The P3 dispatcher's default seams shell out to gh/git, but the R720 box has
no gh and a read-only PAT with no Actions scope — so dispatch_apply_verify
returned no run_id and every task parked at verify ("dispatch unresolved").
Add a GitHub-App auth path: the box mints short-lived (~1h) installation
access tokens from the App private key and uses them for the three dispatch
seams, removing the gh dependency.
- agent_team/github_app.py (new): mint_installation_token (RS256 App JWT,
iss=app_id, iat backdated 60s, exp 9 min; POST /access_tokens) + a lazy
TokenProvider that caches and re-mints near expiry. Secret-safe: the JWT
and token are never logged, never in an exception message, never persisted.
- dispatcher.py: app_branch_pusher / app_workflow_dispatcher / app_run_locator
(additive; gh/git _default_* left untouched). Push auth rides a host-scoped
http.extraHeader via GIT_CONFIG_* env (token never in argv/ps); the REST
run locator maps id->databaseId / created_at->createdAt into select_run_id
and surfaces 4xx promptly instead of silently exhausting the poll window.
- coordinator.py: default_dispatch_node_factory binds the App seams when
AGENT_TEAM_GH_APP_ID / _INSTALLATION_ID / _PRIVATE_KEY are all set; partial
or unreadable config logs one warning and falls back to gh-default (never
raises at serve-start).
- requirements.txt: pin PyJWT, cryptography, requests (App seams + CI fetcher).
- DEPLOY-R720.md / README.md: App dispatch config, permission/scope audit,
env-precedence check, key rotation/revocation + incident response.
Tests: +18 (test_github_app.py new; dispatcher/coordinator additions) covering
JWT claims, cache/re-mint, token-scrub-on-error, REST field mapping + run-name
correlation, and the partial-env inert fallback. Full suite 1523 passing.
Security follow-ups from the per-PR review (non-blocking MEDIUMs):
- Eager _get_token() at make_app build time so a missing AGENT_TEAM_API_TOKEN
fails fast instead of serving requests first (matches the docstring contract).
- Disable /docs, /redoc, /openapi.json (no auth dependency in FastAPI) — the
API is VPN-only/127.0.0.1 and should not expose its schema unauthenticated.
- Scrub raw exception text and subprocess stderr from 500 response bodies;
log server-side instead (avoid internal-path/state disclosure).
- Bound /orchestrator/invoke concurrency with a semaphore (429 over the cap)
so an authenticated caller cannot exhaust the box via many 600s subprocesses.
Also pin fastapi/uvicorn in requirements.txt (WS1 dep). With fastapi now
installed in CI, the previously skip-guarded TestClient tests run for real;
the importorskip guard stays as a no-op safety net.
Tests: 23 pass (adds docs-disabled + concurrency-429 cases).
Reworks the P1 sim so the four §7.1 exit criteria are demonstrated against the
ACTUAL mechanic, not a model (resolves the verifier's "sim models the ledger,
not the LangGraph integration" finding).
- New tests/sim/test_p1_graph_integration.py drives the real agent_team.graph
StateGraph (interrupt/Command(resume)) + the real langgraph SqliteSaver
checkpointer + the committed pending_questions compare-and-set, proving:
(a) suspend survives a simulated restart (drop saver/conn, rebuild over the
same checkpoint DB) and resumes; (b) duplicate answer loses the CAS and the
graph never double-advances; (c) a post-deadline answer loses to expire and
the task is not resumed; (d) two concurrent tasks resume to the correct
thread, with a turn-guarded no-double-apply check.
- graph.py: derive a STABLE question_id from uuid5(thread_id, turn). The
clarifier node replays on resume, so the prior fresh-uuid id changed between
the delivered/ledgered question and the qa_history entry — breaking the
§3.3.1 identity contract. Now the delivered id == ledger key == history entry
(unit-tested in test_graph.py).
- harness._connect() now uses the committed schema.connect() (WAL + busy_timeout)
instead of a raw sqlite3.connect, so concurrent responders genuinely serialize;
the criterion-(d) concurrency test no longer swallows OperationalError (it
asserts zero errors + exactly one CAS winner).
- requirements.txt: pin langgraph-checkpoint-sqlite==3.1.0 (design D9 durable
checkpointer), now exercised by the integration test.
Full suite: 564 passed; ruff + format clean.
Multi-model agent routing with 7 agent nodes (Sonnet, Haiku,
GPT-4.1, Gemini, DeepSeek) and 15 pre-loaded Composio tools
for Slack, Notion, GitHub, and Google Drive integration.
