Single source of truth for repository engineering metrics. Last updated: 2026-08-30 Repository revision:
5d0571689ffc4d50aaf9c94dd2ecada856530033Regenerate:python3 scripts/generate_project_metrics.py
The local results below distinguish collected, passed, and skipped tests. Browser tests are collection-verified here; they are executed separately by the Playwright CI job. Coverage and CI evidence are pinned to the verified commit above; coverage is marked unverified if this document is regenerated at another revision.
| Metric | Value |
|---|---|
| Application modules | 19 |
| REST endpoints | 84 |
| WebSocket endpoints | 2 |
| Non-browser tests collected | 417 |
| Non-browser tests passed | 407 |
| Non-browser tests skipped | 10 |
| Browser tests collected | 9 |
| Total tests collected | 426 |
Verified coverage (src/voxforge, non-browser suite) |
76.40% |
| ADRs | 7 |
| Architecture documents | 26 |
| Benchmark documents | 4 |
- Local non-browser command:
PYTHONPATH=. pytest -q --ignore=tests/browser --cov=src/voxforge - Browser collection command:
PYTHONPATH=. pytest tests/browser --collect-only -q - Verified GitHub CI: run 33298953452 passed on
2026-08-30 for commit
5d057168.
agent_configagent_orchestratoralertsauthconversationdashboardevaluationhandoffknowledgelivekit_gatewaymcp_tool_routermemoryonboardingoutcomesreplaysession_managersttttsvoice_gateway
| Transport | Count | Entry points |
|---|---|---|
| REST | 84 | /api/v1/* routers |
| WebSocket | 2 | /api/v1/ws/voice, /lk/{path} |
Run non-browser tests: make test
Run Playwright tests: make test-browser
| Category | Location |
|---|---|
| Unit | tests/unit/ |
| Integration | tests/integration/ |
- ADR-001-programmatic-voice-pipeline-runner.md
- ADR-002-onboarding-session-lifecycle.md
- ADR-003-mcp-runtime-discovery.md
- ADR-004-livekit-transport-adapter.md
- ADR-005-enterprise-knowledge-base.md
- ADR-006-human-handoff.md
- ADR-008-retrieval-pipeline.md
- README.md
- agent-config-versioning.md
- agent-orchestrator.md
- alerts.md
- authentication.md
- ci-hardening.md
- customer-support-tools.md
- dashboard.md
- evaluation-engine.md
- failure-recovery.md
- human-handoff.md
- integrity-concurrency.md
- knowledge-base.md
- livekit-integration.md
- livekit-webrtc.md
- mcp-runtime-discovery.md
- mcp-tool-router.md
- memory.md
- metrics.md
- observability.md
- onboarding-voice-pipeline.md
- outcomes.md
- rate-limiting.md
- replay.md
- session-consistency.md
- voice-pipeline.md
| Capability | Providers |
|---|---|
| STT | deepgram, mock |
| LLM | openai, mock |
| TTS | cartesia, mock |
| Embeddings | openai (text-embedding-3-small) |
| WebRTC transport | LiveKit |
| Voice transport | WebSocket, LiveKit WebRTC |
MCP servers are runtime-discovered from MCP_SERVERS_CONFIG (stdio transport). Static
tool metadata is used as a degraded fallback when discovery fails. Inspect live status via:
GET /api/v1/tools/mcp/healthGET /api/v1/tools/mcp/servers
No servers are hardcoded in the repository; operators declare servers in environment config.
| Phase | Status |
|---|---|
| Phase 0 — Stabilization | Complete |
| Phase 1 — Onboarding voice pipeline | Complete |
| Phase 2 — CI hardening | Complete |
| Phase 3 — MCP runtime discovery | Complete |
| Phase 4 — LiveKit transport adapter | Complete |
| Phase 5 — Public deployment | Complete |
| Production hardening & load testing | Planned |