- PLAN.md: add task 2b — scripts/podman-agent.mjs writes git signals directly to MongoDB (changedFiles, diffStat, branch, recentCommit) every 15s; Hermes reads merged vision+git state for event detection - mongodb.md: extend engineer_states with git fields, document two-writer upsert pattern (vision fields vs git fields, no conflict) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FFbfi4Cmb7BY75Wtne7bZn
4.5 KiB
MongoDB Atlas Integration Spec
MongoDB Atlas is PodMan's shared memory. It stores live engineer state, the ownership map that enables continual learning, coordination events, and nudge history.
Collections
engineer_states
Latest context per engineer. Two writers, one collection — vision pipeline upserts vision fields, git watcher script upserts git fields independently. Hermes reads the merged document for event detection.
{
_id: string, // engineerId (stable across sessions)
podId: string,
name: string, // display name
// --- Vision fields (written by Hermes via POST /ingest) ---
currentFile: string | null, // active file inferred from screen
inferredTask: string | null, // what engineer appears to be doing
terminalVisible: boolean,
recentTerminalOutput: string | null,
confidence: number, // Gemini Vision confidence (0–1)
visionUpdatedAt: Date,
// --- Git fields (written directly by scripts/podman-agent.mjs) ---
changedFiles: string[], // files with uncommitted changes (git status)
diffStat: string | null, // e.g. "auth/middleware.ts | 24 +++++"
recentCommit: string | null, // most recent commit message
branch: string | null, // current branch name
gitUpdatedAt: Date,
// --- Shared ---
updatedAt: Date // most recent write from either source
}
Index: { podId: 1, updatedAt: -1 }
Two writers, no conflict: vision upsert uses $set on vision fields only; git upsert uses $set on git fields only. MongoDB upsert semantics merge them cleanly.
Usage: Hermes reads all documents for a given podId after each update to run event detection. Both vision and git context are available in the same document — changedFiles provides ground truth, currentFile provides screen context.
ownership_map
Tracks who works on which files. Built up over the session. Persists across sessions — this is the continual learning artifact.
{
_id: string, // `${podId}:${file}`
podId: string,
file: string,
primaryOwner: string, // engineerId with most recent activity on this file
contributors: string[], // all engineerIds observed on this file
observationCount: number, // total frames where this file was seen
lastSeenAt: Date
}
Index: { podId: 1, file: 1 } (unique)
Upsert logic:
- On each context update where
currentFileis non-null:- Increment
observationCount - Update
primaryOwnerto the engineer with the most recentlastSeenAton this file - Add engineerId to
contributorsif not present - Update
lastSeenAt
- Increment
Continual learning: Hermes loads this collection on startup for the pod. If history exists, it pre-populates the in-memory ownership cache before the first frame arrives.
events
Every coordination event detected by Hermes.
{
_id: ObjectId,
podId: string,
type: 'DEPENDENCY_READY' | 'BLOCKER_DETECTED' | 'DUPLICATE_WORK',
involvedEngineers: string[],
file: string | null,
reason: string, // 1-sentence explanation from Gemini
nudgeSent: boolean, // false if suppressed by cooldown
detectedAt: Date
}
Index: { podId: 1, detectedAt: -1 }
nudges
Every voice nudge sent to the room.
{
_id: ObjectId,
podId: string,
eventId: ObjectId, // ref to events collection
targetEngineers: string[],
message: string, // the spoken text
sentAt: Date
}
Index: { podId: 1, sentAt: -1 }
Cooldown check: before sending a nudge, Hermes queries this collection for any nudge in the last 3 minutes for the same podId. If found, suppresses the new nudge and marks the event as nudgeSent: false.
Hermes startup sequence
1. Connect to Atlas using MONGODB_URI
2. Load ownership_map for this podId
3. Build in-memory cache: Map<file, { primaryOwner, contributors }>
4. Begin accepting /ingest requests
Atlas configuration
- Cluster tier: M0 (free) is sufficient for hackathon scale
- Region: same as DigitalOcean deployment (e.g. NYC1)
- Auth: connection string in
MONGODB_URIenv var - Collections created automatically on first write (no schema migration needed)
What MongoDB does NOT store
- Raw screenshot frames (too large — frames are processed in-memory by Hermes and discarded)
- Full Gemini response objects (only extracted fields are stored)
- Session recordings