must NEVER crash the pipeline.
|
||||||
|
WebSocket errors should be silently caught.
|
||||||
|
"""
|
||||||
|
from reflector.hatchet.progress import emit_progress_async
|
||||||
|
|
||||||
|
with patch("reflector.hatchet.progress.get_ws_manager") as mock_get_ws:
|
||||||
|
mock_ws = MagicMock()
|
||||||
|
mock_ws.send_json = AsyncMock(side_effect=Exception("WebSocket error"))
|
||||||
|
mock_get_ws.return_value = mock_ws
|
||||||
|
|
||||||
|
# Should not raise - exceptions are caught
|
||||||
|
await emit_progress_async(
|
||||||
|
transcript_id="test-transcript-123",
|
||||||
|
step="finalize",
|
||||||
|
status="completed",
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
|
@pytest.mark.asyncio
|
||||||
|
async def test_pipeline_steps_mapping_complete():
|
||||||
|
"""Test the PIPELINE_STEPS mapping includes all expected steps.
|
||||||
|
|
||||||
|
Useful: Catches when someone adds a new pipeline step but forgets
|
||||||
|
to add it to the progress mapping, resulting in missing UI updates.
|
||||||
|
"""
|
||||||
|
from reflector.hatchet.progress import PIPELINE_STEPS, TOTAL_STEPS
|
||||||
|
|
||||||
|
expected_steps = [
|
||||||
|
"get_recording",
|
||||||
|
"get_participants",
|
||||||
|
"pad_track",
|
||||||
|
"mixdown_tracks",
|
||||||
|
"generate_waveform",
|
||||||
|
"transcribe_track",
|
||||||
|
"merge_transcripts",
|
||||||
|
"detect_topics",
|
||||||
|
"generate_title",
|
||||||
|
"generate_summary",
|
||||||
|
"finalize",
|
||||||
|
"cleanup_consent",
|
||||||
|
"post_zulip",
|
||||||
|
"send_webhook",
|
||||||
|
]
|
||||||
|
|
||||||
|
for step in expected_steps:
|
||||||
|
assert step in PIPELINE_STEPS, f"Missing step in PIPELINE_STEPS: {step}"
|
||||||
|
assert 1 <= PIPELINE_STEPS[step] <= TOTAL_STEPS
|
||||||
50
server/uv.lock
generated
50
server/uv.lock
generated
@@ -283,15 +283,6 @@ wheels = [
|
|||||||
{ url = "https://files.pythonhosted.org/packages/c5/7c/83ff6046176a675e6a1e8aeefed8892cd97fe7c46af93cc540d1b24b8323/asteroid_filterbanks-0.4.0-py3-none-any.whl", hash = "sha256:4932ac8b6acc6e08fb87cbe8ece84215b5a74eee284fe83acf3540a72a02eaf5", size = 29912 },
|
{ url = "https://files.pythonhosted.org/packages/c5/7c/83ff6046176a675e6a1e8aeefed8892cd97fe7c46af93cc540d1b24b8323/asteroid_filterbanks-0.4.0-py3-none-any.whl", hash = "sha256:4932ac8b6acc6e08fb87cbe8ece84215b5a74eee284fe83acf3540a72a02eaf5", size = 29912 },
|
||||||
]
|
]
|
||||||
|
|
||||||
[[package]]
|
|
||||||
name = "astor"
|
|
||||||
version = "0.8.1"
|
|
||||||
source = { registry = "https://pypi.org/simple" }
|
|
||||||
sdist = { url = "https://files.pythonhosted.org/packages/5a/21/75b771132fee241dfe601d39ade629548a9626d1d39f333fde31bc46febe/astor-0.8.1.tar.gz", hash = "sha256:6a6effda93f4e1ce9f618779b2dd1d9d84f1e32812c23a29b3fff6fd7f63fa5e", size = 35090 }
|
|
||||||
wheels = [
|
|
||||||
{ url = "https://files.pythonhosted.org/packages/c3/88/97eef84f48fa04fbd6750e62dcceafba6c63c81b7ac1420856c8dcc0a3f9/astor-0.8.1-py2.py3-none-any.whl", hash = "sha256:070a54e890cefb5b3739d19f30f5a5ec840ffc9c50ffa7d23cc9fc1a38ebbfc5", size = 27488 },
|
|
||||||
]
|
|
||||||
|
|
||||||
[[package]]
|
[[package]]
|
||||||
name = "async-timeout"
|
name = "async-timeout"
|
||||||
version = "5.0.1"
|
version = "5.0.1"
|
||||||
@@ -629,27 +620,6 @@ wheels = [
|
|||||||
{ url = "https://files.pythonhosted.org/packages/e3/51/9b208e85196941db2f0654ad0357ca6388ab3ed67efdbfc799f35d1f83aa/colorlog-6.9.0-py3-none-any.whl", hash = "sha256:5906e71acd67cb07a71e779c47c4bcb45fb8c2993eebe9e5adcd6a6f1b283eff", size = 11424 },
|
{ url = "https://files.pythonhosted.org/packages/e3/51/9b208e85196941db2f0654ad0357ca6388ab3ed67efdbfc799f35d1f83aa/colorlog-6.9.0-py3-none-any.whl", hash = "sha256:5906e71acd67cb07a71e779c47c4bcb45fb8c2993eebe9e5adcd6a6f1b283eff", size = 11424 },
|
||||||
]
|
]
|
||||||
|
|
||||||
[[package]]
|
|
||||||
name = "conductor-python"
|
|
||||||
version = "1.2.3"
|
|
||||||
source = { registry = "https://pypi.org/simple" }
|
|
||||||
dependencies = [
|
|
||||||
{ name = "astor" },
|
|
||||||
{ name = "certifi" },
|
|
||||||
{ name = "dacite" },
|
|
||||||
{ name = "deprecated" },
|
|
||||||
{ name = "prometheus-client" },
|
|
||||||
{ name = "python-dateutil" },
|
|
||||||
{ name = "requests" },
|
|
||||||
{ name = "shortuuid" },
|
|
||||||
{ name = "six" },
|
|
||||||
{ name = "typing-extensions" },
|
|
||||||
]
|
|
||||||
sdist = { url = "https://files.pythonhosted.org/packages/8f/cb/b20991a9b4da4fa6e21cb596ef5b3a7ed61ce2a00a6ec63a901c01f01272/conductor_python-1.2.3.tar.gz", hash = "sha256:50c5e8414e6606a964ecf92b263f1a6af397ad6948640cc8e1e953b2a1615777", size = 175152 }
|
|
||||||
wheels = [
|
|
||||||
{ url = "https://files.pythonhosted.org/packages/ba/a1/8aeed6eb02d80bc86ea41b807e36d765c79943aaccd2247912181e21d0c9/conductor_python-1.2.3-py3-none-any.whl", hash = "sha256:07feb303d4732916c979e8d5567259b1ff7ff0d4c6b2f5f627473cfeaa026b94", size = 296261 },
|
|
||||||
]
|
|
||||||
|
|
||||||
[[package]]
|
[[package]]
|
||||||
name = "contourpy"
|
name = "contourpy"
|
||||||
version = "1.3.3"
|
version = "1.3.3"
|
||||||
@@ -797,15 +767,6 @@ wheels = [
|
|||||||
{ url = "https://files.pythonhosted.org/packages/e7/05/c19819d5e3d95294a6f5947fb9b9629efb316b96de511b418c53d245aae6/cycler-0.12.1-py3-none-any.whl", hash = "sha256:85cef7cff222d8644161529808465972e51340599459b8ac3ccbac5a854e0d30", size = 8321 },
|
{ url = "https://files.pythonhosted.org/packages/e7/05/c19819d5e3d95294a6f5947fb9b9629efb316b96de511b418c53d245aae6/cycler-0.12.1-py3-none-any.whl", hash = "sha256:85cef7cff222d8644161529808465972e51340599459b8ac3ccbac5a854e0d30", size = 8321 },
|
||||||
]
|
]
|
||||||
|
|
||||||
[[package]]
|
|
||||||
name = "dacite"
|
|
||||||
version = "1.9.2"
|
|
||||||
source = { registry = "https://pypi.org/simple" }
|
|
||||||
sdist = { url = "https://files.pythonhosted.org/packages/55/a0/7ca79796e799a3e782045d29bf052b5cde7439a2bbb17f15ff44f7aacc63/dacite-1.9.2.tar.gz", hash = "sha256:6ccc3b299727c7aa17582f0021f6ae14d5de47c7227932c47fec4cdfefd26f09", size = 22420 }
|
|
||||||
wheels = [
|
|
||||||
{ url = "https://files.pythonhosted.org/packages/94/35/386550fd60316d1e37eccdda609b074113298f23cef5bddb2049823fe666/dacite-1.9.2-py3-none-any.whl", hash = "sha256:053f7c3f5128ca2e9aceb66892b1a3c8936d02c686e707bee96e19deef4bc4a0", size = 16600 },
|
|
||||||
]
|
|
||||||
|
|
||||||
[[package]]
|
[[package]]
|
||||||
name = "databases"
|
name = "databases"
|
||||||
version = "0.8.0"
|
version = "0.8.0"
|
||||||
@@ -3231,7 +3192,6 @@ dependencies = [
|
|||||||
{ name = "alembic" },
|
{ name = "alembic" },
|
||||||
{ name = "av" },
|
{ name = "av" },
|
||||||
{ name = "celery" },
|
{ name = "celery" },
|
||||||
{ name = "conductor-python" },
|
|
||||||
{ name = "databases", extra = ["aiosqlite", "asyncpg"] },
|
{ name = "databases", extra = ["aiosqlite", "asyncpg"] },
|
||||||
{ name = "fastapi", extra = ["standard"] },
|
{ name = "fastapi", extra = ["standard"] },
|
||||||
{ name = "fastapi-pagination" },
|
{ name = "fastapi-pagination" },
|
||||||
@@ -3309,7 +3269,6 @@ requires-dist = [
|
|||||||
{ name = "alembic", specifier = ">=1.11.3" },
|
{ name = "alembic", specifier = ">=1.11.3" },
|
||||||
{ name = "av", specifier = ">=10.0.0" },
|
{ name = "av", specifier = ">=10.0.0" },
|
||||||
{ name = "celery", specifier = ">=5.3.4" },
|
{ name = "celery", specifier = ">=5.3.4" },
|
||||||
{ name = "conductor-python", specifier = ">=1.2.3" },
|
|
||||||
{ name = "databases", extras = ["aiosqlite", "asyncpg"], specifier = ">=0.7.0" },
|
{ name = "databases", extras = ["aiosqlite", "asyncpg"], specifier = ">=0.7.0" },
|
||||||
{ name = "fastapi", extras = ["standard"], specifier = ">=0.100.1" },
|
{ name = "fastapi", extras = ["standard"], specifier = ">=0.100.1" },
|
||||||
{ name = "fastapi-pagination", specifier = ">=0.12.6" },
|
{ name = "fastapi-pagination", specifier = ">=0.12.6" },
|
||||||
@@ -3754,15 +3713,6 @@ wheels = [
|
|||||||
{ url = "https://files.pythonhosted.org/packages/e0/f9/0595336914c5619e5f28a1fb793285925a8cd4b432c9da0a987836c7f822/shellingham-1.5.4-py2.py3-none-any.whl", hash = "sha256:7ecfff8f2fd72616f7481040475a65b2bf8af90a56c89140852d1120324e8686", size = 9755 },
|
{ url = "https://files.pythonhosted.org/packages/e0/f9/0595336914c5619e5f28a1fb793285925a8cd4b432c9da0a987836c7f822/shellingham-1.5.4-py2.py3-none-any.whl", hash = "sha256:7ecfff8f2fd72616f7481040475a65b2bf8af90a56c89140852d1120324e8686", size = 9755 },
|
||||||
]
|
]
|
||||||
|
|
||||||
[[package]]
|
|
||||||
name = "shortuuid"
|
|
||||||
version = "1.0.13"
|
|
||||||
source = { registry = "https://pypi.org/simple" }
|
|
||||||
sdist = { url = "https://files.pythonhosted.org/packages/8c/e2/bcf761f3bff95856203f9559baf3741c416071dd200c0fc19fad7f078f86/shortuuid-1.0.13.tar.gz", hash = "sha256:3bb9cf07f606260584b1df46399c0b87dd84773e7b25912b7e391e30797c5e72", size = 9662 }
|
|
||||||
wheels = [
|
|
||||||
{ url = "https://files.pythonhosted.org/packages/c0/44/21d6bf170bf40b41396480d8d49ad640bca3f2b02139cd52aa1e272830a5/shortuuid-1.0.13-py3-none-any.whl", hash = "sha256:a482a497300b49b4953e15108a7913244e1bb0d41f9d332f5e9925dba33a3c5a", size = 10529 },
|
|
||||||
]
|
|
||||||
|
|
||||||
[[package]]
|
[[package]]
|
||||||
name = "silero-vad"
|
name = "silero-vad"
|
||||||
version = "6.0.0"
|
version = "6.0.0"
|
||||||
|
|||||||
Reference in New Issue
Block a user