Merge pull request #5 from airblackbox/refresh/current-stack

Rebuild platform around the consolidated stack

Author: shotwellj

.env.example

@@ -1,12 +1,20 @@
 # AIR Platform Configuration
 # Copy to .env and fill in your values
 
-# LLM Provider (required for live replay)
+# LLM provider key — sent by YOUR client code, not stored by the gateway.
+# Only needed to run the live round-trip integration test.
 OPENAI_API_KEY=sk-...
 
-# MinIO (defaults work for local dev)
+# Upstream provider the gateway proxies to (default: OpenAI)
+# PROVIDER_URL=https://api.openai.com
+
+# MinIO prompt vault credentials (defaults work for local dev)
 MINIO_ROOT_USER=minioadmin
 MINIO_ROOT_PASSWORD=minioadmin
 
-# Optional: Gateway authentication
+# Optional: require clients to send X-Gateway-Key with this value
 # GATEWAY_KEY=your-secret-key
+
+# Optional: pin the HMAC signing key for the audit chain.
+# When unset, the gateway generates one and persists it in ./runs/.air-signing-key
+# TRUST_SIGNING_KEY=

.github/workflows/ci.yml

@@ -21,9 +21,44 @@ jobs:
           python-version: "3.12"
 
       - name: Lint test files
+        run: python -m py_compile tests/*.py
+
+  integration:
+    runs-on: ubuntu-latest
+    needs: validate
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Start the stack
+        run: docker compose up -d --quiet-pull
+
+      - name: Wait for gateway health
         run: |
-          python -m py_compile tests/conftest.py
-          python -m py_compile tests/test_health.py
-          python -m py_compile tests/test_episode_flow.py
-          python -m py_compile tests/test_policy_flow.py
-          python -m py_compile tests/test_end_to_end.py
+          for i in $(seq 1 30); do
+            curl -sf http://localhost:8080/health && exit 0
+            sleep 2
+          done
+          echo "Gateway never became healthy" >&2
+          docker compose logs
+          exit 1
+
+      - name: Set up Python 3.12
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install test dependencies
+        run: pip install -r requirements.txt
+
+      - name: Run integration tests
+        env:
+          AIR_REQUIRE_STACK: "1"
+        run: python -m pytest tests/ -v
+
+      - name: Stack logs on failure
+        if: failure()
+        run: docker compose logs
+
+      - name: Tear down
+        if: always()
+        run: docker compose down -v

.gitignore

@@ -3,3 +3,4 @@ __pycache__
 *.pyc
 .pytest_cache
 *.egg-info
+runs/

CHANGELOG.md

@@ -4,6 +4,40 @@ All notable changes to **AIR Platform** will be documented in this file.
 
 This project adheres to [Semantic Versioning](https://semver.org/).
 
+## [0.2.0] — 2026-07-13
+
+### Changed
+- Rebuilt Docker Compose around the consolidated stack: gateway now pulled as the
+  signed `ghcr.io/airblackbox/airblackbox` image; collector built directly from the
+  public repo. No local sibling checkouts required — `docker compose up` works from
+  a fresh clone.
+- Rewrote README for the current ecosystem (airblackbox, air-gate, air-blackbox-mcp,
+  compliance-action, tombstone); removed links to archived repositories.
+- Replaced integration tests targeting the retired Episode Store / Policy Engine /
+  Eval Harness services with tests for the gateway audit surface (`/health`,
+  `/v1/audit`, `/v1/analytics`, `/v1/audit/export`) plus an optional live LLM
+  round-trip (needs `OPENAI_API_KEY`).
+- Tests now skip with instructions when the stack is down (hard-fail in CI via
+  `AIR_REQUIRE_STACK=1`).
+- CI now boots the full compose stack and runs the integration suite.
+
+### Added
+- `collector.yaml` — local-only OTel pipeline config mounted over the image default
+  (normalize → prompt vault → redact/metrics → batch). Nothing leaves your machine.
+- `guardrails.yaml` — gateway guardrails config (budgets, loop/retry protection,
+  tool allow/blocklists).
+
+### Fixed
+- Invalid `build-backend` in pyproject.toml (`setuptools.backends._legacy:_Backend`
+  → `setuptools.build_meta`).
+- Worked around duplicate `processors:` key in the upstream collector config that
+  crashed the collector on boot (fix also submitted to the main repo).
+
+### Removed
+- Obsolete tests and smoke script for archived services (`test_production.py`,
+  `test_e2e_stack.py`, `test_end_to_end.py`, `test_episode_flow.py`,
+  `test_policy_flow.py`).
+
 ## [0.1.0] — 2026-02-22
 
 - Docker Compose orchestration for full AIR Blackbox stack

Makefile

@@ -12,9 +12,8 @@ down:
 clean:
 	docker compose down -v
 
-# Run integration tests
+# Run integration tests against the running stack
 test:
-	docker compose exec eval-harness python3 -m pytest /app/tests/ -v
 	python3 -m pytest tests/ -v
 
 # View logs for all services
@@ -24,7 +23,5 @@ logs:
 # Check service health
 status:
 	@echo "=== Gateway ===" && curl -sf http://localhost:8080/health || echo "DOWN"
-	@echo "\n=== Episode Store ===" && curl -sf http://localhost:8100/v1/health || echo "DOWN"
-	@echo "\n=== Policy Engine ===" && curl -sf http://localhost:8200/v1/health || echo "DOWN"
-	@echo "\n=== MinIO ===" && curl -sf http://localhost:9000/minio/health/live || echo "DOWN"
+	@echo "\n=== MinIO ===" && curl -sf http://localhost:9000/minio/health/live > /dev/null && echo "UP" || echo "DOWN"
 	@echo "\n=== Jaeger ===" && curl -sf http://localhost:16686/ > /dev/null && echo "UP" || echo "DOWN"