Imported from tommy260803/ReunionesAuto (AGENTS.md). Install upstream with npx skills add tommy260803/ReunionesAuto. Copyright stays with the author.

AGENTS.md - ReunionesAuto (Zoom2)

Current Stack

• Frontend: Next.js 16 App Router, React 19, TypeScript, Tailwind CSS 4 and Recharts in frontend/.
• Backend: FastAPI in backend/app/ with a raw requests Supabase REST client.
• Database: Supabase PostgreSQL. Versioned SQL scripts live in querys para supabase/.
• Automation: n8n workflows in json n8n/ for Zoom, meeting creation and AI summaries.
• Legacy: app.py is the former Streamlit implementation and is not the active application.

Setup

1. Run SQL scripts in order: query1.txt, query2.txt, query3.txt, query4.txt, insert_sample_tasks.sql, query5_metricas.txt, query6_reuniones_participantes.sql, query7_resumenes_modulo.sql, query8_metricas_inferenciales.sql, query14_actas_modulo.sql.
2. Import the required workflows from json n8n/ and configure their credentials.
3. Configure root .env, including SUPABASE_SERVICE_ROLE_KEY, n8n webhook URLs, N8N_CALLBACK_SECRET, N8N_WORKFLOW_VERSION, and GROQ_API_KEY (or OPENAI_API_KEY) for AI-powered acta generation.
4. Install backend dependencies from backend/requirements.lock.txt for a reproducible environment.
5. Install frontend dependencies with npm ci in frontend/.

Commands

# Backend
backend\.venv\Scripts\python.exe -m uvicorn app.main:app --reload --port 8080 --app-dir backend
backend\.venv\Scripts\python.exe -m unittest discover -s backend\tests -t backend -v

# Frontend
npm run dev --prefix frontend
npm run build --prefix frontend

Statistical Metrics

• metricas_n8n uses one row per invocation and is written with the Supabase service role.
• Production analysis excludes legacy, demo and test rows.
• Latency analysis uses successful terminal end-to-end durations from one selected endpoint.
• The primary latency test is two-sided Welch's t-test.
• Outcomes use uncorrected Pearson chi-square when every expected cell is at least 5, otherwise two-sided Fisher exact.
• A non-significant result is not described as stability or equivalence.
• Do not use seed_metricas_estadisticas.sql as empirical article evidence.

Key Conventions

• Internal telemetry outcomes are pending, success, error, timeout and cancelled.
• Store timestamps in UTC and analyze inclusive calendar dates as half-open UTC intervals.
• One asynchronous job is correlated through correlation_id and finalized by its callback.
• Admin access is checked by the backend dependency in backend/app/core/dependencies.py.
• Never expose SUPABASE_SERVICE_ROLE_KEY to the frontend.

Actas Module

• New table actas with auto-incremented numero, meeting FK, structured sections (decisiones, tareas, proximos_pasos).
• Backend: backend/app/actas/ with router, schemas. Services: document_parser.py (PDF/Word extraction), acta_generator.py (LLM-powered acta generation).
• Frontend: /actas page with upload (PDF/Word), text paste, edit, finalize, and filter by meeting type.
• Supports all meeting types: virtual, presencial, mixta.
• LLM config: GROQ_API_KEY or OPENAI_API_KEY in .env, defaults to Groq llama-3.3-70b-versatile.