monadical/internalai-agent
13
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: migrate to skills-based approach

Browse Source
This commit is contained in:
Mathieu Virbel
2026-02-19 11:36:32 -06:00
parent 42a9b5289f
commit eefac81e57
15 changed files with 2565 additions and 224 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
.agents/skills/checkout/SKILL.md Normal file
View File

@@ -0,0 +1,52 @@
---
name: checkout
description: Build a weekly checkout/review covering Sunday through today. Gathers meetings, emails, Zulip conversations, and Gitea activity, then produces a structured summary.
disable-model-invocation: true
---
# Weekly Review Builder
Build my weekly checkout covering Sunday through today.
1. **Get my identity** with `contactdb_get_me` to obtain my contact_id
2. **Determine date range**: Sunday to today (use `date -d "last sunday" +%Y-%m-%d`)
3. **Gather activity in parallel**:
- **Dataindex**: Launch **one subagent per day** (Sunday through today). Each subagent should query `dataindex_query_entities` for that specific day with my contact_id, looking for meetings, calendar events, emails, documents. Return day-by-day summary.
- **Threaded Conversations**: Launch **one subagent per day** (Sunday through today). Each subagent should:
1. Query `dataindex_query_entities` for entity_type `threaded_conversation` for that specific day with my contact_id
2. For each conversation found, fetch all `conversation_message` entities using the conversation ID as parent_id filter
3. Return messages I participated in with context
- **Gitea**: Launch one subagent to run `~/bin/gitea-activity -s START -e END` and extract commits, PRs (opened/merged/approved), and repositories worked on
4. **Query dataindex directly** for the full week as backup to ensure nothing is missed
**Build the checkout with this structure:**
```
# Weekly Review: [Date Range]
## Objectives
- List 2-3 high-level goals for the week based on the main themes of work
****Major Achievements****
- Bullet points of concrete deliverables, grouped by theme
- Focus on shipped features, solved problems, infrastructure built
****Code Activity****
- Stats line: X commits across Y repositories, Z PRs total (N merged, M open)
- **New Repositories**: `[name](url)` - brief description
- **Pull Requests Merged**: `[#N Title](url)` - one per line with descriptive title
- **Pull Requests Opened (not merged)**: `[#N](url)` - include status if known (approved, draft, etc.)
****Team Interactions****
- **Meeting Type (Nx)**: Brief description of purpose/outcome
With: Key participants
- **Notable conversations**: Date, participants, main subject discussed
```
**Rules:**
- Use `****Title****` format for section headers (not ##)
- All PRs and repositories must be markdown links `[name](url)`
- List merged PRs first, then open/unmerged ones
- Only include meaningful interactions (skip routine standups unless notable decisions made)
- No "who am I" header, no summary section at the end
- Focus on outcomes and business value, not just activity volume

49
.agents/skills/company/SKILL.md Normal file
View File

@@ -0,0 +1,49 @@
---
name: company
description: Monadical company context. Use when you need to understand the organization structure, Zulip stream layout, communication tools, meeting/calendar relationships, or internal product names.
user-invocable: false
---
# Company Context
## About Monadical
Monadical is a software consultancy founded in 2016. The company operates across multiple locations: Montreal and Vancouver (Canada), and Medellin and Cali (Colombia). The team builds internal products alongside client work.
### Internal Products
- **Reflector** — Meeting recording and transcription tool (produces meeting entities in DataIndex)
- **GreyHaven / InternalAI platform** — A local-first platform that aggregates personal data, resolve contact to do automation and analysis
## Communication Tools
| Tool | Role | Data in DataIndex? |
|------------|-----------------------------|---------------------|
| Zulip | Primary internal chat | Yes (connector: `zulip`) |
| Fastmail/Email | External communication | Yes (connector: `mbsync_email`) |
| Calendar | Scheduling (ICS feeds) | Yes (connector: `ics_calendar`) |
| Reflector | Meeting recordings | Yes (connector: `reflector`) |
| HedgeDoc | Collaborative documents | Yes (connector: `hedgedoc`) |
## How the company is working
We use zulip as our main hub for communication. Zulip have channels (top level) and topic (low level). Depending the channels, differents behavior have to be adopted.
### Zulip channels
Here is a list of zulip stream prefix with context on how the company is organized:
- InternalAI (zulip:stream:193) is about this specific platform.
- Leads (zulip:stream:78) is where we talk about our leads/client. We usually create one topic per lead/client - So if you are searching information about a client, always have a look if a related topic exist, that match the client or the company name.
- Checkins (zulip:stream:24) are usually one topic per employee. This is where an employee indicate what it did or will do during a period of time, or just some status update. Not everybody is using the system on regular basis.
- Devcap (zulip:stream:156) is where we are talking about our investment / due diligence before investing. One topic per company.
- General (zulip:stream:21) is where we talk about different topic on various subject, company wide or services.
- Enginerring (zulip:stream:25) is where we talk about enginerring issue / services / new tool to try
- Learning (zulip:stream:31) is where we share links about new tools / ideas or stuff to learn about
- Reflector (zulip:stream:155) dedicated stream about reflector development and usage
- GreyHaven is separated in multiple topics: branding is in (zulip:stream:206), leads specific to greyhaven (zulip:stream:208) with one topic per lead, and marketing (zulip:stream:212)
### Meeting and Calendar
Some persons in the company have a dedicated room for their meeting in reflector. This can be seen in `room_name` in `meeting` entity.
For person like Max, dataindex have calendar information, and he mostly have a related meeting that will be in reflector. However, there is no direct relation between calendar information and reflector meeting. A correlation has to be done to figure out which meeting is it when talking about an event.

105
.agents/skills/connectors/SKILL.md Normal file
View File

@@ -0,0 +1,105 @@
---
name: connectors
description: Reference for all data connectors and their entity type mappings. Use when determining which connector produces which entity types, understanding connector-specific fields, or choosing the right data source for a query.
user-invocable: false
---
# Connectors and Data Sources
Each connector ingests data from an external source into DataIndex. Connectors run periodic background syncs to keep data fresh.
Use `list_connectors()` at runtime to see which connectors are actually configured — not all connectors below may be active in every deployment.
## Connector → Entity Type Mapping
| Connector ID | Entity Types Produced | Description |
|------------------|-----------------------------------------------------------------|----------------------------------|
| `reflector` | `meeting` | Meeting recordings + transcripts |
| `ics_calendar` | `calendar_event` | ICS calendar feed events |
| `mbsync_email` | `email` | Email via mbsync IMAP sync |
| `zulip` | `conversation`, `conversation_message`, `threaded_conversation` | Zulip chat streams and topics |
| `babelfish` | `conversation_message`, `threaded_conversation` | Chat translation bridge |
| `hedgedoc` | `document` | HedgeDoc collaborative documents |
| `contactdb` | `contact` | Synced from ContactDB (static) |
| `browser_history`| `webpage` | Browser extension page visits |
| `api_document` | `document` | API-ingested documents (static) |
## Per-Connector Details
### `reflector` — Meeting Recordings
Ingests meetings from Reflector, Monadical's meeting recording tool.
- **Entity type:** `meeting`
- **Key fields:** `transcript`, `summary`, `participants`, `start_time`, `end_time`, `room_name`
- **Use cases:** Find meetings someone attended, search meeting transcripts, get summaries
- **Tip:** Filter with `contact_ids` to find meetings involving specific people. The `transcript` field contains speaker-diarized text.
### `ics_calendar` — Calendar Events
Parses ICS calendar feeds (Google Calendar, Outlook, etc.).
- **Entity type:** `calendar_event`
- **Key fields:** `start_time`, `end_time`, `attendees`, `location`, `description`, `calendar_name`
- **Use cases:** Check upcoming events, find events with specific attendees, review past schedule
- **Tip:** Multiple calendar feeds may be configured as separate connectors (e.g., `personal_calendar`, `work_calendar`). Use `list_connectors()` to discover them.
### `mbsync_email` — Email
Syncs email via mbsync (IMAP).
- **Entity type:** `email`
- **Key fields:** `text_content`, `from_contact_id`, `to_contact_ids`, `cc_contact_ids`, `thread_id`, `has_attachments`
- **Use cases:** Find emails from/to someone, search email content, track email threads
- **Tip:** Use `from_contact_id` and `to_contact_ids` with `contact_ids` filter. For thread grouping, use the `thread_id` field.
### `zulip` — Chat
Ingests Zulip streams, topics, and messages.
- **Entity types:**
- `conversation` — A Zulip stream/channel with recent messages
- `conversation_message` — Individual chat messages
- `threaded_conversation` — A topic thread within a stream
- **Key fields:** `message`, `mentioned_contact_ids`, `recent_messages`
- **Use cases:** Find discussions about a topic, track who said what, find @-mentions
- **Tip:** Use `threaded_conversation` to find topic-level discussions. Use `conversation_message` with `mentioned_contact_ids` to find messages that mention specific people.
### `babelfish` — Translation Bridge
Ingests translated chat messages from the Babelfish service.
- **Entity types:** `conversation_message`, `threaded_conversation`
- **Use cases:** Similar to Zulip but for translated cross-language conversations
- **Tip:** Query alongside `zulip` connector for complete conversation coverage.
### `hedgedoc` — Collaborative Documents
Syncs documents from HedgeDoc (collaborative markdown editor).
- **Entity type:** `document`
- **Key fields:** `content`, `description`, `url`, `revision_id`
- **Use cases:** Find documents by content, track document revisions
- **Tip:** Use `search()` for semantic document search rather than `query_entities` text filter.
### `contactdb` — Contact Sync (Static)
Mirrors contacts from ContactDB into DataIndex for unified search.
- **Entity type:** `contact`
- **Note:** This is a read-only mirror. Use ContactDB MCP tools directly for contact operations.
### `browser_history` — Browser Extension (Static)
Captures visited webpages from a browser extension.
- **Entity type:** `webpage`
- **Key fields:** `url`, `visit_time`, `text_content`
- **Use cases:** Find previously visited pages, search page content
### `api_document` — API Documents (Static)
Documents ingested via the REST API (e.g., uploaded PDFs, imported files).
- **Entity type:** `document`
- **Note:** These are ingested via `POST /api/v1/ingest/documents`, not periodic sync.

160
.agents/skills/contactdb/SKILL.md Normal file
View File

@@ -0,0 +1,160 @@
---
name: contactdb
description: ContactDB REST API reference. Use when resolving people to contact_ids, searching contacts by name/email, or accessing relationships, notes, and platform identities.
user-invocable: false
---
# ContactDB API Reference
ContactDB is the people directory. It stores contacts, their platform identities, relationships, notes, and links. Every person across all data sources resolves to a single ContactDB `contact_id`.
**Base URL:** `http://localhost:42000/contactdb-api` (via Caddy) or `http://localhost:42800` (direct)
## Core Entities
### Contact
The central entity — represents a person.
| Field | Type | Description |
|----------------------|---------------------|------------------------------------------------|
| `id` | int | Unique contact ID |
| `name` | string | Display name |
| `emails` | EmailField[] | `{type, value, preferred}` |
| `phones` | PhoneField[] | `{type, value, preferred}` |
| `bio` | string? | Short biography |
| `avatar_url` | string? | Profile image URL |
| `personal_info` | PersonalInfo | Birthday, partner, children, role, company, location, how_we_met |
| `interests` | string[] | Topics of interest |
| `values` | string[] | Personal values |
| `tags` | string[] | User-assigned tags |
| `profile_description`| string? | Extended description |
| `is_placeholder` | bool | Auto-created stub (not yet fully resolved) |
| `is_service_account` | bool | Non-human account (bot, no-reply) |
| `stats` | ContactStats | Interaction statistics (see below) |
| `enrichment_data` | dict | Data from enrichment providers |
| `platform_identities`| PlatformIdentity[] | Identities on various platforms |
| `created_at` | datetime | When created |
| `updated_at` | datetime | Last modified |
| `merged_into_id` | int? | If merged, target contact ID |
| `deleted_at` | datetime? | Soft-delete timestamp |
### ContactStats
| Field | Type | Description |
|--------------------------|---------------|--------------------------------------|
| `total_messages` | int | Total messages across platforms |
| `platforms_count` | int | Number of platforms active on |
| `last_interaction_at` | string? | ISO datetime of last interaction |
| `interaction_count_30d` | int | Interactions in last 30 days |
| `interaction_count_90d` | int | Interactions in last 90 days |
| `hotness` | HotnessScore? | Composite engagement score (0-100) |
### PlatformIdentity
Links a contact to a specific platform account.
| Field | Type | Description |
|--------------------|-----------|------------------------------------------|
| `id` | int | Identity record ID |
| `contact_id` | int | Parent contact |
| `source` | string | Data provenance (e.g., `dataindex_zulip`)|
| `platform` | string | Platform name (e.g., `email`, `zulip`) |
| `platform_user_id` | string | User ID on that platform |
| `display_name` | string? | Name shown on that platform |
| `avatar_url` | string? | Platform-specific avatar |
| `bio` | string? | Platform-specific bio |
| `extra_data` | dict | Additional platform-specific data |
| `first_seen_at` | datetime | When first observed |
| `last_seen_at` | datetime | When last observed |
### Relationship
Tracks connections between contacts.
| Field | Type | Description |
|------------------------|-----------|--------------------------------------|
| `id` | int | Relationship ID |
| `from_contact_id` | int | Source contact |
| `to_contact_id` | int | Target contact |
| `relationship_type` | string | Type (e.g., "colleague", "client") |
| `since_date` | date? | When relationship started |
| `relationship_metadata`| dict | Additional metadata |
### Note
Free-text notes attached to a contact.
| Field | Type | Description |
|--------------|----------|----------------------|
| `id` | int | Note ID |
| `contact_id` | int | Parent contact |
| `content` | string | Note text |
| `created_by` | string | Who wrote it |
| `created_at` | datetime | When created |
### Link
External URLs associated with a contact.
| Field | Type | Description |
|--------------|----------|--------------------------|
| `id` | int | Link ID |
| `contact_id` | int | Parent contact |
| `type` | string | Link type (e.g., "github", "linkedin") |
| `label` | string | Display label |
| `url` | string | URL |
## REST Endpoints
### GET `/api/contacts` — List/search contacts
Primary way to find contacts. Returns `{contacts: [...], total, limit, offset}`.
**Query parameters:**
| Parameter | Type | Description |
|------------------------|---------------|----------------------------------------------|
| `search` | string? | Search in name and bio |
| `is_placeholder` | bool? | Filter by placeholder status |
| `is_service_account` | bool? | Filter by service account status |
| `sort_by` | string? | `"hotness"`, `"name"`, or `"updated_at"` |
| `min_hotness` | float? | Minimum hotness score (0-100) |
| `max_hotness` | float? | Maximum hotness score (0-100) |
| `platforms` | string[]? | Contacts with ALL specified platforms (AND) |
| `last_interaction_from`| string? | ISO datetime lower bound |
| `last_interaction_to` | string? | ISO datetime upper bound |
| `limit` | int | Max results (1-100, default 50) |
| `offset` | int | Pagination offset (default 0) |
### GET `/api/contacts/me` — Get self contact
Returns the platform operator's own contact record. **Call this first** in most workflows to get your own `contact_id`.
### GET `/api/contacts/{id}` — Get contact by ID
Get full details for a single contact by numeric ID.
### GET `/api/contacts/by-email/{email}` — Get contact by email
Look up a contact by email address.
### Other Endpoints
| Method | Path | Description |
|--------|-----------------------------------------|----------------------------------|
| POST | `/api/contacts` | Create contact |
| PUT | `/api/contacts/{id}` | Update contact |
| DELETE | `/api/contacts/{id}` | Delete contact |
| POST | `/api/contacts/merge` | Merge two contacts |
| GET | `/api/contacts/{id}/relationships` | List relationships |
| GET | `/api/contacts/{id}/notes` | List notes |
| GET | `/api/contacts/{id}/links` | List links |
| GET | `/api/platform-identities/contacts/{id}`| List platform identities |
## Usage Pattern
1. **Start with `GET /api/contacts/me`** to get the operator's contact ID
2. **Search by name** with `GET /api/contacts?search=Alice`
3. **Use contact IDs** from results as filters in DataIndex queries (`contact_ids` parameter)
4. **Paginate** large result sets with `offset` increments

223
.agents/skills/dataindex/SKILL.md Normal file
View File

@@ -0,0 +1,223 @@
---
name: dataindex
description: DataIndex REST API reference. Use when querying unified data (emails, meetings, calendar events, Zulip conversations, documents) via GET /query, POST /search, or GET /entities/{id}.
user-invocable: false
---
# DataIndex API Reference
DataIndex aggregates data from all connected sources (email, calendar, Zulip, meetings, documents) into a unified query interface. Every piece of data is an **entity** with a common base structure plus type-specific fields.
**Base URL:** `http://localhost:42000/dataindex/api/v1` (via Caddy) or `http://localhost:42180/api/v1` (direct)
## Entity Types
All entities share these base fields:
| Field | Type | Description |
|----------------------|-------------|---------------------------------------------|
| `id` | string | Format: `connector_name:native_id` |
| `entity_type` | string | One of the types below |
| `timestamp` | datetime | When the entity occurred |
| `contact_ids` | string[] | ContactDB IDs of people involved |
| `connector_id` | string | Which connector produced this |
| `title` | string? | Display title |
| `parent_id` | string? | Parent entity (e.g., thread for a message) |
| `raw_data` | dict | Original source data (excluded by default) |
### `calendar_event`
From ICS calendar feeds.
| Field | Type | Description |
|-----------------------|-------------|--------------------------------|
| `start_time` | datetime? | Event start |
| `end_time` | datetime? | Event end |
| `all_day` | bool | All-day event flag |
| `description` | string? | Event description |
| `location` | string? | Event location |
| `attendees` | dict[] | Attendee list |
| `organizer_contact_id`| string? | ContactDB ID of organizer |
| `status` | string? | Event status |
| `calendar_name` | string? | Source calendar name |
| `meeting_url` | string? | Video call link |
### `meeting`
From Reflector (recorded meetings with transcripts).
| Field | Type | Description |
|--------------------|---------------------|-----------------------------------|
| `start_time` | datetime? | Meeting start |
| `end_time` | datetime? | Meeting end |
| `participants` | MeetingParticipant[]| People in the meeting |
| `meeting_platform` | string? | Platform (e.g., "jitsi") |
| `transcript` | string? | Full transcript text |
| `summary` | string? | AI-generated summary |
| `meeting_url` | string? | Meeting link |
| `recording_url` | string? | Recording link |
| `location` | string? | Physical location |
| `room_name` | string? | Virtual room name (also indicates meeting location — see below) |
**MeetingParticipant** fields: `display_name`, `contact_id?`, `platform_user_id?`, `email?`, `speaker?`
> **`room_name` as location indicator:** The `room_name` field often encodes where the meeting took place (e.g., a Jitsi room name like `standup-office-bogota`). Use it to infer the meeting location when `location` is not set.
> **Participant and contact coverage is incomplete.** Meeting data comes from Reflector, which only tracks users who are logged into the Reflector platform. This means:
>
> - **`contact_ids`** only contains ContactDB IDs for Reflector-logged participants who were matched to a known contact. It will often be a **subset** of the actual attendees — do not assume it is the full list.
> - **`participants`** is more complete than `contact_ids` but still only includes people detected by Reflector. Not all participants have accounts or could be identified — some attendees may be entirely absent from this list.
> - **`contact_id` within a participant** may be `null` if the person was detected but couldn't be matched to a ContactDB entry.
>
> **Consequence for queries:** Filtering meetings by `contact_ids` will **miss meetings** where the person attended but wasn't logged into Reflector or wasn't resolved. To get better coverage, combine multiple strategies:
>
> 1. Filter by `contact_ids` for resolved participants
> 2. Search `participants[].display_name` client-side for name matches
> 3. Use `POST /search` with the person's name to search meeting transcripts and summaries
### `email`
From mbsync email sync.
| Field | Type | Description |
|--------------------|-----------|--------------------------------------|
| `thread_id` | string? | Email thread grouping |
| `text_content` | string? | Plain text body |
| `html_content` | string? | HTML body |
| `snippet` | string? | Preview snippet |
| `from_contact_id` | string? | Sender's ContactDB ID |
| `to_contact_ids` | string[] | Recipient ContactDB IDs |
| `cc_contact_ids` | string[] | CC recipient ContactDB IDs |
| `has_attachments` | bool | Has attachments flag |
| `attachments` | dict[] | Attachment metadata |
### `conversation`
A Zulip stream/channel.
| Field | Type | Description |
|--------------------|---------|----------------------------------------|
| `recent_messages` | dict[] | Recent messages in the conversation |
### `conversation_message`
A single message in a Zulip conversation.
| Field | Type | Description |
|-------------------------|-----------|-----------------------------------|
| `message` | string? | Message text content |
| `mentioned_contact_ids` | string[] | ContactDB IDs of mentioned people |
### `threaded_conversation`
A Zulip topic thread (group of messages under a topic).
| Field | Type | Description |
|--------------------|---------|----------------------------------------|
| `recent_messages` | dict[] | Recent messages in the thread |
### `document`
From HedgeDoc, API ingestion, or other document sources.
| Field | Type | Description |
|----------------|-----------|------------------------------|
| `content` | string? | Document body text |
| `description` | string? | Document description |
| `mimetype` | string? | MIME type |
| `url` | string? | Source URL |
| `revision_id` | string? | Revision identifier |
### `webpage`
From browser history extension.
| Field | Type | Description |
|----------------|-----------|------------------------------|
| `url` | string | Page URL |
| `visit_time` | datetime | When visited |
| `text_content` | string? | Page text content |
## REST Endpoints
### GET `/api/v1/query` — Exhaustive Filtered Enumeration
Use when you need **all** entities matching specific criteria. Supports pagination.
**When to use:** "List all meetings since January", "Get all emails from Alice", "Count calendar events this week"
**Query parameters:**
| Parameter | Type | Description |
|------------------|---------------|------------------------------------------------|
| `entity_types` | string (repeat) | Filter by type — repeat param for multiple: `?entity_types=email&entity_types=meeting` |
| `contact_ids` | string | Comma-separated ContactDB IDs: `"1,42"` |
| `connector_ids` | string | Comma-separated connector IDs: `"zulip,reflector"` |
| `date_from` | string | ISO datetime lower bound (UTC if no timezone) |
| `date_to` | string | ISO datetime upper bound |
| `search` | string? | Text filter on content fields |
| `parent_id` | string? | Filter by parent entity |
| `id_prefix` | string? | Filter entities by ID prefix (e.g., `zulip:stream:155`) |
| `thread_id` | string? | Filter emails by thread ID |
| `room_name` | string? | Filter meetings by room name |
| `limit` | int | Max results per page (default 50) |
| `offset` | int | Pagination offset (default 0) |
| `sort_by` | string | `"timestamp"` (default), `"title"`, `"contact_activity"`, etc. |
| `sort_order` | string | `"desc"` (default) or `"asc"` |
| `include_raw_data`| bool | Include raw_data field (default false) |
**Response format:**
```json
{
"items": [...],
"total": 152,
"page": 1,
"size": 50,
"pages": 4
}
```
**Pagination:** loop with offset increments until `offset >= total`. See the [notebook-patterns skill](.agents/skills/notebook-patterns/SKILL.md) for a reusable helper.
### POST `/api/v1/search` — Semantic Search
Use when you need **relevant** results for a natural-language question. Returns ranked text chunks. No pagination — set a higher `limit` instead.
**When to use:** "What was discussed about the product roadmap?", "Find conversations about hiring"
**Request body (JSON):**
```json
{
"search_text": "product roadmap decisions",
"entity_types": ["meeting", "threaded_conversation"],
"contact_ids": ["1", "42"],
"date_from": "2025-01-01T00:00:00Z",
"date_to": "2025-06-01T00:00:00Z",
"connector_ids": ["reflector", "zulip"],
"limit": 20
}
```
**Response:** `{results: [...chunks], total_count}` — each chunk has `entity_ids`, `entity_type`, `connector_id`, `content`, `timestamp`.
### GET `/api/v1/entities/{id}` — Get Entity by ID
Retrieve full details of a single entity. The `entity_id` format is `connector_name:native_id`.
### GET `/api/v1/connectors/status` — Connector Status
Get sync status for all connectors (last sync time, entity count, health).
## Common Query Recipes
| Question | entity_type + connector_id |
|---------------------------------------|------------------------------------------|
| Meetings I attended | `meeting` + `reflector`, with your contact_id |
| Upcoming calendar events | `calendar_event` + `ics_calendar`, date_from=now |
| Emails from someone | `email` + `mbsync_email`, with their contact_id |
| Zulip threads about a topic | `threaded_conversation` + `zulip`, search="topic" |
| All documents | `document` + `hedgedoc` |
| Chat messages mentioning someone | `conversation_message` + `zulip`, with contact_id |
| What was discussed about X? | Use `POST /search` with `search_text` |

808
.agents/skills/notebook-patterns/SKILL.md Normal file
View File

@@ -0,0 +1,808 @@
---
name: notebook-patterns
description: Marimo notebook patterns for InternalAI data analysis. Use when creating or editing marimo notebooks — covers cell scoping, async cells, pagination helpers, analysis patterns, and do/don't rules.
user-invocable: false
---
# Marimo Notebook Patterns
This guide covers how to create [marimo](https://marimo.io) notebooks for data analysis against the InternalAI platform APIs. Marimo notebooks are plain `.py` files with reactive cells — no `.ipynb` format, no Jupyter dependency.
## Marimo Basics
A marimo notebook is a Python file with `@app.cell` decorated functions. Each cell returns values as a tuple, and other cells receive them as function parameters — marimo builds a reactive DAG automatically.
```python
import marimo
app = marimo.App()
@app.cell
def cell_one():
x = 42
return (x,)
@app.cell
def cell_two(x):
# Re-runs automatically when x changes
result = x * 2
return (result,)
```
**Key rules:**
- Cells declare dependencies via function parameters
- Cells return values as tuples: `return (var1, var2,)`
- The **last expression at the top level** of a cell is displayed as rich output in the marimo UI (dataframes render as tables, dicts as collapsible trees). Expressions inside `if`/`else`/`for` blocks do **not** count — see [Cell Output Must Be at the Top Level](#cell-output-must-be-at-the-top-level) below
- Use `mo.md("# heading")` for formatted markdown output (import `mo` once in setup — see below)
- No manual execution order; the DAG determines it
- **Variable names must be unique across cells.** Every variable assigned at the top level of a cell is tracked by marimo's DAG. If two cells both define `resp`, marimo raises `MultipleDefinitionError` and refuses to run. Prefix cell-local variables with `_` (e.g., `_resp`, `_rows`, `_data`) to make them **private** to that cell — marimo ignores `_`-prefixed names.
- **All imports must go in the `setup` cell.** Every `import` statement creates a top-level variable (e.g., `import asyncio` defines `asyncio`). If two cells both `import asyncio`, marimo raises `MultipleDefinitionError`. Place **all** imports in a single setup cell and pass them as cell parameters. Do NOT `import marimo as mo` or `import asyncio` in multiple cells — import once in `setup`, then receive via `def my_cell(mo, asyncio):`.
### Cell Variable Scoping — Example
This is the **most common mistake**. Any variable assigned at the top level of a cell (not inside a `def` or comprehension) is tracked by marimo. If two cells assign the same name, the notebook refuses to run.
**BROKEN** — `resp` is defined at top level in both cells:
```python
# Cell A
@app.cell
def search_meetings(client, DATAINDEX):
resp = client.post(f"{DATAINDEX}/search", json={...}) # defines 'resp'
resp.raise_for_status()
results = resp.json()["results"]
return (results,)
# Cell B
@app.cell
def fetch_details(client, DATAINDEX, results):
resp = client.get(f"{DATAINDEX}/entities/{results[0]}") # also defines 'resp' → ERROR
meeting = resp.json()
return (meeting,)
```
> **Error:** `MultipleDefinitionError: variable 'resp' is defined in multiple cells`
**FIXED** — prefix cell-local variables with `_`:
```python
# Cell A
@app.cell
def search_meetings(client, DATAINDEX):
_resp = client.post(f"{DATAINDEX}/search", json={...}) # _resp is cell-private
_resp.raise_for_status()
results = _resp.json()["results"]
return (results,)
# Cell B
@app.cell
def fetch_details(client, DATAINDEX, results):
_resp = client.get(f"{DATAINDEX}/entities/{results[0]}") # _resp is cell-private, no conflict
meeting = _resp.json()
return (meeting,)
```
**Rule of thumb:** if a variable is only used within the cell to compute a return value, prefix it with `_`. Only leave names unprefixed if another cell needs to receive them.
> **Note:** Variables inside nested `def` functions are naturally local and don't need `_` prefixes — e.g., `resp` inside a `def fetch_all(...)` helper is fine because it's scoped to the function, not the cell.
### Cell Output Must Be at the Top Level
Marimo only renders the **last expression at the top level** of a cell as rich output. An expression buried inside an `if`/`else`, `for`, `try`, or any other block is **not** displayed — it's silently discarded.
**BROKEN** — `_df` inside the `if` branch is never rendered, and `mo.md()` inside `if`/`else` is also discarded:
```python
@app.cell
def show_results(results, mo):
if results:
_df = pl.DataFrame(results)
mo.md(f"**Found {len(results)} results**")
_df # Inside an if block — marimo does NOT display this
else:
mo.md("**No results found**") # Also inside a block — NOT displayed
return
```
**FIXED** — split into separate cells. Each cell displays exactly **one thing** at the top level:
```python
# Cell 1: build the data, return it
@app.cell
def build_results(results, pl):
results_df = pl.DataFrame(results) if results else None
return (results_df,)
# Cell 2: heading — mo.md() is the top-level expression (use ternary for conditional text)
@app.cell
def show_results_heading(results_df, mo):
mo.md(f"**Found {len(results_df)} results**" if results_df is not None else "**No results found**")
# Cell 3: table — DataFrame is the top-level expression
@app.cell
def show_results_table(results_df):
results_df # Top-level expression — marimo renders this as interactive table
```
**Rules:**
- Each cell should display **one thing** — either `mo.md()` OR a DataFrame, never both
- `mo.md()` must be a **top-level expression**, not inside `if`/`else`/`for`/`try` blocks
- Build conditional text using variables or ternary expressions, then call `mo.md(_text)` at the top level
- For DataFrames, use a standalone display cell: `def show_table(df): df`
### Async Cells
When a cell uses `await` (e.g., for `llm_call` or `asyncio.gather`), you **must** declare it as `async def`:
```python
@app.cell
async def analyze(meetings, llm_call, ResponseModel, asyncio):
async def _score(meeting):
return await llm_call(prompt=..., response_model=ResponseModel)
results = await asyncio.gather(*[_score(_m) for _m in meetings])
return (results,)
```
Note that `asyncio` is imported in the `setup` cell and received here as a parameter — never `import asyncio` inside individual cells.
If you write `await` in a non-async cell, marimo cannot parse the cell and saves it as an `_unparsable_cell` string literal — the cell won't run, and you'll see `SyntaxError: 'return' outside function` or similar errors. See [Fixing `_unparsable_cell`](#fixing-_unparsable_cell) below.
### Cells That Define Classes Must Return Them
If a cell defines Pydantic models (or any class) that other cells need, it **must** return them:
```python
# BaseModel and Field are imported in the setup cell and received as parameters
@app.cell
def models(BaseModel, Field):
class MeetingSentiment(BaseModel):
overall_sentiment: str
sentiment_score: int = Field(description="Score from -10 to +10")
class FrustrationExtraction(BaseModel):
has_frustrations: bool
frustrations: list[dict]
return MeetingSentiment, FrustrationExtraction # Other cells receive these as parameters
```
A bare `return` (or no return) means those classes are invisible to the rest of the notebook.
### Fixing `_unparsable_cell`
When marimo can't parse a cell into a proper `@app.cell` function, it saves the raw code as `app._unparsable_cell("...", name="cell_name")`. These cells **won't run** and show errors like `SyntaxError: 'return' outside function`.
**Common causes:**
1. Using `await` without making the cell `async def`
2. Using `return` in code that marimo failed to wrap into a function (usually a side effect of cause 1)
**How to fix:** Convert the `_unparsable_cell` string back into a proper `@app.cell` decorated function:
```python
# BROKEN — saved as _unparsable_cell because of top-level await
app._unparsable_cell("""
results = await asyncio.gather(...)
return results
""", name="my_cell")
# FIXED — proper async cell function (asyncio imported in setup, received as parameter)
@app.cell
async def my_cell(some_dependency, asyncio):
results = await asyncio.gather(...)
return (results,)
```
**Key differences to note when converting:**
- Wrap the code in an `async def` function (if it uses `await`)
- Add cell dependencies as function parameters (including imports like `asyncio`)
- Return values as tuples: `return (var,)` not `return var`
- Prefix cell-local variables with `_`
- Never add `import` statements inside the cell — all imports belong in `setup`
### Inline Dependencies with PEP 723
Use PEP 723 `/// script` metadata so `uv run` auto-installs dependencies:
```python
# /// script
# requires-python = ">=3.12"
# dependencies = [
# "marimo",
# "httpx",
# "polars",
# "mirascope[openai]",
# "pydantic",
# "python-dotenv",
# ]
# ///
```
### Checking Notebooks Before Running
Always run `marimo check` before opening or running a notebook. It catches common issues — duplicate variable definitions, `_unparsable_cell` blocks, branch expressions that won't display, and more — without needing to start the full editor:
```bash
uvx marimo check notebook.py # Check a single notebook
uvx marimo check workflows/ # Check all notebooks in a directory
uvx marimo check --fix notebook.py # Auto-fix fixable issues
```
**Run this after every edit.** A clean `marimo check` (no output, exit code 0) means the notebook is structurally valid. Any errors must be fixed before running.
### Running Notebooks
```bash
uvx marimo edit notebook.py # Interactive editor (best for development)
uvx marimo run notebook.py # Read-only web app
uv run notebook.py # Script mode (terminal output)
```
### Inspecting Cell Outputs
In `marimo edit`, every cell's return value is displayed as rich output below the cell. This is the primary way to introspect API responses:
- **Dicts/lists** render as collapsible JSON trees — click to expand nested fields
- **Polars/Pandas DataFrames** render as interactive sortable tables
- **Strings** render as plain text
To inspect a raw API response, just make it the last expression:
```python
@app.cell
def inspect_response(client, DATAINDEX):
_resp = client.get(f"{DATAINDEX}/query", params={
"entity_types": "meeting", "limit": 2,
})
_resp.json() # This gets displayed as a collapsible JSON tree
```
To inspect an intermediate value alongside other work, use `mo.accordion` or return it:
```python
@app.cell
def debug_meetings(meetings, mo):
mo.md(f"**Count:** {len(meetings)}")
# Show first item structure for inspection
mo.accordion({"First meeting raw": mo.json(meetings[0])}) if meetings else None
```
## Notebook Skeleton
Every notebook against InternalAI follows this structure:
```python
# /// script
# requires-python = ">=3.12"
# dependencies = [
# "marimo",
# "httpx",
# "polars",
# "mirascope[openai]",
# "pydantic",
# "python-dotenv",
# ]
# ///
import marimo
app = marimo.App()
@app.cell
def params():
"""User parameters — edit these to change the workflow's behavior."""
SEARCH_TERMS = ["greyhaven"]
DATE_FROM = "2026-01-01T00:00:00Z"
DATE_TO = "2026-02-01T00:00:00Z"
TARGET_PERSON = None # Set to a name like "Alice" to filter by person, or None for all
return DATE_FROM, DATE_TO, SEARCH_TERMS, TARGET_PERSON
@app.cell
def config():
BASE = "http://localhost:42000"
CONTACTDB = f"{BASE}/contactdb-api"
DATAINDEX = f"{BASE}/dataindex/api/v1"
return (CONTACTDB, DATAINDEX,)
@app.cell
def setup():
from dotenv import load_dotenv
load_dotenv(".env") # Load .env from the project root
import asyncio # All imports go here — never import inside other cells
import httpx
import marimo as mo
import polars as pl
from pydantic import BaseModel, Field
client = httpx.Client(timeout=30)
return (asyncio, client, mo, pl, BaseModel, Field,)
# --- your IN / ETL / OUT cells here ---
if __name__ == "__main__":
app.run()
```
> **`load_dotenv(".env")`** reads the `.env` file explicitly by name. This makes `LLM_API_KEY` and other env vars available to `os.getenv()` calls in `lib/llm.py` without requiring the shell to have them pre-set. Always include `python-dotenv` in PEP 723 dependencies and call `load_dotenv(".env")` early in the setup cell.
**The `params` cell must always be the first cell** after `app = marimo.App()`. It contains all user-configurable constants (search terms, date ranges, target names, etc.) as plain Python values. This way the user can tweak the workflow by editing a single cell at the top — no need to hunt through the code for hardcoded values.
## Pagination Helper
The DataIndex `GET /query` endpoint paginates with `limit` and `offset`. Always paginate — result sets can be large.
```python
@app.cell
def helpers(client):
def fetch_all(url, params):
"""Fetch all pages from a paginated DataIndex endpoint."""
all_items = []
limit = params.get("limit", 50)
params = {**params, "limit": limit, "offset": 0}
while True:
resp = client.get(url, params=params)
resp.raise_for_status()
data = resp.json()
all_items.extend(data["items"])
if params["offset"] + limit >= data["total"]:
break
params["offset"] += limit
return all_items
def resolve_contact(name, contactdb_url):
"""Find a contact by name, return their ID."""
resp = client.get(f"{contactdb_url}/api/contacts", params={"search": name})
resp.raise_for_status()
contacts = resp.json()["contacts"]
if not contacts:
raise ValueError(f"No contact found for '{name}'")
return contacts[0]
return (fetch_all, resolve_contact,)
```
## Pattern 1: Emails Involving a Specific Person
Emails have `from_contact_id`, `to_contact_ids`, and `cc_contact_ids`. The query API's `contact_ids` filter matches entities where the contact appears in **any** of these roles.
```python
@app.cell
def find_person(resolve_contact, CONTACTDB):
target = resolve_contact("Alice", CONTACTDB)
target_id = target["id"]
target_name = target["name"]
return (target_id, target_name,)
@app.cell
def fetch_emails(fetch_all, DATAINDEX, target_id):
emails = fetch_all(f"{DATAINDEX}/query", {
"entity_types": "email",
"contact_ids": str(target_id),
"date_from": "2025-01-01T00:00:00Z",
"sort_order": "desc",
})
return (emails,)
@app.cell
def email_table(emails, target_id, target_name, pl):
email_df = pl.DataFrame([{
"date": e["timestamp"][:10],
"subject": e.get("title", "(no subject)"),
"direction": (
"sent" if str(target_id) == str(e.get("from_contact_id"))
else "received"
),
"snippet": (e.get("snippet") or e.get("text_content") or "")[:100],
} for e in emails])
return (email_df,)
@app.cell
def show_emails(email_df, target_name, mo):
mo.md(f"## Emails involving {target_name} ({len(email_df)} total)")
@app.cell
def display_email_table(email_df):
email_df # Renders as interactive table in marimo edit
```
## Pattern 2: Meetings with a Specific Participant
Meetings have a `participants` list where each entry may or may not have a resolved `contact_id`. The query API's `contact_ids` filter only matches **resolved** participants.
**Strategy:** Query by `contact_ids` to get meetings with resolved participants, then optionally do a client-side check on `participants[].display_name` or `transcript` for unresolved ones.
> **Always include `room_name` in meeting tables.** The `room_name` field contains the virtual room name (e.g., `standup-office-bogota`) and often indicates where the meeting took place. It's useful context when `title` is generic or missing — include it as a column alongside `title`.
```python
@app.cell
def fetch_meetings(fetch_all, DATAINDEX, target_id, my_id):
# Get meetings where the target appears in contact_ids
resolved_meetings = fetch_all(f"{DATAINDEX}/query", {
"entity_types": "meeting",
"contact_ids": str(target_id),
"date_from": "2025-01-01T00:00:00Z",
})
return (resolved_meetings,)
@app.cell
def meeting_table(resolved_meetings, target_name, pl):
_rows = []
for _m in resolved_meetings:
_participants = _m.get("participants", [])
_names = [_p["display_name"] for _p in _participants]
_rows.append({
"date": (_m.get("start_time") or _m["timestamp"])[:10],
"title": _m.get("title", "Untitled"),
"room_name": _m.get("room_name", ""),
"participants": ", ".join(_names),
"has_transcript": _m.get("transcript") is not None,
"has_summary": _m.get("summary") is not None,
})
meeting_df = pl.DataFrame(_rows)
return (meeting_df,)
```
To also find meetings where the person was present but **not resolved** (guest), search the transcript:
```python
@app.cell
def search_unresolved(client, DATAINDEX, target_name):
# Semantic search for the person's name in meeting transcripts
_resp = client.post(f"{DATAINDEX}/search", json={
"search_text": target_name,
"entity_types": ["meeting"],
"limit": 50,
})
_resp.raise_for_status()
transcript_hits = _resp.json()["results"]
return (transcript_hits,)
```
## Pattern 3: Calendar Events → Meeting Correlation
Calendar events and meetings are separate entities from different connectors. To find which calendar events had a corresponding recorded meeting, match by time overlap.
```python
@app.cell
def fetch_calendar_and_meetings(fetch_all, DATAINDEX, my_id):
events = fetch_all(f"{DATAINDEX}/query", {
"entity_types": "calendar_event",
"contact_ids": str(my_id),
"date_from": "2025-01-01T00:00:00Z",
"sort_by": "timestamp",
"sort_order": "asc",
})
meetings = fetch_all(f"{DATAINDEX}/query", {
"entity_types": "meeting",
"contact_ids": str(my_id),
"date_from": "2025-01-01T00:00:00Z",
})
return (events, meetings,)
@app.cell
def correlate(events, meetings, pl):
from datetime import datetime, timedelta
def _parse_dt(s):
if not s:
return None
return datetime.fromisoformat(s.replace("Z", "+00:00"))
# Index meetings by start_time for matching
_meeting_by_time = {}
for _m in meetings:
_start = _parse_dt(_m.get("start_time"))
if _start:
_meeting_by_time[_start] = _m
_rows = []
for _ev in events:
_ev_start = _parse_dt(_ev.get("start_time"))
_ev_end = _parse_dt(_ev.get("end_time"))
if not _ev_start:
continue
# Find meeting within 15-min window of calendar event start
_matched = None
for _m_start, _m in _meeting_by_time.items():
if abs((_m_start - _ev_start).total_seconds()) < 900:
_matched = _m
break
_rows.append({
"date": _ev_start.strftime("%Y-%m-%d"),
"time": _ev_start.strftime("%H:%M"),
"event_title": _ev.get("title", "(untitled)"),
"has_recording": _matched is not None,
"meeting_title": _matched.get("title", "") if _matched else "",
"attendee_count": len(_ev.get("attendees", [])),
})
calendar_df = pl.DataFrame(_rows)
return (calendar_df,)
```
## Pattern 4: Full Interaction Timeline for a Person
Combine emails, meetings, and Zulip messages into a single chronological view.
```python
@app.cell
def fetch_all_interactions(fetch_all, DATAINDEX, target_id):
all_entities = fetch_all(f"{DATAINDEX}/query", {
"contact_ids": str(target_id),
"date_from": "2025-01-01T00:00:00Z",
"sort_by": "timestamp",
"sort_order": "desc",
})
return (all_entities,)
@app.cell
def interaction_timeline(all_entities, target_name, pl):
_rows = []
for _e in all_entities:
_etype = _e["entity_type"]
_summary = ""
if _etype == "email":
_summary = _e.get("snippet") or _e.get("title") or ""
elif _etype == "meeting":
_summary = _e.get("summary") or _e.get("title") or ""
elif _etype == "conversation_message":
_summary = (_e.get("message") or "")[:120]
elif _etype == "threaded_conversation":
_summary = _e.get("title") or ""
elif _etype == "calendar_event":
_summary = _e.get("title") or ""
else:
_summary = _e.get("title") or _e["entity_type"]
_rows.append({
"date": _e["timestamp"][:10],
"type": _etype,
"source": _e["connector_id"],
"summary": _summary[:120],
})
timeline_df = pl.DataFrame(_rows)
return (timeline_df,)
@app.cell
def show_timeline(timeline_df, target_name, mo):
mo.md(f"## Interaction Timeline: {target_name} ({len(timeline_df)} events)")
@app.cell
def display_timeline(timeline_df):
timeline_df
```
## Pattern 5: LLM Filtering with `lib.llm`
When you need to classify, score, or extract structured information from each entity (e.g. "is this meeting about project X?", "rate the relevance of this email"), use the `llm_call` helper from `workflows/lib`. It sends each item to an LLM and parses the response into a typed Pydantic model.
**Prerequisites:** Copy `.env.example` to `.env` and fill in your `LLM_API_KEY`. Add `mirascope`, `pydantic`, and `python-dotenv` to the notebook's PEP 723 dependencies.
```python
# /// script
# requires-python = ">=3.12"
# dependencies = [
# "marimo",
# "httpx",
# "polars",
# "mirascope[openai]",
# "pydantic",
# "python-dotenv",
# ]
# ///
```
### Setup cell — load `.env` and import `llm_call`
```python
@app.cell
def setup():
from dotenv import load_dotenv
load_dotenv(".env") # Makes LLM_API_KEY available to lib/llm.py
import asyncio
import httpx
import marimo as mo
import polars as pl
from pydantic import BaseModel, Field
from lib.llm import llm_call
client = httpx.Client(timeout=30)
return (asyncio, client, llm_call, mo, pl, BaseModel, Field,)
```
### Define a response model
Create a Pydantic model that describes the structured output you want from the LLM:
```python
@app.cell
def models(BaseModel, Field):
class RelevanceScore(BaseModel):
relevant: bool
reason: str
score: int # 0-10
return (RelevanceScore,)
```
### Filter entities through the LLM
Iterate over fetched entities and call `llm_call` for each one. Since `llm_call` is async, use `asyncio.gather` to process items concurrently:
```python
@app.cell
async def llm_filter(meetings, llm_call, RelevanceScore, pl, mo, asyncio):
_topic = "Greyhaven"
async def _score(meeting):
_text = meeting.get("summary") or meeting.get("title") or ""
_result = await llm_call(
prompt=f"Is this meeting about '{_topic}'?\n\nMeeting: {_text}",
response_model=RelevanceScore,
system_prompt="Score the relevance of this meeting to the given topic. Set relevant=true if score >= 5.",
)
return {**meeting, "llm_relevant": _result.relevant, "llm_reason": _result.reason, "llm_score": _result.score}
scored_meetings = await asyncio.gather(*[_score(_m) for _m in meetings])
relevant_meetings = [_m for _m in scored_meetings if _m["llm_relevant"]]
mo.md(f"**LLM filter:** {len(relevant_meetings)}/{len(meetings)} meetings relevant to '{_topic}'")
return (relevant_meetings,)
```
### Tips for LLM filtering
- **Keep prompts short** — only include the fields the LLM needs (title, summary, snippet), not the entire raw entity.
- **Use structured output** — always pass a `response_model` so you get typed fields back, not free-text.
- **Batch wisely** — `asyncio.gather` sends all requests concurrently. For large datasets (100+ items), process in chunks to avoid rate limits.
- **Cache results** — LLM calls are slow and cost money. If iterating on a notebook, consider storing scored results in a cell variable so you don't re-score on every edit.
## Do / Don't — Quick Reference for LLM Agents
When generating marimo notebooks, follow these rules strictly. Violations cause `MultipleDefinitionError` at runtime.
### Do
- **Prefix cell-local variables with `_`** — `_resp`, `_rows`, `_m`, `_data`, `_chunk`. Marimo ignores `_`-prefixed names so they won't clash across cells.
- **Put all imports in the `setup` cell** and pass them as cell parameters: `def my_cell(client, mo, pl, asyncio):`. Never `import` inside other cells — even `import asyncio` in two async cells causes `MultipleDefinitionError`.
- **Give returned DataFrames unique names** — `email_df`, `meeting_df`, `timeline_df`. Never use a bare `df` that might collide with another cell.
- **Return only values other cells need** — everything else should be `_`-prefixed and stays private to the cell.
- **Import stdlib modules in `setup` too** — even `from datetime import datetime` creates a top-level name. If two cells both import `datetime`, marimo errors. Import it once in `setup` and receive it as a parameter, or use it inside a `_`-prefixed helper function where it's naturally scoped.
- **Every non-utility cell must show a preview** — see the "Cell Output Previews" section below.
- **Use separate display cells for DataFrames** — the build cell returns the DataFrame and shows a `mo.md()` count/heading; a standalone display cell (e.g., `def show_table(df): df`) renders it as an interactive table the user can sort and filter.
- **Include `room_name` when listing meetings** — the virtual room name provides useful context about where the meeting took place (e.g., `standup-office-bogota`). Show it as a column alongside `title`.
- **Keep cell output expressions at the top level** — if a cell conditionally displays a DataFrame, initialize `_output = None` before the `if`/`else`, assign inside the branches, then put `_output` as the last top-level expression. Expressions inside `if`/`else`/`for` blocks are silently ignored by marimo.
- **Put all user parameters in a `params` cell as the first cell** — date ranges, search terms, target names, limits. Never hardcode these values deeper in the notebook.
- **Declare cells as `async def` when using `await`** — `@app.cell` followed by `async def cell_name(...)`. This includes cells using `asyncio.gather`, `await llm_call(...)`, or any async API.
- **Return classes/models from cells that define them** — if a cell defines `class MyModel(BaseModel)`, return it so other cells can use it as a parameter: `return (MyModel,)`.
- **Use `python-dotenv` to load `.env`** — add `python-dotenv` to PEP 723 dependencies and call `load_dotenv(".env")` early in the setup cell (before importing `lib.llm`). This ensures `LLM_API_KEY` and other env vars are available without requiring them to be pre-set in the shell.
### Don't
- **Don't define the same variable name in two cells** — even `resp = ...` in cell A and `resp = ...` in cell B is a fatal error.
- **Don't `import` inside non-setup cells** — every `import X` defines a top-level variable `X`. If two cells both `import asyncio`, marimo raises `MultipleDefinitionError` and refuses to run. Put all imports in the `setup` cell and receive them as function parameters.
- **Don't use generic top-level names** like `df`, `rows`, `resp`, `data`, `result` — either prefix with `_` or give them a unique descriptive name.
- **Don't return temporary variables** — if `_rows` is only used to build a DataFrame, keep it `_`-prefixed and only return the DataFrame.
- **Don't use `await` in a non-async cell** — this causes marimo to save the cell as `_unparsable_cell` (a string literal that won't execute). Always use `async def` for cells that call async functions.
- **Don't define classes in a cell without returning them** — a bare `return` or no return makes classes invisible to the DAG. Other cells can't receive them as parameters.
- **Don't put display expressions inside `if`/`else`/`for` blocks** — marimo only renders the last top-level expression. A DataFrame inside an `if` branch is silently discarded. Use the `_output = None` pattern instead (see [Cell Output Must Be at the Top Level](#cell-output-must-be-at-the-top-level)).
## Cell Output Previews
Every cell that fetches, transforms, or produces data **must display a preview** so the user can validate results at each step. The only exceptions are **utility cells** (config, setup, helpers) that only define constants or functions.
Think from the user's perspective: when they open the notebook in `marimo edit`, each cell should tell them something useful — a count, a sample, a summary. Silent cells that do work but show nothing are hard to debug and validate.
### What to show
| Cell type | What to preview |
|-----------|----------------|
| API fetch (list of items) | `mo.md(f"**Fetched {len(items)} meetings**")` |
| DataFrame build | The DataFrame itself as last expression (renders as interactive table) |
| Scalar result | `mo.md(f"**Contact:** {name} (id={contact_id})")` |
| Search / filter | `mo.md(f"**{len(hits)} results** matching '{term}'")` |
| Final output | Full DataFrame or `mo.md()` summary as last expression |
### Example: fetch cell with preview
**Bad** — cell runs silently, user sees nothing:
```python
@app.cell
def fetch_meetings(fetch_all, DATAINDEX, my_id):
meetings = fetch_all(f"{DATAINDEX}/query", {
"entity_types": "meeting",
"contact_ids": str(my_id),
})
return (meetings,)
```
**Good** — cell shows a count so the user knows it worked:
```python
@app.cell
def fetch_meetings(fetch_all, DATAINDEX, my_id, mo):
meetings = fetch_all(f"{DATAINDEX}/query", {
"entity_types": "meeting",
"contact_ids": str(my_id),
})
mo.md(f"**Fetched {len(meetings)} meetings**")
return (meetings,)
```
### Example: transform cell with table preview
**Bad** — builds DataFrame but doesn't display it:
```python
@app.cell
def build_table(meetings, pl):
_rows = [{"date": _m["timestamp"][:10], "title": _m.get("title", "")} for _m in meetings]
meeting_df = pl.DataFrame(_rows)
return (meeting_df,)
```
**Good** — the build cell shows a `mo.md()` count, and a **separate display cell** renders the DataFrame as an interactive table:
```python
@app.cell
def build_table(meetings, pl, mo):
_rows = [{"date": _m["timestamp"][:10], "title": _m.get("title", "")} for _m in meetings]
meeting_df = pl.DataFrame(_rows).sort("date")
mo.md(f"### Meetings ({len(meeting_df)} results)")
return (meeting_df,)
@app.cell
def show_meeting_table(meeting_df):
meeting_df # Renders as interactive sortable table
```
### Separate display cells for DataFrames
When a cell builds a DataFrame, use **two cells**: one that builds and returns it (with a `mo.md()` summary), and a standalone display cell that renders it as a table. This keeps the build logic clean and gives the user an interactive table they can sort and filter in the marimo UI.
```python
# Cell 1: build and return the DataFrame, show a count
@app.cell
def build_sentiment_table(analyzed_meetings, pl, mo):
_rows = [...]
sentiment_df = pl.DataFrame(_rows).sort("date", descending=True)
mo.md(f"### Sentiment Analysis ({len(sentiment_df)} meetings)")
return (sentiment_df,)
# Cell 2: standalone display — just the DataFrame, nothing else
@app.cell
def show_sentiment_table(sentiment_df):
sentiment_df
```
This pattern makes every result inspectable. The `mo.md()` cell gives a quick count/heading; the display cell lets the user explore the full data interactively.
### Utility cells (no preview needed)
Config, setup, and helper cells that only define constants or functions don't need previews:
```python
@app.cell
def config():
BASE = "http://localhost:42000"
CONTACTDB = f"{BASE}/contactdb-api"
DATAINDEX = f"{BASE}/dataindex/api/v1"
return CONTACTDB, DATAINDEX
@app.cell
def helpers(client):
def fetch_all(url, params):
...
return (fetch_all,)
```
## Tips
- Use `marimo edit` during development to see cell outputs interactively
- Make raw API responses the last expression in a cell to inspect their structure
- Use `polars` over `pandas` for better performance and type safety
- Set `timeout=30` on httpx clients — some queries over large date ranges are slow
- Name cells descriptively — function names appear in the marimo sidebar

364
.agents/skills/project-history/SKILL.md Normal file
View File

@@ -0,0 +1,364 @@
---
name: project-history
description: Build initial historical timeline for a project. Queries all datasources and creates week-by-week analysis files up to a sync date. Requires project-init to have been run first (datasources.md must exist).
disable-model-invocation: true
argument-hint: [project-name] [date-from] [date-to]
---
# Build Project History
**When to use:** After `/project-init` has been run and the user has reviewed `datasources.md`. This skill gathers historical data and builds the week-by-week timeline.
**Precondition:** `projects/$0/datasources.md` must exist. If it doesn't, run `/project-init $0` first.
## Step 1: Read Datasources
Read `projects/$0/datasources.md` to determine:
- Which Zulip stream IDs and search terms to query
- Which git repository to clone/pull
- Which meeting room names to filter by
- Which entity types to prioritize
## Step 2: Gather Historical Data
Query data for the period `$1` to `$2`.
### A. Query Zulip
For each PRIMARY stream in datasources.md:
```python
# Paginate through all threaded conversations
GET /api/v1/query
entity_types=threaded_conversation
connector_ids=zulip
date_from=$1
date_to=$2
search={project-search-term}
limit=100
offset=0
```
### B. Clone/Pull Git Repository
```bash
# First time
git clone --depth 200 {url} ./tmp/$0-clone
# Or if already cloned
cd ./tmp/$0-clone && git pull
# Extract commit history for the period
git log --since="$1" --until="$2" --format="%H|%an|%ae|%ad|%s" --date=short
git log --since="$1" --until="$2" --format="%an" | sort | uniq -c | sort -rn
```
### C. Query Meeting Recordings
For each PRIMARY meeting room in datasources.md:
```python
GET /api/v1/query
entity_types=meeting
date_from=$1
date_to=$2
room_name={room-name}
limit=100
```
Also do a semantic search for broader coverage:
```python
POST /api/v1/search
search_text={project-name}
entity_types=["meeting"]
date_from=$1
date_to=$2
limit=50
```
## Step 3: Analyze by Week
For each week in the period, create a week file. Group the gathered data into calendar weeks (Monday-Sunday).
For each week, analyze:
1. **Key Decisions** — Strategic choices, architecture changes, vendor selections, security responses
2. **Technical Work** — Features developed, bug fixes, infrastructure changes, merges/PRs
3. **Team Activity** — Who was active, new people, departures, role changes
4. **Blockers** — Issues, delays, dependencies
### Week file template
**File:** `projects/$0/timeline/{year-month}/week-{n}.md`
```markdown
# $0 - Week {n}, {Month} {Year}
**Period:** {date-range}
**Status:** [Active/Quiet/Blocked]
## Key Decisions
### Decision Title
- **Decision:** What was decided
- **Date:** {date}
- **Who:** {decision-makers}
- **Impact:** Why it matters
- **Context:** Background
## Technical Work
- [{Date}] {Description} - {Who}
## Team Activity
### Core Contributors
- **Name:** Focus area
### Occasional Contributors
- Name: What they contributed
## GitHub Activity
**Commits:** {count}
**Focus Areas:**
- Area 1
**Key Commits:**
- Hash: Description (Author)
## Zulip Activity
**Active Streams:**
- Stream: Topics discussed
## Current Blockers
1. Blocker description
## Milestones Reached
If any milestones were completed this week, document with business objective:
- **Milestone:** What was achieved
- **Business Objective:** WHY this matters (search for this in discussions, PRs, meetings)
- **Impact:** Quantifiable results if available
## Next Week Focus
- Priority 1
## Notes
- Context and observations
- Always try to capture the WHY behind decisions and milestones
```
### Categorization principles
**Key Decisions:**
- Technology migrations
- Architecture changes
- Vendor switches
- Security incidents
- Strategic pivots
**Technical Work:**
- Feature implementations
- Bug fixes
- Infrastructure changes
- Refactoring
**Skip Unless Meaningful:**
- Routine check-ins
- Minor documentation updates
- Social chat
### Contributor types
**Core Contributors:** Regular commits (multiple per week), active in technical discussions, making architectural decisions, reviewing PRs.
**Occasional Contributors:** Sporadic commits, topic-specific involvement, testing/QA, feedback only.
## Step 4: Create/Update Timeline Index
**File:** `projects/$0/timeline/index.md`
```markdown
# $0 Timeline Index
## {Year}
### {Quarter}
- [Month Week 1](./{year}-{month}/week-1.md)
- [Month Week 2](./{year}-{month}/week-2.md)
## Key Milestones
| Date | Milestone | Business Objective | Status |
|------|-----------|-------------------|--------|
| Mar 2025 | SQLite → PostgreSQL migration | Improve query performance (107ms→27ms) and enable concurrent access for scaling | Complete |
| Jul 2025 | Chakra UI 3 migration | Modernize UI component library and improve accessibility | Complete |
## Summary by Quarter
### Q{X} {Year}
- **Milestone 1:** What happened + Business objective
- **Milestone 2:** What happened + Business objective
```
## Step 5: Create Project Dashboard (project.md)
**File:** `projects/$0/project.md`
Create the **living document** — the entry point showing current status:
```markdown
# $0 Project
**One-liner:** [Brief description]
**Status:** [Active/On Hold/Deprecated]
**Last Updated:** [Date]
---
## This Week's Focus
### Primary Objective
[What the team is working on right now - from the most recent week]
### Active Work
- [From recent commits and discussions]
### Blockers
- [Any current blockers]
---
## Last Week's Focus
### Delivered
- ✅ [What was completed]
### Decisions Made
- [Key decisions from last week]
---
## Team
### Core Contributors (Active)
| Name | Focus | Availability |
|------|-------|--------------|
| [From git analysis] | [Area] | Full-time/Part-time |
### Occasional Contributors
- [Name] - [Role]
---
## Milestones
### In Progress 🔄
| Milestone | Target | Business Objective |
|-----------|--------|-------------------|
| [Active milestones from the data] | [Date] | [WHY this matters] |
### Recently Completed ✅
| Milestone | Date | Business Objective |
|-----------|------|-------------------|
| [Recently completed] | [Date] | [WHY this mattered] |
### Lost in Sight / Paused ⏸️
| Milestone | Status | Reason |
|-----------|--------|--------|
| [If any] | Paused | [Why] |
---
## Recent Decisions
### Week [N] (Current)
- **[Decision]** - [Context from data]
---
## Quick Links
- [📊 Timeline](./timeline/index.md) - Week-by-week history
- [📋 Background](./background.md) - Project architecture
- [🔌 Data Sources](./datasources.md) - How to gather information
---
*This is a living document. It reflects the current state and changes frequently.*
```
**Fill in from the analyzed data:**
- Team members from git contributors
- Current focus from the most recent week's activity
- Milestones from major features/deployments found in the data
- Recent decisions from meeting transcripts and Zulip discussions
## Step 6: Update Sync State
Update `projects/$0/sync-state.md`:
```markdown
# Sync State
status: history_complete
created_at: {original date}
last_sync_date: $2
initial_history_from: $1
initial_history_to: $2
```
## Common Patterns
### Security Incident
```markdown
### Security Incident: {CVE-ID}
- **Discovered:** {date}
- **Severity:** CRITICAL/HIGH/MEDIUM
- **Who:** {discoverers}
- **Impact:** {description}
- **Actions:**
1. Immediate fix
2. Secrets rotated
3. Monitoring added
```
### Technology Migration
```markdown
### Migration: {Old} -> {New}
- **Decision:** {date}
- **Who:** {decision-makers}
- **Timeline:** {duration}
- **Rationale:** {why} ← Always include the business objective
- **Status:** Complete/In Progress/Planned
```
**Important:** When documenting any milestone or decision, always search for and include the WHY:
- Performance improvements (quantify if possible: "reduced from X to Y")
- Business capabilities enabled ("allows concurrent access for scaling")
- User experience improvements ("improves accessibility")
- Risk mitigation ("addresses security vulnerability")
- Cost reduction ("eliminates cloud dependency")
Look for this context in: meeting recordings, Zulip planning threads, PR descriptions, release notes.
### Team Change
```markdown
### Team: {Name} {Joined/Left/Role Change}
- **Date:** {date}
- **From:** {old role} (if applicable)
- **To:** {new role}
- **Impact:** {on project}
```
## Key Rules
- **Link to sources**: Always reference commit hashes, PR numbers, Zulip topic names, meeting dates
- **Be explicit about exclusions**: Document what streams/sources you're NOT analyzing and why
- **Write once**: Week files are historical records — don't modify them after creation
- **Paginate all queries**: Result sets can be large, always loop through all pages

264
.agents/skills/project-init/SKILL.md Normal file
View File

@@ -0,0 +1,264 @@
---
name: project-init
description: Initialize a new project analysis. Creates directory structure, discovers relevant data sources (Zulip streams, git repos, meeting rooms), and writes datasources.md, background.md skeleton, and sync-state.md.
disable-model-invocation: true
argument-hint: [project-name]
---
# Initialize Project Analysis
**When to use:** Starting analysis of a new project. This skill sets up the project structure and discovers data sources. It does NOT gather historical data — use `/project-history` for that after reviewing the datasources.
## Step 1: Create Project Structure
```bash
mkdir -p projects/$0/timeline
```
## Step 2: Discover and Document Data Sources
Investigate what data sources exist for this project. Use the [connectors skill](../connectors/SKILL.md) and [company skill](../company/SKILL.md) for reference.
### Discovery process
1. **Zulip streams**: Search DataIndex for `threaded_conversation` entities matching the project name. Note which stream IDs appear. Cross-reference with the company skill's Zulip channel list to identify primary vs. secondary streams.
2. **Git repositories**: Ask the user for the repository URL, or search Gitea/GitHub if accessible.
3. **Meeting rooms**: Search DataIndex for `meeting` entities matching the project name. Note which `room_name` values appear — these are the relevant meeting rooms.
4. **Search terms**: Identify the project name, key technologies, and domain-specific terms that surface relevant data.
5. **Entity type priority**: Determine which entity types are most relevant (typically `threaded_conversation`, `meeting`, and possibly `email`).
### Write datasources.md
**File:** `projects/$0/datasources.md`
```markdown
# $0 - Data Sources
## Zulip Streams
### PRIMARY Streams (Analyze All)
| Stream ID | Name | Topics | Priority | What to Look For |
|-----------|------|--------|----------|------------------|
| XXX | stream-name | N topics | CRITICAL | Development discussions |
### SECONDARY Streams (Selective)
| Stream ID | Name | Topics to Analyze | Context |
|-----------|------|-------------------|---------|
| YYY | integration-stream | specific-topic | Integration work |
### EXCLUDE
- stream-id-1: reason
- stream-id-2: reason
## Git Repository
**URL:** https://...
**Commands:**
```
git clone {url} ./tmp/$0-clone
cd ./tmp/$0-clone
git log --format="%H|%an|%ae|%ad|%s" --date=short > commits.csv
git log --format="%an|%ae" | sort | uniq -c | sort -rn
```
## Meeting Rooms
### PRIMARY
- room-name: Project-specific discussions
### SECONDARY (Context Only)
- allhands: General updates
### EXCLUDE
- personal-rooms: Other projects
## Search Terms
### Primary
- project-name
- key-technology-1
### Technical
- architecture-term-1
## Entity Types Priority
1. threaded_conversation (Zulip)
2. meeting (recordings)
3. [Exclude: calendar, email, document if not relevant]
```
## Step 3: Create Project Dashboard (Living Document)
**File:** `projects/$0/project.md`
This is the **entry point** — the living document showing current status.
```markdown
# $0 Project
**One-liner:** [Brief description]
**Status:** [Active/On Hold/Deprecated]
**Repository:** URL
**Last Updated:** [Date]
---
## This Week's Focus
### Primary Objective
[What the team is working on right now]
### Active Work
- [Current task 1]
- [Current task 2]
### Blockers
- [Any blockers]
---
## Last Week's Focus
### Delivered
- ✅ [What was completed]
### Decisions Made
- [Key decisions from last week]
---
## Team
### Core Contributors (Active)
| Name | Focus | Availability |
|------|-------|--------------|
| [Name] | [Area] | Full-time/Part-time |
### Occasional Contributors
- [Name] - [Role]
---
## Milestones
### In Progress 🔄
| Milestone | Target | Business Objective |
|-----------|--------|-------------------|
| [Name] | [Date] | [WHY this matters] |
### Recently Completed ✅
| Milestone | Date | Business Objective |
|-----------|------|-------------------|
| [Name] | [Date] | [WHY this mattered] |
### Lost in Sight / Paused ⏸️
| Milestone | Status | Reason |
|-----------|--------|--------|
| [Name] | Paused | [Why paused] |
---
## Recent Decisions
### Week [N] (Current)
- **[Decision]** - [Context]
### Week [N-1]
- **[Decision]** - [Context]
---
## Quick Links
- [📊 Timeline](./timeline/index.md) - Week-by-week history
- [📋 Background](./background.md) - Project architecture and details
- [🔌 Data Sources](./datasources.md) - How to gather information
- [⚙️ Sync State](./sync-state.md) - Last sync information
---
*This is a living document. It reflects the current state and changes frequently.*
```
## Step 4: Create Background Skeleton
**File:** `projects/$0/background.md`
Static/architecture information that rarely changes.
```markdown
# $0 - Background
**Type:** [Web app/Mobile app/Library/Service]
**Repository:** URL
## What is $0?
[Brief description of what the project does]
## Architecture
### Components
- Component 1 - Purpose
- Component 2 - Purpose
### Technology Stack
- Technology 1 - Usage
- Technology 2 - Usage
## Data Sources
See: [datasources.md](./datasources.md)
## Timeline Structure
Weekly timeline files are organized in `timeline/` directory.
## How This Project Is Updated
1. Gather Data: Query Zulip, Git, meetings
2. Update Timeline: Create week-by-week entries
3. Update Project Dashboard: Refresh [project.md](./project.md)
For current status, see: [project.md](./project.md)
```
## Step 4: Create Timeline Index
**File:** `projects/$0/timeline/index.md`
```markdown
# $0 Timeline Index
## Key Milestones
| Date | Milestone | Status |
|------|-----------|--------|
| [To be filled by project-history] | | |
## Summary by Quarter
[To be filled by project-history]
```
## Step 5: Initialize Sync State
**File:** `projects/$0/sync-state.md`
```markdown
# Sync State
status: initialized
created_at: [today's date]
last_sync_date: null
initial_history_from: null
initial_history_to: null
```
## Done
After this skill completes, the user should:
1. **Review `datasources.md`** — confirm the streams, repos, and meeting rooms are correct
2. **Edit `background.md`** — fill in any known project details
3. **Run `/project-history $0 [date-from] [date-to]`** — to build the initial historical timeline

344
.agents/skills/project-sync/SKILL.md Normal file
View File

@@ -0,0 +1,344 @@
---
name: project-sync
description: Sync a project timeline using subagents for parallelism. Splits work by week and datasource to stay within context limits. Handles both first-time and incremental syncs.
disable-model-invocation: true
argument-hint: [project-name]
---
# Project Sync
**When to use:** Keep a project timeline up to date. Works whether the project has been synced before or not.
**Precondition:** `projects/$0/datasources.md` must exist. If it doesn't, run `/project-init $0` first.
## Architecture: Coordinator + Subagents
This skill is designed for **subagent execution** to stay within context limits. The main agent acts as a **coordinator** that delegates data-intensive work to subagents.
```
Coordinator
├── Phase 1: Gather (parallel subagents, one per datasource)
│ ├── Subagent: Zulip → writes tmp/$0-sync/zulip.md
│ ├── Subagent: Git → writes tmp/$0-sync/git.md
│ └── Subagent: Meetings → writes tmp/$0-sync/meetings.md
│
├── Phase 2: Synthesize (parallel subagents, one per week)
│ ├── Subagent: Week 1 → writes timeline/{year-month}/week-{n}.md
│ ├── Subagent: Week 2 → writes timeline/{year-month}/week-{n}.md
│ └── ...
│
└── Phase 3: Finalize (coordinator directly)
├── timeline/index.md (add links to new weeks)
├── project.md (update living document)
└── sync-state.md (update sync status)
```
---
## Coordinator Steps
### Step 1: Determine Sync Range
Check whether `projects/$0/sync-state.md` exists.
**Case A — First sync (no sync-state.md):**
Default range is **last 12 months through today**. If the user provided explicit dates as extra arguments (`$1`, `$2`), use those instead.
**Case B — Incremental sync (sync-state.md exists):**
Read `last_sync_date` from `projects/$0/sync-state.md`. Range is `last_sync_date` to today.
### Step 2: Read Datasources
Read `projects/$0/datasources.md` to determine:
- Zulip stream IDs and search terms
- Git repository URL
- Meeting room names
- Entity types to prioritize
### Step 3: Prepare Scratch Directory
```bash
mkdir -p tmp/$0-sync
```
This directory holds intermediate outputs from Phase 1 subagents. It is ephemeral — delete it after the sync completes.
### Step 4: Compute Week Boundaries
Split the sync range into ISO calendar weeks (Monday–Sunday). Produce a list of `(week_number, week_start, week_end, year_month)` tuples. This list drives Phase 2.
---
## Phase 1: Gather Data (parallel subagents)
Launch **one subagent per datasource**, all in parallel. Each subagent covers the **full sync range** and writes its output to a scratch file. The output must be organized by week so Phase 2 subagents can consume it.
### Subagent: Zulip
**Input:** Sync range, PRIMARY stream IDs and search terms from datasources.md.
**Important:** `threaded_conversation` entities only contain the **last 50 messages** in a topic. To get complete message history for a week, you must query `conversation_message` entities.
**Task:** Two-step process for each PRIMARY stream:
**Step 1:** List all thread IDs in the stream using `id_prefix`:
```
GET /api/v1/query
entity_types=threaded_conversation
connector_ids=zulip
id_prefix=zulip:stream:{stream_id}
limit=100
offset=0
```
This returns all thread entities (e.g., `zulip:stream:155:topic_name`). Save these IDs.
**Step 2:** For each week in the sync range, query messages from each thread:
```
GET /api/v1/query
entity_types=conversation_message
connector_ids=zulip
parent_id={thread_id} # e.g., zulip:stream:155:standalone
date_from={week_start}
date_to={week_end}
limit=100
offset=0
```
Paginate through all messages for each thread/week combination.
**Output:** Write `tmp/$0-sync/zulip.md` with results grouped by week:
```markdown
## Week {n} ({week_start} to {week_end})
### Stream: {stream_name}
- **Topic:** {topic} ({date}, {message_count} messages, {participant_count} participants)
{brief summary or key quote}
```
### Subagent: Git
**Input:** Sync range, git repository URL from datasources.md.
**Task:**
**Important:** Git commands may fail due to gitconfig permission issues. Use a temporary HOME directory:
```bash
# Set temporary HOME to avoid gitconfig permission issues
export HOME=$(pwd)/.tmp-home
mkdir -p ./tmp
# Clone if needed, pull if exists
if [ -d ./tmp/$0-clone ]; then
export HOME=$(pwd)/.tmp-home && cd ./tmp/$0-clone && git pull
else
export HOME=$(pwd)/.tmp-home && git clone --depth 500 {url} ./tmp/$0-clone
cd ./tmp/$0-clone
fi
# Get commits in the date range
export HOME=$(pwd)/.tmp-home && git log --since="{range_start}" --until="{range_end}" --format="%H|%an|%ae|%ad|%s" --date=short
# Get contributor statistics
export HOME=$(pwd)/.tmp-home && git log --since="{range_start}" --until="{range_end}" --format="%an" | sort | uniq -c | sort -rn
```
**Output:** Write `tmp/$0-sync/git.md` with results grouped by week:
```markdown
## Week {n} ({week_start} to {week_end})
**Commits:** {count}
**Contributors:** {name} ({count}), {name} ({count})
### Key Commits
- `{short_hash}` {subject} — {author} ({date})
```
### Subagent: Meetings
**Input:** Sync range, meeting room names from datasources.md.
**Task:** For each PRIMARY room, query meetings and run semantic search:
```
GET /api/v1/query
entity_types=meeting
date_from={range_start}
date_to={range_end}
room_name={room-name}
limit=100
POST /api/v1/search
search_text={project-name}
entity_types=["meeting"]
date_from={range_start}
date_to={range_end}
limit=50
```
**Output:** Write `tmp/$0-sync/meetings.md` with results grouped by week:
```markdown
## Week {n} ({week_start} to {week_end})
### Meeting: {title} ({date}, {room})
**Participants:** {names}
**Summary:** {brief summary}
**Key points:**
- {point}
```
---
## Phase 2: Synthesize Week Files (parallel subagents)
After all Phase 1 subagents complete, launch **one subagent per week**, all in parallel. Each produces a single week file.
### Subagent: Week {n}
**Input:** The relevant `## Week {n}` sections extracted from each of:
- `tmp/$0-sync/zulip.md`
- `tmp/$0-sync/git.md`
- `tmp/$0-sync/meetings.md`
Pass only the sections for this specific week — do NOT pass the full files.
**Task:** Merge and analyze the data from all three sources. Categorize into:
1. **Key Decisions** — Technology migrations, architecture changes, vendor switches, security incidents, strategic pivots
2. **Technical Work** — Feature implementations, bug fixes, infrastructure changes
3. **Team Activity** — Core vs. occasional contributors, role changes
4. **Blockers** — Issues, delays, dependencies
**Milestones:** When documenting milestones, capture BOTH:
- **WHAT** — The technical achievement (e.g., "PostgreSQL migration")
- **WHY** — The business objective (e.g., "to improve query performance from 107ms to 27ms and enable concurrent access for scaling")
Search for business objectives in: meeting discussions about roadmap, Zulip threads about planning, PR descriptions, release notes, and any "why are we doing this" conversations.
**Skip unless meaningful:** Routine check-ins, minor documentation updates, social chat.
**Output:** Write `projects/$0/timeline/{year-month}/week-{n}.md` using the week file template from [project-history](../project-history/SKILL.md). Also return a **3-5 line summary** to the coordinator for use in Phase 3.
Create the month directory first if needed: `mkdir -p projects/$0/timeline/{year-month}`
---
## Phase 3: Finalize (coordinator directly)
The coordinator collects the summaries returned by all Phase 2 subagents. These summaries are small enough to fit in the coordinator's context.
### Step 5: Update Timeline Index
Add links to new week files in `projects/$0/timeline/index.md`. Append entries under the appropriate year/quarter sections. Update milestones if any were reached.
### Step 6: Update Project Dashboard (project.md)
**File:** `projects/$0/project.md`
This is the **living document** — update it with current status from the week summaries:
**Update these sections:**
1. **This Week's Focus** - What the team is actively working on now
2. **Last Week's Focus** - What was completed in the most recent week
3. **Team** - Current contributors and their focus areas
4. **Milestones** - Update status and add new ones with business objectives
5. **Recent Decisions** - Key decisions from the last 2-3 weeks
**Milestone Format:**
```markdown
### In Progress 🔄
| Milestone | Target | Business Objective |
|-----------|--------|-------------------|
| Standalone deployment | Feb 2026 | Enable non-developers to self-host without complex setup |
### Recently Completed ✅
| Milestone | Date | Business Objective |
|-----------|------|-------------------|
| PostgreSQL migration | Mar 2025 | Improve performance (107ms→27ms) and enable scaling |
### Lost in Sight / Paused ⏸️
| Milestone | Status | Reason |
|-----------|--------|--------|
| Feature X | Paused | Resources reallocated to higher priority |
```
**Note:** Milestones in this company change frequently — update status (in progress/done/paused) as needed.
### Step 7: Update Sync State
Create or update `projects/$0/sync-state.md`:
**First sync (Case A):**
```markdown
# Sync State
status: synced
created_at: {today's date}
last_sync_date: {today's date}
initial_history_from: {range_start}
initial_history_to: {range_end}
last_incremental_sync: {today's date}
```
**Incremental sync (Case B):**
```markdown
# Sync State
status: synced
created_at: {original value}
last_sync_date: {today's date}
initial_history_from: {original value}
initial_history_to: {original value}
last_incremental_sync: {today's date}
```
### Step 8: Cleanup
```bash
rm -rf tmp/$0-sync
```
### Step 9: Summary Report
Output a brief summary:
```markdown
## Sync Summary: {Date}
### Period Covered
{range_start} to {range_end}
### Key Changes
1. Decision: {brief description}
2. Feature: {what was built}
3. Team: {who joined/left}
### Metrics
- {n} new commits
- {n} active contributors
- {n} weeks analyzed
- {n} new Zulip threads
- {n} meetings recorded
### Current Status
[Status description]
```
---
## Key Rules
- **Link to sources**: Always reference commit hashes, PR numbers, Zulip topic names, meeting dates
- **Be explicit about exclusions**: Document what you're NOT analyzing and why
- **Write once**: Week files are historical records — don't modify existing ones, only create new ones
- **Paginate all queries**: Always loop through all pages of results
- **Distinguish contributor types**: Core (regular activity) vs. occasional (sporadic)
- **Subagent isolation**: Each subagent should be self-contained. Pass only the data it needs — never the full scratch files
- **Fail gracefully**: If a datasource subagent fails (e.g., git clone errors, API down), the coordinator should continue with available data and note the gap in the summary

105
.agents/skills/workflow/SKILL.md Normal file
View File

@@ -0,0 +1,105 @@
---
name: workflow
description: Create a marimo notebook for data analysis. Use when the request involves analysis over time periods, large data volumes, or when the user asks to "create a workflow".
disable-model-invocation: true
argument-hint: [topic]
---
# Workflow — Create a Marimo Notebook
## When to create a marimo notebook
Any request that involves **analysis over a period of time** (e.g., "meetings this month", "emails since January", "interaction trends") is likely to return a **large volume of data** — too much to process inline. In these cases, **always produce a marimo notebook** (a `.py` file following the patterns in the [notebook-patterns skill](.agents/skills/notebook-patterns/SKILL.md)).
Also create a notebook when the user asks to "create a workflow", "write a workflow", or "build an analysis".
If you're unsure whether a question is simple enough to answer directly or needs a notebook, **ask the user**.
## Always create a new workflow
When the user requests a workflow, **always create a new notebook file**. Do **not** modify or re-run an existing workflow unless the user explicitly asks you to (e.g., "update workflow 001", "fix the sentiment notebook", "re-run the existing analysis"). Each new request gets its own sequentially numbered file — even if it covers a similar topic to an earlier workflow.
## File naming and location
All notebooks go in the **`workflows/`** directory. Use a sequential number prefix so workflows stay ordered by creation:
```
workflows/<NNN>_<topic>_<scope>.py
```
- `<NNN>` — zero-padded sequence number (`001`, `002`, …). Look at existing files in `workflows/` to determine the next number.
- `<topic>` — what is being analyzed, in snake_case (e.g., `greyhaven_meetings`, `alice_emails`, `hiring_discussions`)
- `<scope>` — time range or qualifier (e.g., `january`, `q1_2026`, `last_30d`, `all_time`)
**Examples:**
```
workflows/001_greyhaven_meetings_january.py
workflows/002_alice_emails_q1_2026.py
workflows/003_hiring_discussions_last_30d.py
workflows/004_team_interaction_timeline_all_time.py
```
**Before creating a new workflow**, list existing files in `workflows/` to find the highest number and increment it.
## Plan before you implement
Before writing any notebook, **always propose a plan first** and get the user's approval. The plan should describe:
1. **Goal** — What question are we answering?
2. **Data sources** — Which entity types and API endpoints will be used?
3. **Algorithm / ETL steps** — Step-by-step description of the data pipeline: what gets fetched, how it's filtered, joined, or aggregated, and what the final output looks like.
4. **Output format** — Table columns, charts, or summary statistics the user will see.
Only proceed to implementation after the user confirms the plan.
## Validate before delivering
After writing or editing a notebook, **always run `uvx marimo check`** to verify it has no structural errors (duplicate variables, undefined names, branch expressions, etc.):
```bash
uvx marimo check workflows/NNN_topic_scope.py
```
A clean check (no output, exit code 0) means the notebook is valid. Fix any errors before delivering the notebook to the user.
## Steps
1. **Identify people** — Use ContactDB to resolve names/emails to `contact_id` values. For "me"/"my" questions, always start with `GET /api/contacts/me`.
2. **Find data** — Use DataIndex `GET /query` (exhaustive, paginated) or `POST /search` (semantic, ranked) with `contact_ids`, `entity_types`, `date_from`/`date_to`, `connector_ids` filters.
3. **Analyze** — For simple answers, process the API response directly. For complex multi-step analysis, build a marimo notebook (see the [notebook-patterns skill](.agents/skills/notebook-patterns/SKILL.md) for detailed patterns).
## Quick Example (Python)
> "Find all emails involving Alice since January"
```python
import httpx
CONTACTDB = "http://localhost:42000/contactdb-api"
DATAINDEX = "http://localhost:42000/dataindex/api/v1"
client = httpx.Client(timeout=30)
# 1. Resolve "Alice" to a contact_id
resp = client.get(f"{CONTACTDB}/api/contacts", params={"search": "Alice"})
alice_id = resp.json()["contacts"][0]["id"] # e.g. 42
# 2. Fetch all emails involving Alice (with pagination)
emails = []
offset = 0
while True:
resp = client.get(f"{DATAINDEX}/query", params={
"entity_types": "email",
"contact_ids": str(alice_id),
"date_from": "2025-01-01T00:00:00Z",
"limit": 50,
"offset": offset,
})
data = resp.json()
emails.extend(data["items"])
if offset + 50 >= data["total"]:
break
offset += 50
print(f"Found {len(emails)} emails involving Alice")
```