fix(federation): add restart policies + M2-04 OID assertion criterion

Address review feedback on PR #490:
- Add `restart: unless-stopped` to postgres-federated, valkey-federated,
  step-ca services so they auto-recover after host reboot / docker restart.
- Update FED-M2-04 acceptance: must wire federation.tpl template into
  mosaic-fed provisioner config AND include unit/integration test asserting
  issued certs contain BOTH custom OIDs (grantId + subjectUserId). Carry-
  forward guard from M2-02 review against silent OID stripping.

.gitignore

@@ -9,3 +9,6 @@ coverage
 *.tsbuildinfo
 .pnpm-store
 docs/reports/
+
+# Step-CA dev password — real file is gitignored; commit only the .example
+infra/step-ca/dev-password

docker-compose.federated.yml

@@ -27,6 +27,7 @@ services:
   postgres-federated:
     image: pgvector/pgvector:pg17
     profiles: [federated]
+    restart: unless-stopped
     ports:
       - '${PG_FEDERATED_HOST_PORT:-5433}:5432'
     environment:
@@ -45,6 +46,7 @@ services:
   valkey-federated:
     image: valkey/valkey:8-alpine
     profiles: [federated]
+    restart: unless-stopped
     ports:
       - '${VALKEY_FEDERATED_HOST_PORT:-6380}:6379'
     volumes:
@@ -55,6 +57,64 @@ services:
       timeout: 3s
       retries: 5
 
+  # ---------------------------------------------------------------------------
+  # Step-CA — Mosaic Federation internal certificate authority
+  #
+  # Image: pinned to 0.27.4 (latest stable as of late 2025).
+  # `latest` is forbidden per Mosaic image policy (immutable tag required for
+  # reproducible deployments and digest-first promotion in CI).
+  #
+  # Profile: `federated` — this service must not start in non-federated dev.
+  #
+  # Password:
+  #   Dev:  bind-mount ./infra/step-ca/dev-password (gitignored; copy from
+  #         ./infra/step-ca/dev-password.example and customise locally).
+  #   Prod: replace the bind-mount with a Docker secret:
+  #           secrets:
+  #             ca_password:
+  #               external: true
+  #         and reference it as `/run/secrets/ca_password` (same path the
+  #         init script already uses).
+  #
+  # Provisioner: "mosaic-fed" (consumed by apps/gateway/src/federation/ca.service.ts)
+  # ---------------------------------------------------------------------------
+  step-ca:
+    image: smallstep/step-ca:0.27.4
+    profiles: [federated]
+    restart: unless-stopped
+    ports:
+      - '${STEP_CA_HOST_PORT:-9000}:9000'
+    volumes:
+      - step_ca_data:/home/step
+      # init script — executed as the container entrypoint
+      - ./infra/step-ca/init.sh:/usr/local/bin/mosaic-step-ca-init.sh:ro
+      # X.509 template skeleton (wired in M2-04)
+      - ./infra/step-ca/templates:/etc/step-ca-templates:ro
+      # Dev password file — GITIGNORED; copy from dev-password.example
+      # In production, replace this with a Docker secret (see comment above).
+      - ./infra/step-ca/dev-password:/run/secrets/ca_password:ro
+    entrypoint: ['/bin/sh', '/usr/local/bin/mosaic-step-ca-init.sh']
+    healthcheck:
+      # The healthcheck requires the root cert to exist, which is only true
+      # after init.sh has completed on first boot. start_period gives init
+      # time to finish before Docker starts counting retries.
+      test:
+        [
+          'CMD',
+          'step',
+          'ca',
+          'health',
+          '--ca-url',
+          'https://localhost:9000',
+          '--root',
+          '/home/step/certs/root_ca.crt',
+        ]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+      start_period: 30s
+
 volumes:
   pg_federated_data:
   valkey_federated_data:
+  step_ca_data:

docker/gateway.Dockerfile

@@ -5,18 +5,27 @@ RUN corepack enable
 
 FROM base AS builder
 WORKDIR /app
+# Copy workspace manifests first for layer-cached install
 COPY pnpm-workspace.yaml pnpm-lock.yaml package.json ./
 COPY apps/gateway/package.json ./apps/gateway/
 COPY packages/ ./packages/
+COPY plugins/ ./plugins/
 RUN pnpm install --frozen-lockfile
 COPY . .
-RUN pnpm --filter @mosaic/gateway build
+# Build gateway and all of its workspace dependencies via turbo dependency graph
+RUN pnpm turbo run build --filter @mosaicstack/gateway...
+# Produce a self-contained deploy artifact: flat node_modules, no pnpm symlinks
+# --legacy is required for pnpm v10 when inject-workspace-packages is not set
+RUN pnpm --filter @mosaicstack/gateway --prod deploy --legacy /deploy
 
 FROM base AS runner
 WORKDIR /app
 ENV NODE_ENV=production
+# Use the pnpm deploy output — resolves all deps into a flat, self-contained node_modules
+COPY --from=builder /deploy/node_modules ./node_modules
+COPY --from=builder /deploy/package.json ./package.json
+# dist is declared in package.json "files" so pnpm deploy copies it into /deploy;
+# copy from builder explicitly as belt-and-suspenders
 COPY --from=builder /app/apps/gateway/dist ./dist
-COPY --from=builder /app/apps/gateway/package.json ./package.json
-COPY --from=builder /app/node_modules ./node_modules
 EXPOSE 4000
 CMD ["node", "dist/main.js"]

docs/federation/TASKS.md

@@ -47,11 +47,12 @@ Goal: Two federated-tier gateways stood up on Portainer at `mos-test-1.woltje.co
 > **Tracking issue:** #482.
 
 | id                    | status      | description                                                                                                                                                                                                             | issue | agent  | branch                                | depends_on   | estimate | notes                                                                                                                                              |
-| ---------------- | ----------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----- | ------ | ------------------------------------- | ------------ | -------- | -------------------------------------------------------------------------------------------------------------------------------------------------- |
+| --------------------- | ----------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----- | ------ | ------------------------------------- | ------------ | -------- | -------------------------------------------------------------------------------------------------------------------------------------------------- |
-| FED-M2-DEPLOY-01 | not-started | Verify `gateway:fed-v0.1.0-m1` image was published by `.woodpecker/publish.yml` on tag push; if not, investigate and remediate. Document image URI in deployment artifact.                                              | #482  | sonnet | feat/federation-deploy-image-verify   | —            | 2K       | publish.yml registers `gateway:$CI_COMMIT_TAG` destination; should already exist at `git.mosaicstack.dev/mosaicstack/stack/gateway:fed-v0.1.0-m1`. |
+| FED-M2-DEPLOY-01      | done        | Verify `gateway:fed-v0.1.0-m1` image was published by `.woodpecker/publish.yml` on tag push; if not, investigate and remediate. Document image URI in deployment artifact.                                              | #482  | sonnet | (verified inline, no PR)              | —            | 2K       | Tag exists; digest `sha256:9b72e202a9eecc27d31920b87b475b9e96e483c0323acc57856be4b1355db1ec` captured for digest-pinned deploys.                   |
-| FED-M2-DEPLOY-02 | not-started | Author Portainer git-stack compose file `deploy/portainer/federated-test.stack.yml` (gateway + PG-pgvector + Valkey, env-driven). Use immutable tag, not `latest`.                                                      | #482  | sonnet | feat/federation-deploy-stack-template | DEPLOY-01    | 5K       | Stack must be parameterizable via env (`STACK_DOMAIN`, `BETTERAUTH_SECRET`, etc.) so one template serves both hosts.                               |
+| FED-M2-DEPLOY-02      | done        | Author Portainer git-stack compose file `deploy/portainer/federated-test.stack.yml` (gateway + PG-pgvector + Valkey, env-driven). Use immutable tag, not `latest`.                                                      | #482  | sonnet | feat/federation-deploy-stack-template | DEPLOY-01    | 5K       | Shipped in PR #485. Digest-pinned. Env: STACK_NAME, HOST_FQDN, POSTGRES_PASSWORD, BETTER_AUTH_SECRET, BETTER_AUTH_URL.                             |
-| FED-M2-DEPLOY-03 | not-started | Deploy stack to mos-test-1.woltje.com via `~/.config/mosaic/tools/portainer/`. Verify M1 acceptance: federated-tier boot succeeds; `mosaic gateway doctor --json` returns green; pgvector `vector(3)` round-trip works. | #482  | sonnet | feat/federation-deploy-test-1         | DEPLOY-02    | 3K       | Requires `PORTAINER_URL` + `PORTAINER_API_KEY` env (vault-loaded). DNS for mos-test-1 must resolve before deploy.                                  |
+| FED-M2-DEPLOY-IMG-FIX | in-progress | Gateway image runtime broken (ERR_MODULE_NOT_FOUND for `dotenv`); Dockerfile copies `.pnpm/` store but not `apps/gateway/node_modules` symlinks. Switch to `pnpm deploy` for self-contained runtime.                    | #482  | sonnet | (subagent in flight)                  | DEPLOY-02    | 4K       | Subagent `a78a9ab0ddae91fbc` in flight. Triggers Kaniko rebuild on merge; capture new digest; bump stack template in follow-up PR before redeploy. |
-| FED-M2-DEPLOY-04 | not-started | Deploy stack to mos-test-2.woltje.com via Portainer wrapper. Same M1 acceptance probes as DEPLOY-03.                                                                                                                    | #482  | sonnet | feat/federation-deploy-test-2         | DEPLOY-02    | 3K       | Independent of DEPLOY-03 (parallelizable). Same secret material with distinct domain + secrets per host.                                           |
+| FED-M2-DEPLOY-03      | blocked     | Deploy stack to mos-test-1.woltje.com via `~/.config/mosaic/tools/portainer/`. Verify M1 acceptance: federated-tier boot succeeds; `mosaic gateway doctor --json` returns green; pgvector `vector(3)` round-trip works. | #482  | sonnet | feat/federation-deploy-test-1         | IMG-FIX      | 3K       | Stack created on Portainer endpoint 3 (Swarm `local`), but blocked on image fix. Container fails on boot until IMG-FIX merges + redeploy.          |
+| FED-M2-DEPLOY-04      | blocked     | Deploy stack to mos-test-2.woltje.com via Portainer wrapper. Same M1 acceptance probes as DEPLOY-03.                                                                                                                    | #482  | sonnet | feat/federation-deploy-test-2         | IMG-FIX      | 3K       | Same status as DEPLOY-03. Stack created; blocked on image fix.                                                                                     |
 | FED-M2-DEPLOY-05      | not-started | Document deployment in `docs/federation/TEST-INFRA.md`: hosts, image tags, secrets sourcing, redeploy procedure, teardown. Update MISSION-MANIFEST with deployment status.                                              | #482  | haiku  | feat/federation-deploy-docs           | DEPLOY-03,04 | 3K       | Operator-facing doc; mentions but does not duplicate `tools/portainer/README.md`.                                                                  |
 
 **Deploy workstream estimate:** ~16K tokens
@@ -63,11 +64,11 @@ Goal: Two federated-tier gateways stood up on Portainer at `mos-test-1.woltje.co
 Goal: An admin can create a federation grant; counterparty enrolls; cert is signed by Step-CA with SAN OIDs for `grantId` + `subjectUserId`. No runtime federation traffic flows yet (that's M3).
 
 | id        | status      | description                                                                                                                                                                                      | issue | agent  | branch                             | depends_on       | estimate | notes                                                                                                                                                                                                                                                                                                                                                                                                 |
-| --------- | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----- | ------ | ---------------------------------- | ---------------- | -------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
+| --------- | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----- | ------ | ---------------------------------- | ---------------- | -------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| FED-M2-01 | not-started | DB migration: `federation_grants`, `federation_peers`, `federation_audit_log` tables + enum types (`grant_status`, `peer_state`). Drizzle schema + migration generation; migration tests.        | #461  | sonnet | feat/federation-m2-schema          | —                | 5K       | `federation_audit_log` is created but not yet written to (audit logic is M4). Reserve `query_hash`, `outcome`, `bytes_out` columns.       |
+| FED-M2-01 | needs-qa    | DB migration: `federation_grants`, `federation_peers`, `federation_audit_log` tables + enum types (`grant_status`, `peer_state`). Drizzle schema + migration generation; migration tests.        | #461  | sonnet | feat/federation-m2-schema          | —                | 5K       | PR #486 open. First review NEEDS CHANGES (missing DESC indexes + reserved cols). Remediation subagent `a673dd9355dc26f82` in flight in worktree `agent-a4404ac1`.                                                                                                                                                                                                                                     |
 | FED-M2-02 | not-started | Add Step-CA sidecar to `docker-compose.federated.yml`: official `smallstep/step-ca` image, persistent CA volume, JWK provisioner config baked into init script.                                  | #461  | sonnet | feat/federation-m2-stepca          | DEPLOY-02        | 4K       | Profile-gated under `federated`. CA password from secret; dev compose uses dev-only password file.                                                                                                                                                                                                                                                                                                    |
 | FED-M2-03 | not-started | Scope JSON schema + validator: `resources` allowlist, `excluded_resources`, `include_teams`, `include_personal`, `max_rows_per_query`. Vitest unit tests for valid + invalid scopes.             | #461  | sonnet | feat/federation-m2-scope-schema    | —                | 4K       | Validator independent of CA — reusable from grant CRUD + (later) M3 scope enforcement.                                                                                                                                                                                                                                                                                                                |
-| FED-M2-04 | not-started | `apps/gateway/src/federation/ca.service.ts`: Step-CA client (CSR submission, OID-bearing cert retrieval). Mocked + integration tests against real Step-CA container.                             | #461  | sonnet | feat/federation-m2-ca-service      | M2-02            | 6K       | SAN OIDs: `grantId` (custom OID 1.3.6.1.4.1.99999.1) + `subjectUserId` (1.3.6.1.4.1.99999.2). Document OID assignments in PRD/SETUP.      |
+| FED-M2-04 | not-started | `apps/gateway/src/federation/ca.service.ts`: Step-CA client (CSR submission, OID-bearing cert retrieval). Mocked + integration tests against real Step-CA container.                             | #461  | sonnet | feat/federation-m2-ca-service      | M2-02            | 6K       | SAN OIDs: `grantId` (custom OID 1.3.6.1.4.1.99999.1) + `subjectUserId` (1.3.6.1.4.1.99999.2). Document OID assignments in PRD/SETUP. **Acceptance**: must (a) wire `federation.tpl` template into `mosaic-fed` provisioner config and (b) include a unit/integration test asserting issued certs contain BOTH OIDs — fails-loud guard against silent OID stripping (carry-forward from M2-02 review). |
 | FED-M2-05 | not-started | Sealed storage for `client_key_pem` reusing existing `provider_credentials` sealing key. Tests prove DB-at-rest is ciphertext, not PEM. Key rotation path documented (deferred impl).            | #461  | sonnet | feat/federation-m2-key-sealing     | M2-01            | 5K       | Separate from M2-06 to keep crypto seam isolated; reviewer focus is sealing only.                                                                                                                                                                                                                                                                                                                     |
 | FED-M2-06 | not-started | `grants.service.ts`: CRUD + status transitions (`pending` → `active` → `revoked`); integrates M2-03 (scope) + M2-05 (sealing). Unit tests cover all transitions including invalid ones.          | #461  | sonnet | feat/federation-m2-grants-service  | M2-03, M2-05     | 6K       | Business logic only — CSR + cert work delegated to M2-04. Revocation handler is M6.                                                                                                                                                                                                                                                                                                                   |
 | FED-M2-07 | not-started | `enrollment.controller.ts`: short-lived single-use token endpoint; CSR signing; updates grant `pending` → `active`; emits enrollment audit (table-only write, M4 tightens).                      | #461  | sonnet | feat/federation-m2-enrollment      | M2-04, M2-06     | 6K       | Tokens single-use with 410 on replay; tokens TTL'd at 15min; rate-limited at request layer (M4 introduces guard, M2 uses simple lock).                                                                                                                                                                                                                                                                |

docs/scratchpads/mvp-20260312.md

@@ -523,3 +523,92 @@ Independent security review surfaced three high-impact and four medium findings;
 - #8: confirm `packages/config/dist` not git-tracked
 
 **Next mission step:** FED-M2 (Step-CA + grant schema + admin CLI). Per TASKS.md scope rule, M2 will be decomposed when it enters active planning. Issue #461 tracks scope.
+
+## Session 20 — 2026-04-21 — FED-M2 kickoff
+
+### Decisions
+
+- **Workstream split**: parallel CODE (M2-01..M2-13, ~72K) + DEPLOY (DEPLOY-01..DEPLOY-05, ~16K) tracks; re-converge at M2-10 E2E.
+- **Test hosts**: `mos-test-1.woltje.com` (querying side / Server A), `mos-test-2.woltje.com` (serving side / Server B). Wildcard `*.woltje.com` A→174.137.97.162 already exists; Traefik wildcard cert covers both subdomains. No DNS or cert work needed pre-deploy.
+- **Portainer access**: requires `PORTAINER_INSECURE=1` flag added to mosaic wrappers (self-signed cert at `https://10.1.1.43:9443`). PR pending on `feat/mosaic-portainer-tls-flag`.
+- **Image policy**: deploy by digest (immutable) per Mosaic policy. `gateway:fed-v0.1.0-m1` digest = `sha256:9b72e202a9eecc27d31920b87b475b9e96e483c0323acc57856be4b1355db1ec`.
+
+### DEPLOY-01 — image manifest verified
+
+- Tag `fed-v0.1.0-m1` exists at `git.mosaicstack.dev/mosaicstack/stack/gateway`
+- Digest: `sha256:9b72e202a9eecc27d31920b87b475b9e96e483c0323acc57856be4b1355db1ec`
+- 9 layers, ~530MB total
+- Use this digest in DEPLOY-02 stack template (do NOT reference `:fed-v0.1.0-m1` tag in stack — pin to digest)
+
+### Registry auth note
+
+- Gitea container registry uses Bearer token flow (`/v2/token?service=container_registry&scope=repository:<repo>:pull`)
+- Username: `jarvis` (NOT `mosaicstack`); password: `gitea.mosaicstack.token` from credentials.json
+- Direct `Authorization: Bearer <pat>` does NOT work — must exchange PAT for registry token first
+
+### Active PRs
+
+- #483 — docs: M2 mission planning (TASKS decomposition + manifest update) — CI running
+- (pending) `feat/mosaic-portainer-tls-flag` — wrapper PORTAINER_INSECURE flag (sonnet subagent in progress)
+- (pending) `feat/federation-m2-schema` — FED-M2-01 DB schema migration (sonnet subagent in progress)
+
+### MISSION-MANIFEST layout fix
+
+- Initial M2 commit had Test Infrastructure block inserted by lint-staged prettier between "Last Updated" and "Parent Mission" — split mission frontmatter
+- Fixed in 3d001fdb: moved Parent Mission back to frontmatter, kept Test Infrastructure as standalone H2 between Mission and Context
+
+## Session 21 — 2026-04-21/22 — DEPLOY-02 merged, gateway image bug discovered, M2-01 in remediation
+
+### PRs merged
+
+- **#483** — docs(federation): M2 mission planning (TASKS decomposition + manifest update)
+- **#484** — feat(mosaic-portainer): PORTAINER_INSECURE flag for self-signed TLS (wrapper sync to `~/.config/mosaic/tools/portainer/` done manually due to broken `mosaic upgrade` `set -o pipefail` on dash)
+- **#485** — feat(deploy): portainer stack template `deploy/portainer/federated-test.stack.yml` for federation test instances [DEPLOY-02]
+
+### Stack deployed (mos-test-1, mos-test-2)
+
+- Both stacks created on Portainer endpoint 3 (`local` Swarm @ 10.1.1.43, the only endpoint with traefik-public + woltje.com wildcard cert)
+- Swarm ID `l7z67tfpd4bvj4979ufpkyi50`
+- Image pinned to digest `sha256:9b72e202a9eecc27d31920b87b475b9e96e483c0323acc57856be4b1355db1ec`
+- Traefik labels target `${HOST_FQDN}` per env
+
+### CRITICAL FINDING — gateway image runtime-broken
+
+- `docker run` against `gateway:fed-v0.1.0-m1` fails immediately:  
+  `Error [ERR_MODULE_NOT_FOUND]: Cannot find package 'dotenv' imported from /app/dist/main.js`
+- Root cause: `docker/gateway.Dockerfile` copies `/app/node_modules` from builder — but pnpm puts deps in the content-addressed `.pnpm/` store with symlinks at `apps/gateway/node_modules/*`. The runner stage misses the symlinks → Node can't resolve workspace deps.
+- M1 release was never runtime-tested as a stripped container; CI passed because tests run in dev tree where pnpm symlinks are intact.
+- **Fix in flight** (subagent `a78a9ab0ddae91fbc`): switch builder to `pnpm --filter @mosaic/gateway --prod deploy /deploy`, then runner copies `/deploy/node_modules` + `/deploy/dist` + `/deploy/package.json`.
+
+### M2-01 schema review verdict — NEEDS CHANGES
+
+- PR #486 (`feat/federation-m2-schema`) — independent reviewer (sonnet) found 2 real issues:
+  1. `federation_audit_log` time-range indexes missing `.desc()` on `created_at` (3 places)
+  2. Reserved columns missing per TASKS.md M2-01 spec: `query_hash`, `outcome`, `bytes_out` (M4 will write; spec said reserve now)
+- Also notes (advisory): subject_user_id correctly `text` (matches BetterAuth users.id; spec defect, not code defect); peer→grant cascade test not present (would be trivial to add)
+- **Remediation in flight** (subagent `a673dd9355dc26f82` in worktree `agent-a4404ac1`): apply DESC + reserved cols, regenerate migration in place (preferred) or stack 0009 (fallback), force-push, post PR comment.
+
+### Process notes
+
+- Branch race incident: schema subagent + wrapper subagent both ran in main checkout → schema files appeared on wrapper branch. Recovered by TaskStop, `git checkout --` to clean, respawned schema subagent with `isolation: "worktree"`. **Rule going forward:** any subagent doing code edits gets `isolation: "worktree"` unless work is single-file and the orchestrator confirms no other branch will touch overlapping files.
+- `pr-create.sh` shell-quotes backticks badly → use `tea pr create --repo mosaicstack/stack` directly (matches CLI-skill behavior). Will leave a followup to harden pr-create.sh.
+- Gitea registry auth: bearer-token exchange flow (`/v2/token?service=container_registry&scope=repository:<repo>:pull`) — direct `Authorization: Bearer <pat>` returns 401.
+- Portainer Swarm stack create endpoint: `POST /api/stacks/create/swarm/string?endpointId=<id>` (NOT `/api/stacks?type=1` — deprecated and rejected with 400).
+
+### In-flight at compaction boundary
+
+- Subagent `a78a9ab0ddae91fbc` — Dockerfile pnpm-deploy fix → PR (not yet opened at handoff)
+- Subagent `a673dd9355dc26f82` — M2-01 schema remediation (DESC + reserved cols) → force-push to PR #486
+- Both will trigger CI; orchestrator must independently re-review fixes (especially the security-adjacent schema work) per "always verify subagent claims" rule.
+
+### Next after subagents return
+
+1. Independent re-review of schema remediation (different subagent, fresh context)
+2. Merge #486 if green
+3. Merge Dockerfile fix PR if green → triggers Kaniko CI rebuild → capture new digest
+4. Update `deploy/portainer/federated-test.stack.yml` to new digest in a small PR
+5. Redeploy mos-test-1 + mos-test-2 (Portainer stack update via API)
+6. Verify HTTPS reachability + `/health` endpoint at both hosts
+7. DEPLOY-03/04 acceptance probes (`mosaic gateway doctor --json`, pgvector `vector(3)` round-trip)
+8. DEPLOY-05: author `docs/federation/TEST-INFRA.md`
+9. M2-02 (Step-CA sidecar) kicks off after image health is green

infra/step-ca/dev-password.example

@@ -0,0 +1 @@
+dev-only-step-ca-password-do-not-use-in-production

infra/step-ca/init.sh

@@ -0,0 +1,60 @@
+#!/bin/sh
+# infra/step-ca/init.sh
+#
+# Idempotent first-boot initialiser for the Mosaic Federation CA.
+#
+# On the first run (no /home/step/config/ca.json present) this script:
+#   1. Initialises Step-CA with a JWK provisioner named "mosaic-fed".
+#   2. Writes the CA configuration to the persistent volume at /home/step.
+#
+# On subsequent runs (config already exists) this script skips init and
+# starts the CA directly.
+#
+# The provisioner name "mosaic-fed" is consumed by:
+#   apps/gateway/src/federation/ca.service.ts  (added in M2-04)
+#
+# Password source:
+#   Dev:  mounted from ./infra/step-ca/dev-password via bind mount.
+#   Prod: mounted from a Docker secret at /run/secrets/ca_password.
+#
+# OID template:
+#   infra/step-ca/templates/federation.tpl is copied into the CA config
+#   directory so the JWK provisioner can reference it.  The template
+#   skeleton is wired in M2-04 when the CA service lands the SAN-bearing
+#   CSR work.
+
+set -e
+
+CA_CONFIG="/home/step/config/ca.json"
+PASSWORD_FILE="/run/secrets/ca_password"
+
+if [ ! -f "${CA_CONFIG}" ]; then
+  echo "[step-ca init] First boot detected — initialising Mosaic Federation CA..."
+
+  step ca init \
+    --name "Mosaic Federation CA" \
+    --dns "localhost" \
+    --dns "step-ca" \
+    --address ":9000" \
+    --provisioner "mosaic-fed" \
+    --password-file "${PASSWORD_FILE}" \
+    --provisioner-password-file "${PASSWORD_FILE}" \
+    --no-db
+
+  echo "[step-ca init] CA initialised."
+
+  # Copy the X.509 template into the Step-CA config directory so the
+  # provisioner can reference it in M2-04.
+  if [ -f "/etc/step-ca-templates/federation.tpl" ]; then
+    mkdir -p /home/step/templates
+    cp /etc/step-ca-templates/federation.tpl /home/step/templates/federation.tpl
+    echo "[step-ca init] Federation X.509 template copied to /home/step/templates/."
+  fi
+
+  echo "[step-ca init] Startup complete."
+else
+  echo "[step-ca init] Config already exists — skipping init."
+fi
+
+echo "[step-ca init] Starting Step-CA on :9000..."
+exec step-ca /home/step/config/ca.json --password-file "${PASSWORD_FILE}"

infra/step-ca/templates/federation.tpl

@@ -0,0 +1,48 @@
+{
+  "subject": {{ toJson .Subject }},
+  "sans": {{ toJson .SANs }},
+
+  {{- /*
+    Mosaic Federation X.509 Certificate Template
+    ============================================
+    This template is used by the "mosaic-fed" JWK provisioner to sign
+    federation client certificates.
+
+    Custom OID extensions (per PRD §6):
+      1.3.6.1.4.1.99999.1  — mosaic.federation.grantId   (UUID string)
+      1.3.6.1.4.1.99999.2  — mosaic.federation.subjectUserId (UUID string)
+
+    TODO (M2-04): Wire actual OID extensions below once the CA service
+    (apps/gateway/src/federation/ca.service.ts) lands the SAN-bearing CSR
+    work and the template can be exercised end-to-end.
+
+    Step-CA template reference:
+      https://smallstep.com/docs/step-ca/templates
+
+    Expected final shape of the extensions block (placeholder — not yet
+    activated):
+
+    "extensions": [
+      {
+        "id": "1.3.6.1.4.1.99999.1",
+        "critical": false,
+        "value": {{ toJson (first .Token.mosaic_grant_id) }}
+      },
+      {
+        "id": "1.3.6.1.4.1.99999.2",
+        "critical": false,
+        "value": {{ toJson (first .Token.mosaic_subject_user_id) }}
+      }
+    ],
+
+    The provisioner must pass these values in the ACME/JWK token payload
+    (token claims `mosaic_grant_id` and `mosaic_subject_user_id`) when
+    submitting the CSR.  M2-04 owns that work.
+  */ -}}
+
+  "keyUsage": ["digitalSignature"],
+  "extKeyUsage": ["clientAuth"],
+  "basicConstraints": {
+    "isCA": false
+  }
+}

packages/db/drizzle/0008_smart_lyja.sql

@@ -0,0 +1,75 @@
+CREATE TYPE "public"."grant_status" AS ENUM('active', 'revoked', 'expired');--> statement-breakpoint
+CREATE TYPE "public"."peer_state" AS ENUM('pending', 'active', 'suspended', 'revoked');--> statement-breakpoint
+CREATE TABLE "admin_tokens" (
+	"id" text PRIMARY KEY NOT NULL,
+	"user_id" text NOT NULL,
+	"token_hash" text NOT NULL,
+	"label" text NOT NULL,
+	"scope" text DEFAULT 'admin' NOT NULL,
+	"expires_at" timestamp with time zone,
+	"last_used_at" timestamp with time zone,
+	"created_at" timestamp with time zone DEFAULT now() NOT NULL
+);
+--> statement-breakpoint
+CREATE TABLE "federation_audit_log" (
+	"id" uuid PRIMARY KEY DEFAULT gen_random_uuid() NOT NULL,
+	"request_id" text NOT NULL,
+	"peer_id" uuid,
+	"subject_user_id" text,
+	"grant_id" uuid,
+	"verb" text NOT NULL,
+	"resource" text NOT NULL,
+	"status_code" integer NOT NULL,
+	"result_count" integer,
+	"denied_reason" text,
+	"latency_ms" integer,
+	"created_at" timestamp with time zone DEFAULT now() NOT NULL,
+	"query_hash" text,
+	"outcome" text,
+	"bytes_out" integer
+);
+--> statement-breakpoint
+CREATE TABLE "federation_grants" (
+	"id" uuid PRIMARY KEY DEFAULT gen_random_uuid() NOT NULL,
+	"subject_user_id" text NOT NULL,
+	"peer_id" uuid NOT NULL,
+	"scope" jsonb NOT NULL,
+	"status" "grant_status" DEFAULT 'active' NOT NULL,
+	"expires_at" timestamp with time zone,
+	"created_at" timestamp with time zone DEFAULT now() NOT NULL,
+	"revoked_at" timestamp with time zone,
+	"revoked_reason" text
+);
+--> statement-breakpoint
+CREATE TABLE "federation_peers" (
+	"id" uuid PRIMARY KEY DEFAULT gen_random_uuid() NOT NULL,
+	"common_name" text NOT NULL,
+	"display_name" text NOT NULL,
+	"cert_pem" text NOT NULL,
+	"cert_serial" text NOT NULL,
+	"cert_not_after" timestamp with time zone NOT NULL,
+	"client_key_pem" text,
+	"state" "peer_state" DEFAULT 'pending' NOT NULL,
+	"endpoint_url" text,
+	"last_seen_at" timestamp with time zone,
+	"created_at" timestamp with time zone DEFAULT now() NOT NULL,
+	"revoked_at" timestamp with time zone,
+	CONSTRAINT "federation_peers_common_name_unique" UNIQUE("common_name"),
+	CONSTRAINT "federation_peers_cert_serial_unique" UNIQUE("cert_serial")
+);
+--> statement-breakpoint
+ALTER TABLE "admin_tokens" ADD CONSTRAINT "admin_tokens_user_id_users_id_fk" FOREIGN KEY ("user_id") REFERENCES "public"."users"("id") ON DELETE cascade ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "federation_audit_log" ADD CONSTRAINT "federation_audit_log_peer_id_federation_peers_id_fk" FOREIGN KEY ("peer_id") REFERENCES "public"."federation_peers"("id") ON DELETE set null ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "federation_audit_log" ADD CONSTRAINT "federation_audit_log_subject_user_id_users_id_fk" FOREIGN KEY ("subject_user_id") REFERENCES "public"."users"("id") ON DELETE set null ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "federation_audit_log" ADD CONSTRAINT "federation_audit_log_grant_id_federation_grants_id_fk" FOREIGN KEY ("grant_id") REFERENCES "public"."federation_grants"("id") ON DELETE set null ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "federation_grants" ADD CONSTRAINT "federation_grants_subject_user_id_users_id_fk" FOREIGN KEY ("subject_user_id") REFERENCES "public"."users"("id") ON DELETE cascade ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "federation_grants" ADD CONSTRAINT "federation_grants_peer_id_federation_peers_id_fk" FOREIGN KEY ("peer_id") REFERENCES "public"."federation_peers"("id") ON DELETE cascade ON UPDATE no action;--> statement-breakpoint
+CREATE INDEX "admin_tokens_user_id_idx" ON "admin_tokens" USING btree ("user_id");--> statement-breakpoint
+CREATE UNIQUE INDEX "admin_tokens_hash_idx" ON "admin_tokens" USING btree ("token_hash");--> statement-breakpoint
+CREATE INDEX "federation_audit_log_peer_created_at_idx" ON "federation_audit_log" USING btree ("peer_id","created_at" DESC NULLS LAST);--> statement-breakpoint
+CREATE INDEX "federation_audit_log_subject_created_at_idx" ON "federation_audit_log" USING btree ("subject_user_id","created_at" DESC NULLS LAST);--> statement-breakpoint
+CREATE INDEX "federation_audit_log_created_at_idx" ON "federation_audit_log" USING btree ("created_at" DESC NULLS LAST);--> statement-breakpoint
+CREATE INDEX "federation_grants_subject_status_idx" ON "federation_grants" USING btree ("subject_user_id","status");--> statement-breakpoint
+CREATE INDEX "federation_grants_peer_status_idx" ON "federation_grants" USING btree ("peer_id","status");--> statement-breakpoint
+CREATE INDEX "federation_peers_cert_serial_idx" ON "federation_peers" USING btree ("cert_serial");--> statement-breakpoint
+CREATE INDEX "federation_peers_state_idx" ON "federation_peers" USING btree ("state");

packages/db/drizzle/meta/_journal.json

@@ -57,6 +57,13 @@
       "when": 1774227064500,
       "tag": "0006_swift_shen",
       "breakpoints": true
+    },
+    {
+      "idx": 8,
+      "version": "7",
+      "when": 1776822435828,
+      "tag": "0008_smart_lyja",
+      "breakpoints": true
     }
   ]
 }

packages/db/src/federation.integration.test.ts

@@ -0,0 +1,424 @@
+/**
+ * FED-M2-01 — Integration test: federation DB schema (peers / grants / audit_log).
+ *
+ * Prereq: docker compose -f docker-compose.federated.yml --profile federated up -d
+ *         (or any postgres with the mosaic schema already applied)
+ * Run:    FEDERATED_INTEGRATION=1 pnpm --filter @mosaicstack/db test src/federation.integration.test.ts
+ *
+ * Skipped when FEDERATED_INTEGRATION !== '1'.
+ *
+ * Strategy:
+ *  - Applies the federation migration SQL directly (idempotent: CREATE TYPE/TABLE
+ *    with IF NOT EXISTS guards applied via inline SQL before the migration DDL).
+ *  - Assumes the base schema (users table etc.) already exists in the target DB.
+ *  - All test rows use the `fed-m2-01-` prefix; cleanup in afterAll.
+ *
+ * Coverage:
+ *  1. Federation tables + enums apply cleanly against the existing schema.
+ *  2. Insert a sample user + peer + grant + audit row; verify round-trip.
+ *  3. FK cascade: deleting the user cascades to federation_grants.
+ *  4. FK set-null: deleting the peer sets federation_audit_log.peer_id to NULL.
+ *  5. Enum constraint: inserting an invalid status/state value throws a DB error.
+ *  6. Unique constraint: duplicate cert_serial throws a DB error.
+ */
+
+import postgres from 'postgres';
+import { afterAll, beforeAll, describe, expect, it } from 'vitest';
+
+const run = process.env['FEDERATED_INTEGRATION'] === '1';
+
+const PG_URL = process.env['DATABASE_URL'] ?? 'postgresql://mosaic:mosaic@localhost:5433/mosaic';
+
+/** Recognisable test-row prefix for safe cleanup without full-table truncation. */
+const T = 'fed-m2-01';
+
+// Deterministic IDs (UUID format required for uuid PK columns: 8-4-4-4-12 hex digits).
+const PEER1_ID = `f2000001-0000-4000-8000-000000000001`;
+const PEER2_ID = `f2000002-0000-4000-8000-000000000002`;
+const USER1_ID = `${T}-user-1`;
+
+let sql: ReturnType<typeof postgres> | undefined;
+
+beforeAll(async () => {
+  if (!run) return;
+  sql = postgres(PG_URL, { max: 1, connect_timeout: 10, idle_timeout: 10 });
+
+  // Apply the federation enums and tables idempotently.
+  // This mirrors the migration file but uses IF NOT EXISTS guards so it can run
+  // against a DB that may not have had drizzle migrations tracked.
+  await sql`
+    DO $$ BEGIN
+      CREATE TYPE peer_state AS ENUM ('pending', 'active', 'suspended', 'revoked');
+    EXCEPTION WHEN duplicate_object THEN NULL;
+    END $$
+  `;
+  await sql`
+    DO $$ BEGIN
+      CREATE TYPE grant_status AS ENUM ('active', 'revoked', 'expired');
+    EXCEPTION WHEN duplicate_object THEN NULL;
+    END $$
+  `;
+  await sql`
+    CREATE TABLE IF NOT EXISTS federation_peers (
+      id              uuid PRIMARY KEY DEFAULT gen_random_uuid(),
+      common_name     text NOT NULL,
+      display_name    text NOT NULL,
+      cert_pem        text NOT NULL,
+      cert_serial     text NOT NULL,
+      cert_not_after  timestamp with time zone NOT NULL,
+      client_key_pem  text,
+      state           peer_state NOT NULL DEFAULT 'pending',
+      endpoint_url    text,
+      last_seen_at    timestamp with time zone,
+      created_at      timestamp with time zone NOT NULL DEFAULT now(),
+      revoked_at      timestamp with time zone,
+      CONSTRAINT federation_peers_common_name_unique UNIQUE (common_name),
+      CONSTRAINT federation_peers_cert_serial_unique UNIQUE (cert_serial)
+    )
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_peers_cert_serial_idx ON federation_peers (cert_serial)
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_peers_state_idx ON federation_peers (state)
+  `;
+  await sql`
+    CREATE TABLE IF NOT EXISTS federation_grants (
+      id               uuid PRIMARY KEY DEFAULT gen_random_uuid(),
+      subject_user_id  text NOT NULL REFERENCES users(id) ON DELETE CASCADE,
+      peer_id          uuid NOT NULL REFERENCES federation_peers(id) ON DELETE CASCADE,
+      scope            jsonb NOT NULL,
+      status           grant_status NOT NULL DEFAULT 'active',
+      expires_at       timestamp with time zone,
+      created_at       timestamp with time zone NOT NULL DEFAULT now(),
+      revoked_at       timestamp with time zone,
+      revoked_reason   text
+    )
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_grants_subject_status_idx ON federation_grants (subject_user_id, status)
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_grants_peer_status_idx ON federation_grants (peer_id, status)
+  `;
+  await sql`
+    CREATE TABLE IF NOT EXISTS federation_audit_log (
+      id               uuid PRIMARY KEY DEFAULT gen_random_uuid(),
+      request_id       text NOT NULL,
+      peer_id          uuid REFERENCES federation_peers(id) ON DELETE SET NULL,
+      subject_user_id  text REFERENCES users(id) ON DELETE SET NULL,
+      grant_id         uuid REFERENCES federation_grants(id) ON DELETE SET NULL,
+      verb             text NOT NULL,
+      resource         text NOT NULL,
+      status_code      integer NOT NULL,
+      result_count     integer,
+      denied_reason    text,
+      latency_ms       integer,
+      created_at       timestamp with time zone NOT NULL DEFAULT now(),
+      query_hash       text,
+      outcome          text,
+      bytes_out        integer
+    )
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_audit_log_peer_created_at_idx
+      ON federation_audit_log (peer_id, created_at DESC NULLS LAST)
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_audit_log_subject_created_at_idx
+      ON federation_audit_log (subject_user_id, created_at DESC NULLS LAST)
+  `;
+  await sql`
+    CREATE INDEX IF NOT EXISTS federation_audit_log_created_at_idx
+      ON federation_audit_log (created_at DESC NULLS LAST)
+  `;
+});
+
+afterAll(async () => {
+  if (!sql) return;
+
+  // Cleanup in FK-safe order (children before parents).
+  await sql`DELETE FROM federation_audit_log WHERE request_id LIKE ${T + '%'}`.catch(() => {});
+  await sql`
+    DELETE FROM federation_grants
+    WHERE subject_user_id LIKE ${T + '%'}
+       OR revoked_reason LIKE ${T + '%'}
+  `.catch(() => {});
+  await sql`DELETE FROM federation_peers WHERE common_name LIKE ${T + '%'}`.catch(() => {});
+  await sql`DELETE FROM users WHERE id LIKE ${T + '%'}`.catch(() => {});
+  await sql.end({ timeout: 3 }).catch(() => {});
+});
+
+describe.skipIf(!run)('federation schema — integration', () => {
+  // ── 1. Insert sample rows ──────────────────────────────────────────────────
+
+  it('inserts a user, peer, grant, and audit row without constraint violation', async () => {
+    const certPem = '-----BEGIN CERTIFICATE-----\nMIItest\n-----END CERTIFICATE-----';
+
+    // User — BetterAuth users.id is text (any string, not uuid).
+    await sql!`
+      INSERT INTO users (id, name, email, email_verified, created_at, updated_at)
+      VALUES (${USER1_ID}, ${'M2-01 Test User'}, ${USER1_ID + '@example.com'}, false, now(), now())
+      ON CONFLICT (id) DO NOTHING
+    `;
+
+    // Peer
+    await sql!`
+      INSERT INTO federation_peers
+        (id, common_name, display_name, cert_pem, cert_serial, cert_not_after, state, created_at)
+      VALUES (
+        ${PEER1_ID},
+        ${T + '-gateway-example-com'},
+        ${'Test Peer'},
+        ${certPem},
+        ${T + '-serial-001'},
+        now() + interval '1 year',
+        ${'active'},
+        now()
+      )
+      ON CONFLICT (id) DO NOTHING
+    `;
+
+    // Grant — scope is jsonb; pass as JSON string and cast server-side.
+    const scopeJson = JSON.stringify({
+      resources: ['tasks', 'notes'],
+      operations: ['list', 'get'],
+    });
+    const grants = await sql!`
+      INSERT INTO federation_grants
+        (subject_user_id, peer_id, scope, status, created_at)
+      VALUES (
+        ${USER1_ID},
+        ${PEER1_ID},
+        ${scopeJson}::jsonb,
+        ${'active'},
+        now()
+      )
+      RETURNING id
+    `;
+    expect(grants).toHaveLength(1);
+    const grantId = grants[0]!['id'] as string;
+
+    // Audit log row
+    await sql!`
+      INSERT INTO federation_audit_log
+        (request_id, peer_id, subject_user_id, grant_id, verb, resource, status_code, created_at)
+      VALUES (
+        ${T + '-req-001'},
+        ${PEER1_ID},
+        ${USER1_ID},
+        ${grantId},
+        ${'list'},
+        ${'tasks'},
+        ${200},
+        now()
+      )
+    `;
+
+    // Verify the audit row is present and has correct data.
+    const auditRows = await sql!`
+      SELECT * FROM federation_audit_log WHERE request_id = ${T + '-req-001'}
+    `;
+    expect(auditRows).toHaveLength(1);
+    expect(auditRows[0]!['status_code']).toBe(200);
+    expect(auditRows[0]!['verb']).toBe('list');
+    expect(auditRows[0]!['resource']).toBe('tasks');
+  }, 30_000);
+
+  // ── 2. FK cascade: user delete cascades grants ─────────────────────────────
+
+  it('cascade-deletes federation_grants when the subject user is deleted', async () => {
+    const cascadeUserId = `${T}-cascade-user`;
+    await sql!`
+      INSERT INTO users (id, name, email, email_verified, created_at, updated_at)
+      VALUES (${cascadeUserId}, ${'Cascade User'}, ${cascadeUserId + '@example.com'}, false, now(), now())
+      ON CONFLICT (id) DO NOTHING
+    `;
+    const scopeJson = JSON.stringify({ resources: ['tasks'] });
+    await sql!`
+      INSERT INTO federation_grants
+        (subject_user_id, peer_id, scope, status, revoked_reason, created_at)
+      VALUES (
+        ${cascadeUserId},
+        ${PEER1_ID},
+        ${scopeJson}::jsonb,
+        ${'active'},
+        ${T + '-cascade-test'},
+        now()
+      )
+    `;
+
+    const before = await sql!`
+      SELECT count(*)::int AS cnt FROM federation_grants WHERE subject_user_id = ${cascadeUserId}
+    `;
+    expect(before[0]!['cnt']).toBe(1);
+
+    // Delete user → grants should cascade-delete.
+    await sql!`DELETE FROM users WHERE id = ${cascadeUserId}`;
+
+    const after = await sql!`
+      SELECT count(*)::int AS cnt FROM federation_grants WHERE subject_user_id = ${cascadeUserId}
+    `;
+    expect(after[0]!['cnt']).toBe(0);
+  }, 15_000);
+
+  // ── 3. FK set-null: peer delete sets audit_log.peer_id to NULL ────────────
+
+  it('sets federation_audit_log.peer_id to NULL when the peer is deleted', async () => {
+    // Insert a throwaway peer for this specific cascade test.
+    await sql!`
+      INSERT INTO federation_peers
+        (id, common_name, display_name, cert_pem, cert_serial, cert_not_after, state, created_at)
+      VALUES (
+        ${PEER2_ID},
+        ${T + '-gateway-throwaway-com'},
+        ${'Throwaway Peer'},
+        ${'cert-pem-placeholder'},
+        ${T + '-serial-002'},
+        now() + interval '1 year',
+        ${'active'},
+        now()
+      )
+      ON CONFLICT (id) DO NOTHING
+    `;
+    const reqId = `${T}-req-setnull`;
+    await sql!`
+      INSERT INTO federation_audit_log
+        (request_id, peer_id, subject_user_id, verb, resource, status_code, created_at)
+      VALUES (
+        ${reqId},
+        ${PEER2_ID},
+        ${USER1_ID},
+        ${'get'},
+        ${'tasks'},
+        ${200},
+        now()
+      )
+    `;
+
+    await sql!`DELETE FROM federation_peers WHERE id = ${PEER2_ID}`;
+
+    const rows = await sql!`
+      SELECT peer_id FROM federation_audit_log WHERE request_id = ${reqId}
+    `;
+    expect(rows).toHaveLength(1);
+    expect(rows[0]!['peer_id']).toBeNull();
+  }, 15_000);
+
+  // ── 4. Enum constraint: invalid grant_status rejected ─────────────────────
+
+  it('rejects an invalid grant_status value with a DB error', async () => {
+    const scopeJson = JSON.stringify({ resources: ['tasks'] });
+    await expect(
+      sql!`
+        INSERT INTO federation_grants
+          (subject_user_id, peer_id, scope, status, created_at)
+        VALUES (
+          ${USER1_ID},
+          ${PEER1_ID},
+          ${scopeJson}::jsonb,
+          ${'invalid_status'},
+          now()
+        )
+      `,
+    ).rejects.toThrow();
+  }, 10_000);
+
+  // ── 5. Enum constraint: invalid peer_state rejected ───────────────────────
+
+  it('rejects an invalid peer_state value with a DB error', async () => {
+    await expect(
+      sql!`
+        INSERT INTO federation_peers
+          (common_name, display_name, cert_pem, cert_serial, cert_not_after, state, created_at)
+        VALUES (
+          ${'bad-state-peer'},
+          ${'Bad State'},
+          ${'pem'},
+          ${'bad-serial-999'},
+          now() + interval '1 year',
+          ${'invalid_state'},
+          now()
+        )
+      `,
+    ).rejects.toThrow();
+  }, 10_000);
+
+  // ── 6. Unique constraint: duplicate cert_serial rejected ──────────────────
+
+  it('rejects a duplicate cert_serial with a unique constraint violation', async () => {
+    await expect(
+      sql!`
+        INSERT INTO federation_peers
+          (common_name, display_name, cert_pem, cert_serial, cert_not_after, state, created_at)
+        VALUES (
+          ${T + '-dup-cn'},
+          ${'Dup Peer'},
+          ${'pem'},
+          ${T + '-serial-001'},
+          now() + interval '1 year',
+          ${'pending'},
+          now()
+        )
+      `,
+    ).rejects.toThrow();
+  }, 10_000);
+
+  // ── 7. FK cascade: peer delete cascades to federation_grants ─────────────
+
+  it('cascade-deletes federation_grants when the owning peer is deleted', async () => {
+    const PEER3_ID = `f2000003-0000-4000-8000-000000000003`;
+    const cascadeGrantUserId = `${T}-cascade-grant-user`;
+
+    // Insert a dedicated user and peer for this test.
+    await sql!`
+      INSERT INTO users (id, name, email, email_verified, created_at, updated_at)
+      VALUES (${cascadeGrantUserId}, ${'Cascade Grant User'}, ${cascadeGrantUserId + '@example.com'}, false, now(), now())
+      ON CONFLICT (id) DO NOTHING
+    `;
+    await sql!`
+      INSERT INTO federation_peers
+        (id, common_name, display_name, cert_pem, cert_serial, cert_not_after, state, created_at)
+      VALUES (
+        ${PEER3_ID},
+        ${T + '-gateway-cascade-peer'},
+        ${'Cascade Peer'},
+        ${'cert-pem-cascade'},
+        ${T + '-serial-003'},
+        now() + interval '1 year',
+        ${'active'},
+        now()
+      )
+      ON CONFLICT (id) DO NOTHING
+    `;
+
+    const scopeJson = JSON.stringify({ resources: ['tasks'] });
+    await sql!`
+      INSERT INTO federation_grants
+        (subject_user_id, peer_id, scope, status, created_at)
+      VALUES (
+        ${cascadeGrantUserId},
+        ${PEER3_ID},
+        ${scopeJson}::jsonb,
+        ${'active'},
+        now()
+      )
+    `;
+
+    const before = await sql!`
+      SELECT count(*)::int AS cnt FROM federation_grants WHERE peer_id = ${PEER3_ID}
+    `;
+    expect(before[0]!['cnt']).toBe(1);
+
+    // Delete peer → grants should cascade-delete.
+    await sql!`DELETE FROM federation_peers WHERE id = ${PEER3_ID}`;
+
+    const after = await sql!`
+      SELECT count(*)::int AS cnt FROM federation_grants WHERE peer_id = ${PEER3_ID}
+    `;
+    expect(after[0]!['cnt']).toBe(0);
+
+    // Cleanup
+    await sql!`DELETE FROM users WHERE id = ${cascadeGrantUserId}`.catch(() => {});
+  }, 15_000);
+});

packages/db/src/federation.ts

@@ -0,0 +1,20 @@
+/**
+ * Federation schema re-exports.
+ *
+ * The actual table and enum definitions live in schema.ts (alongside all other
+ * Drizzle tables) to avoid CJS/ESM cross-import issues when drizzle-kit loads
+ * schema files via esbuild-register.  Application code that wants named imports
+ * for federation symbols should import from this file.
+ *
+ * M2-01: DB tables and enums only. No business logic.
+ * M2-03 will add JSON schema validation for the `scope` column.
+ * M4 will write rows to federation_audit_log.
+ */
+
+export {
+  peerStateEnum,
+  grantStatusEnum,
+  federationPeers,
+  federationGrants,
+  federationAuditLog,
+} from './schema.js';

packages/db/src/index.ts

@@ -2,6 +2,7 @@ export { createDb, type Db, type DbHandle } from './client.js';
 export { createPgliteDb } from './client-pglite.js';
 export { runMigrations } from './migrate.js';
 export * from './schema.js';
+export * from './federation.js';
 export {
   eq,
   and,

packages/db/src/schema.ts

@@ -5,6 +5,7 @@
 
 import {
   pgTable,
+  pgEnum,
   text,
   timestamp,
   boolean,
@@ -585,3 +586,194 @@ export const summarizationJobs = pgTable(
   },
   (t) => [index('summarization_jobs_status_idx').on(t.status)],
 );
+
+// ─── Federation ──────────────────────────────────────────────────────────────
+// Enums declared before tables that reference them.
+// All federation definitions live in this file (avoids CJS/ESM cross-import
+// issues when drizzle-kit loads schema files via esbuild-register).
+// Application code imports from `federation.ts` which re-exports from here.
+
+/**
+ * Lifecycle state of a federation peer.
+ * - pending:   registered but not yet approved / TLS handshake not confirmed
+ * - active:    fully operational; mTLS verified
+ * - suspended: temporarily blocked; cert still valid
+ * - revoked:   cert revoked; no traffic allowed
+ */
+export const peerStateEnum = pgEnum('peer_state', ['pending', 'active', 'suspended', 'revoked']);
+
+/**
+ * Lifecycle state of a federation grant.
+ * - active:  grant is in effect
+ * - revoked: manually revoked before expiry
+ * - expired: natural expiry (expires_at passed)
+ */
+export const grantStatusEnum = pgEnum('grant_status', ['active', 'revoked', 'expired']);
+
+/**
+ * A registered peer gateway identified by its Step-CA certificate CN.
+ * Represents both inbound peers (other gateways querying us) and outbound
+ * peers (gateways we query — identified by client_key_pem being set).
+ */
+export const federationPeers = pgTable(
+  'federation_peers',
+  {
+    id: uuid('id').primaryKey().defaultRandom(),
+
+    /** Certificate CN, e.g. "gateway-uscllc-com". Unique — one row per peer identity. */
+    commonName: text('common_name').notNull().unique(),
+
+    /** Human-friendly label shown in admin UI. */
+    displayName: text('display_name').notNull(),
+
+    /** Pinned PEM certificate used for mTLS verification. */
+    certPem: text('cert_pem').notNull(),
+
+    /** Certificate serial number — used for CRL / revocation lookup. */
+    certSerial: text('cert_serial').notNull().unique(),
+
+    /** Certificate expiry — used by the renewal scheduler (FED-M6). */
+    certNotAfter: timestamp('cert_not_after', { withTimezone: true }).notNull(),
+
+    /**
+     * Sealed (encrypted) private key for outbound connections TO this peer.
+     * NULL for inbound-only peer rows (we serve them; we don't call them).
+     */
+    clientKeyPem: text('client_key_pem'),
+
+    /** Current peer lifecycle state. */
+    state: peerStateEnum('state').notNull().default('pending'),
+
+    /** Base URL for outbound queries, e.g. "https://woltje.com:443". NULL for inbound-only peers. */
+    endpointUrl: text('endpoint_url'),
+
+    /** Timestamp of the most recent successful inbound or outbound request. */
+    lastSeenAt: timestamp('last_seen_at', { withTimezone: true }),
+
+    createdAt: timestamp('created_at', { withTimezone: true }).notNull().defaultNow(),
+
+    /** Populated when the cert is revoked; NULL while the peer is active. */
+    revokedAt: timestamp('revoked_at', { withTimezone: true }),
+  },
+  (t) => [
+    // CRL / revocation lookups by serial.
+    index('federation_peers_cert_serial_idx').on(t.certSerial),
+    // Filter peers by state (e.g. find all active peers for outbound routing).
+    index('federation_peers_state_idx').on(t.state),
+  ],
+);
+
+/**
+ * A grant lets a specific peer cert query a specific local user's data within
+ * a defined scope. Scopes are validated by JSON Schema in M2-03; this table
+ * stores them as raw jsonb.
+ */
+export const federationGrants = pgTable(
+  'federation_grants',
+  {
+    id: uuid('id').primaryKey().defaultRandom(),
+
+    /**
+     * The local user whose data this grant exposes.
+     * Cascade delete: if the user account is deleted, revoke all their grants.
+     */
+    subjectUserId: text('subject_user_id')
+      .notNull()
+      .references(() => users.id, { onDelete: 'cascade' }),
+
+    /**
+     * The peer gateway holding the grant.
+     * Cascade delete: if the peer record is removed, the grant is moot.
+     */
+    peerId: uuid('peer_id')
+      .notNull()
+      .references(() => federationPeers.id, { onDelete: 'cascade' }),
+
+    /**
+     * Scope object — validated by JSON Schema (M2-03).
+     * Example: { "resources": ["tasks", "notes"], "operations": ["list", "get"] }
+     */
+    scope: jsonb('scope').notNull(),
+
+    /** Current grant lifecycle state. */
+    status: grantStatusEnum('status').notNull().default('active'),
+
+    /** Optional hard expiry. NULL means the grant does not expire automatically. */
+    expiresAt: timestamp('expires_at', { withTimezone: true }),
+
+    createdAt: timestamp('created_at', { withTimezone: true }).notNull().defaultNow(),
+
+    /** Populated when the grant is explicitly revoked. */
+    revokedAt: timestamp('revoked_at', { withTimezone: true }),
+
+    /** Human-readable reason for revocation (audit trail). */
+    revokedReason: text('revoked_reason'),
+  },
+  (t) => [
+    // Hot path: look up active grants for a subject user (auth middleware).
+    index('federation_grants_subject_status_idx').on(t.subjectUserId, t.status),
+    // Hot path: look up active grants held by a peer (inbound request check).
+    index('federation_grants_peer_status_idx').on(t.peerId, t.status),
+  ],
+);
+
+/**
+ * Append-only audit log of all federation requests.
+ * M4 writes rows here. M2 only creates the table.
+ *
+ * All FKs use SET NULL so audit rows survive peer/user/grant deletion.
+ */
+export const federationAuditLog = pgTable(
+  'federation_audit_log',
+  {
+    id: uuid('id').primaryKey().defaultRandom(),
+
+    /** UUIDv7 from the X-Request-ID header — correlates with OTEL traces. */
+    requestId: text('request_id').notNull(),
+
+    /** Peer that made the request. SET NULL if the peer is later deleted. */
+    peerId: uuid('peer_id').references(() => federationPeers.id, { onDelete: 'set null' }),
+
+    /** Subject user whose data was queried. SET NULL if the user is deleted. */
+    subjectUserId: text('subject_user_id').references(() => users.id, { onDelete: 'set null' }),
+
+    /** Grant under which the request was authorised. SET NULL if the grant is deleted. */
+    grantId: uuid('grant_id').references(() => federationGrants.id, { onDelete: 'set null' }),
+
+    /** Request verb: "list" | "get" | "search". */
+    verb: text('verb').notNull(),
+
+    /** Resource type: "tasks" | "notes" | "memory" | etc. */
+    resource: text('resource').notNull(),
+
+    /** HTTP status code returned to the peer. */
+    statusCode: integer('status_code').notNull(),
+
+    /** Number of items returned (NULL for non-list requests or errors). */
+    resultCount: integer('result_count'),
+
+    /** Why the request was denied (NULL when allowed). */
+    deniedReason: text('denied_reason'),
+
+    /** End-to-end latency in milliseconds. */
+    latencyMs: integer('latency_ms'),
+
+    createdAt: timestamp('created_at', { withTimezone: true }).notNull().defaultNow(),
+
+    // Reserved for M4 — see PRD 7.3
+    /** SHA-256 of the normalised GraphQL/REST query string; written by M4 search. */
+    queryHash: text('query_hash'),
+    /** Request outcome: "allowed" | "denied" | "partial"; written by M4. */
+    outcome: text('outcome'),
+    /** Response payload size in bytes; written by M4. */
+    bytesOut: integer('bytes_out'),
+  },
+  (t) => [
+    // Per-peer request history in reverse chronological order.
+    index('federation_audit_log_peer_created_at_idx').on(t.peerId, t.createdAt.desc()),
+    // Per-user access log in reverse chronological order.
+    index('federation_audit_log_subject_created_at_idx').on(t.subjectUserId, t.createdAt.desc()),
+    // Global time-range scans (dashboards, rate-limit windows).
+    index('federation_audit_log_created_at_idx').on(t.createdAt.desc()),
+  ],
+);