Turn one requirement into a clean SRS using a multi-agent pipeline.
Local demo recording:
Screen.Recording.2026-04-23.163110.mp4
Mode toggle behavior:
docs/Screen Recording 2026-04-23 163110.mp4
- Accepts a requirement from UI or API
- Runs all workflow nodes:
planner -> architect -> developer -> qa -> reviewer -> discussion -> erd -> timeline -> srs
- Saves run history and step snapshots
- Exports SRS as markdown
- Shows ERD diagram in the UI (Mermaid rendered)
- ON: same full node chain, simpler output per node, fastest response
- OFF: same full node chain, slightly richer output, still fast
You can toggle this from the web UI before starting a run.
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
copy .env.example .envMinimum required:
POSTGRES_DSN- one API key:
GOOGLE_API_KEYorGROQ_API_KEY
python run.pyOpen:
http://127.0.0.1:8000/(UI)http://127.0.0.1:8000/health(health check)
curl -X POST http://127.0.0.1:8000/api/v1/build ^
-H "Content-Type: application/json" ^
-d "{\"user_requirement\":\"Build a task API with auth and tests.\",\"fast_mode\":true}"Useful endpoints:
POST /api/v1/buildGET /api/v1/runs/{run_id}GET /api/v1/runs/{run_id}/progressGET /api/v1/runs/{run_id}/srsGET /api/v1/runs/{run_id}/srs.md
FAST_MODE=true(default complexity mode)GOOGLE_API_KEY,GOOGLE_MODELGROQ_API_KEY,GROQ_MODELFALLBACK_TO_OLLAMA_ON_QUOTA_EXCEEDEDSTEP_TIMEOUT_SECONDS,STEP_MAX_RETRIES
app/
agents/ workflow logic
api/ REST endpoints
core/ config/security/export
memory/ DB storage and retrieval
schemas/ pydantic contracts
ui/ web UI
docs/ setup, runbook, plans
tests/ tests
run.py one-click launcher