dsh-trajectory-governance

Agent trajectory governance & anomaly diagnosis for DeepSeek Harness (dsh).

Rebuilds the flat session/event log into a structured, multi-branch trajectory tree, keeps observation-layer snapshots, and runs three temporal anomaly strategies (loop deadlock / invalid retry / goal drift) with results mounted on tree nodes and surfaced in an independent GUI tab.

Zero kernel modification · consume-only event subscription · observation-layer snapshots (no control-plane rollback) · independent SQLite storage · never overwrites the official Trajectory view.

Why

The official Trajectory page is a flat raw-event log. For long-running agent tasks (dozens of turns, subagents, forks) you cannot see:

• which tool call belongs to which sub-task / subagent / fork branch,
• where the agent started drifting from the original goal,
• whether it is looping in place burning tokens (deadlock / invalid retry).

This plugin is an upper observation & reasoning layer: it consumes committed events and derives structure, snapshots, and diagnoses — it never steers, intercepts, or rolls back the agent. Both views coexist.

Architecture

dsh web profile
 ├─ @deepseek-ai/dsh-base (official core)
 └─ dsh-trajectory-governance (this bundle)
      │
      │  ctx.on('session/event' | 'session/created' | 'session/disposed'
      │         | 'subagent/start' | 'subagent/end')   ← emit-mode, post-commit, read-only
      ▼
 ┌─ ingest ───────────────┐   ┌─ tree ─────────────────┐
 │ normalize (scalars)    │──▶│ cross-session trajectory │
 │ sqlite (events/sessions)│   │ tree + branch attach    │
 └────────────────────────┘   └─────────────────────────┘
 ┌─ snapshot ─────────────┐   ┌─ diagnose ──────────────┐
 │ index-only snapshots    │   │ 3 strategies (async)    │
 │ LRU/TTL prune, branches │   │ results → anomalies DB  │
 └────────────────────────┘   └─────────────────────────┘
      │
      ▼
 ┌─ api (host) ────────────┐   ┌─ client (browser) ─────┐
 │ /trajectory-governance/  │◀──│ independent Tab:        │
 │ api/* (JSON, same-origin)│   │ conversation.view slot, │
 └─────────────────────────┘   │ id 'trajectory-…'       │
                               └─────────────────────────┘

Data flow: session/event (durable, post-commit) → normalized + persisted → tree/snapshot/diagnosis services → JSON API → the GUI tab. The diagnosis engine runs asynchronously on an interval (incremental watermark + tail overlap) and never blocks the agent loop.

Install

Requires dsh >= 0.1.0-rc.6. Two ways:

Bundle (published / git):

dsh plugin --profile web add github:dfycaly98931680/dsh-trajectory-governance#<commit-sha>

A git install fetches sources, so the package ships a prepare build; pnpm requires an explicit allow-list for it (add the printed key to the profile's pnpm-workspace.yaml, then re-run):

allowBuilds:
  dsh-trajectory-governance: true

Alternatively install a built tarball (pnpm pack) — no build permission needed.

Dev overlay (no pnpm):

dsh web --patch C:\path\to\dsh-plugin\overlay.dev.yml

Then restart the web app. You should see [trajectory-governance] loaded; storage=... sessions=N events=M anomalies=K in the logs.

Use

1. Run any agent task in the GUI (ideally a long one).
2. Open the 轨迹治理 / Trajectory Governance tab (a separate conversation.view entry — the official Trajectory tab is untouched).
3. Pick a session: the tree renders with branch badges (subagent/fork), snapshot marks (★), and red-highlighted anomaly ranges (⚠).
4. Click a node for the raw event + mounted diagnosis report (type, confidence, severity, description, suggestion).
5. Use the snapshot bar to capture checkpoints at any node, list them, and jump back; delete stale branches via the API/CLI.

CLI inspection utilities (against the plugin's own SQLite):

node scripts/inspect.mjs   [dbPath]                 # sessions/events overview
node scripts/tree.mjs      <sessionId> [dbPath]     # trajectory tree JSON
node scripts/snapshot.mjs  <dbPath> <cmd> [...]     # list|create|context|delete|prune

Configuration

The plugin row accepts config (all optional; deep-merged over defaults):

- insert:
    - id: trajectory-governance
      name: dsh-trajectory-governance
      config:
        storage:
          path: C:/data/trajectory.db   # default ~/.dsh-trajectory-governance/trajectory.db
        diagnosis:
          strategyA:
            enabled: true
            windowSize: 5
            similarityThreshold: 0.85
            resultSimilarityThreshold: 0.8
          strategyB:
            enabled: true
            minRounds: 3
          strategyC:
            enabled: true
            sampleEveryNRounds: 5
            similarityThreshold: 0.5
            consecutiveSamples: 2
            embedder: lexical        # 'lexical' built-in; 'llm' is the extension seam
          analyzer:
            enabled: true
            intervalMs: 10000
          alerts:
            enabled: true
            minConfidence: 0.8
            cooldownMs: 600000
            channels:
              desktop: true          # browser Notification (client side)
              cordisEvent: true      # 'trajectory/anomaly' bus event for the ecosystem
              webhook:
                url: ""              # Feishu/DingTalk/Slack-style text webhook
          actions:
            onLoopDeadlock: notify   # notify | suggest-stop(一键中断按钮) | auto-stop
            onInvalidRetry: notify
            onGoalDrift: notify
          cost:
            enabled: true
            tokenPricePerM: 0.28     # adjust to your provider pricing
            estimateWhenUsageMissing: true

Full schema: DIAGNOSIS_CONFIG_SCHEMA in src/diagnose/config.ts.

Alerts & stop-loss (v0.2)

When a NEW anomaly is detected (confidence ≥ alerts.minConfidence), the plugin:

1. emits trajectory/anomaly on the Cordis bus (ecosystem notification plugins can subscribe) and POSTs a text webhook if configured — the message is in stop-loss language: type, confidence, and how much tokens/money/time the range already consumed (from official assistant/message.usage);
2. the GUI tab polls /api/alerts, shows a desktop notification (browser Notification API, permission button in the bar), and pins abnormal sessions on top of the dropdown sorted by wasted tokens;
3. offers a one-click interrupt (POST /api/actions/interrupt) which calls the official agent.cancel capability — never a monkey-patch. Default tier is notify; auto-stop (loop deadlock only, confidence ≥ 0.9) can be configured for unattended setups.

Breakpoints & waste reports (v0.3)

Breakpoint exit — the legitimate lightweight "rollback". Instead of a control-plane rewind (already covered by dsh-turn-rewind), this plugin offers a diagnosis-driven restart: at any snapshot / anomaly start / selected node, "branch a new session here" resolves the target down to the nearest stable turn/end boundary and calls the official ctx.sessions.fork(source, boundary) — a fresh child session seeded at that point. It never touches workspace files, keeps the observation-layer stance, and turns "you wasted money here" into a one-click action.

Waste attribution report — not another usage-stats dashboard (that niche is saturated): it answers the attribution question nobody else does, "which sessions wasted tokens/money, on which anomaly types":

curl http://127.0.0.1:3080/trajectory-governance/api/report/waste          # all sessions (Markdown + JSON)
curl "http://127.0.0.1:3080/trajectory-governance/api/report/waste?sessionId=<id>"  # one session detail

Both are also reachable from the tab ("浪费报告" / "本会话报告" buttons).

Quick demo (reproduce loop deadlock & goal drift)

Instant (no waiting for a real loop): seed a synthetic session into the live store, refresh the tab, pick demo-session:

node scripts/demo-seed.mjs            # writes into the default live DB

You should see a tree with red-highlighted loop_deadlock (5 identical tool calls) and a severe goal_drift range, plus a snapshot mark (★).

Real agent:

1. Give the agent a long, open-ended refactor prompt (e.g. "重构这个项目为 TypeScript 并补单元测试，先读代码再动手").
2. Let it run until it starts repeating the same tool call (same file, same content) with no visible progress — Strategy A flags loop_deadlock on the trigger node.
3. Ask a follow-up in a different domain ("顺便帮我看看怎么做番茄意面") — after N sampled turns Strategy C flags goal_drift (mild/moderate/severe) against the original baseline.

Verify: the tree tab highlights the ranges in red; scripts/inspect.mjs shows the anomaly counts; scripts/tree.mjs <sessionId> dumps the mounted anomalyInfo.

Screenshots

TODO — add after the first live run: tree tab overview / anomaly card / snapshot list. (Placeholders reserved here.)

Testing

npm install
npm run build && npm test    # 46 tests: normalize/store/subscriber/tree/snapshot/diagnose/api

Boundaries (important)

• Zero kernel modification — no fork, no monkey-patch, no host storage access.
• Consume-only events — session/event etc. are emit-mode, post-commit, fire-and-forget listeners; payloads are never mutated.
• Observation-layer snapshots — a snapshot is an index (nodeId + context hash
  • branchId); context is rebuilt from the event store. This plugin never forks, resumes, or rewinds the agent.
• Independent Tab — registered with its own slot id; the official Trajectory page (id trajectory) is never overwritten.
• Preview API — SESSION_FORMAT_VERSION = 0, no compatibility implied. The event vocabulary is centralized in src/core/event-catalog.ts; unknown types honor the envelope's ignorable marker.

Repository layout

src/
  index.ts                  # Cordis plugin entry (name/apply)
  core/                     # types + 44-type event catalog
  ingest/                   # normalize + session-event subscriber
  persistence/              # node:sqlite store (events/sessions/snapshots/anomalies)
  tree/                     # trajectory tree builder + queries
  snapshot/                 # observation snapshots + branch manager
  diagnose/                 # similarity, config, 3 strategies, engine, mount, runtime
  api/                      # host JSON API (/trajectory-governance/api/*)
  client/                   # independent GUI tab (conversation.view, id 'trajectory-governance')
tests/                      # node:test suites
scripts/                    # inspect / tree / snapshot / smoke
cordis.patch.yml            # official bundle load row
plugin.manifest.yaml        # self-describing metadata (loader uses cordis.patch.yml)

License

MIT — see LICENSE. Part of the awesome-dsh-plugin ecosystem. Topic: dsh-plugin.