selfhostyourtech/reflector
2
0
Fork 0
mirror of https://github.com/Monadical-SAS/reflector.git synced 2026-02-06 18:56:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: parallelize hatchet (#804)

Browse Source 
* parallelize hatchet (no-mistakes)

* dry (no-mistakes) (minimal)

* comments

* self-review

* self-review

* self-review

* self-review

* pr comments

* pr comments

---------

Co-authored-by: Igor Loskutov <igor.loskutoff@gmail.com>
This commit is contained in:
Igor Monadical
2025-12-23 11:03:36 -05:00
committed by GitHub
parent 7c2d0698ed
commit 594bcc09e0
15 changed files with 849 additions and 287 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
server/reflector/hatchet/workflows/track_processing.py
View File

@@ -23,6 +23,7 @@ from hatchet_sdk import Context
from pydantic import BaseModel
from reflector.hatchet.client import HatchetClientManager
from reflector.hatchet.constants import TIMEOUT_AUDIO, TIMEOUT_HEAVY
from reflector.hatchet.workflows.models import PadTrackResult, TranscribeTrackResult
from reflector.logger import logger
from reflector.utils.audio_constants import PRESIGNED_URL_EXPIRATION_SECONDS
@@ -47,7 +48,7 @@ hatchet = HatchetClientManager.get_client()
track_workflow = hatchet.workflow(name="TrackProcessing", input_validator=TrackInput)
@track_workflow.task(execution_timeout=timedelta(seconds=300), retries=3)
@track_workflow.task(execution_timeout=timedelta(seconds=TIMEOUT_AUDIO), retries=3)
async def pad_track(input: TrackInput, ctx: Context) -> PadTrackResult:
"""Pad single audio track with silence for alignment.
@@ -153,7 +154,7 @@ async def pad_track(input: TrackInput, ctx: Context) -> PadTrackResult:
@track_workflow.task(
parents=[pad_track], execution_timeout=timedelta(seconds=600), retries=3
parents=[pad_track], execution_timeout=timedelta(seconds=TIMEOUT_HEAVY), retries=3
)
async def transcribe_track(input: TrackInput, ctx: Context) -> TranscribeTrackResult:
"""Transcribe audio track using GPU (Modal.com) or local Whisper."""
@@ -197,23 +198,20 @@ async def transcribe_track(input: TrackInput, ctx: Context) -> TranscribeTrackRe
transcript = await transcribe_file_with_processor(audio_url, input.language)
# Tag all words with speaker index
words = []
for word in transcript.words:
word_dict = word.model_dump()
word_dict["speaker"] = input.track_index
words.append(word_dict)
word.speaker = input.track_index
ctx.log(
f"transcribe_track complete: track {input.track_index}, {len(words)} words"
f"transcribe_track complete: track {input.track_index}, {len(transcript.words)} words"
)
logger.info(
"[Hatchet] transcribe_track complete",
track_index=input.track_index,
word_count=len(words),
word_count=len(transcript.words),
)
return TranscribeTrackResult(
words=words,
words=transcript.words,
track_index=input.track_index,
)