LangAi Cursor Rules

You are working in the "Company Researcher" repository (Python 3.11+).

Project conventions:
- Primary package: `src/company_researcher/`
- Repository CLI entrypoint: `run_research.py`
- Docs live under `docs/` (index: `docs/README.md`)
- Deployment assets live under `deploy/` (Dockerfile, docker-compose, Kubernetes manifests)

When updating documentation:
- Prefer repo-root commands that work on Windows PowerShell and macOS/Linux
- Keep doc links relative and valid
- If you change an entrypoint (CLI/API), update the docs and deployment manifests in the same change

When updating code:
- Keep changes small and readable
- Add helpful errors for missing optional dependencies (FastAPI/Celery/etc.)

# Cursor Project Rules (.cursorrules)

This repository is **Company Researcher**: a Python multi-agent research system built on **LangGraph + LangChain + Pydantic**.

## Priorities

- Prefer **correctness and traceability** over cleverness.
- Prefer **small diffs** and preserve existing behavior unless explicitly asked.
- Prefer **batching** when reading multiple files (faster + cheaper).

## Coding standards (Python)

- Follow existing tooling in `pyproject.toml`:
  - Formatting: **Black**
  - Linting: **Ruff**
  - Types: **mypy** (gradual)
  - Tests: **pytest**
- Prefer explicit types for public APIs and Pydantic models.
- Prefer small, testable functions.
- Never commit secrets; use environment variables and `env.example`.

## Project conventions

- Source code: `src/`
- Tests: `tests/`
- Generated artifacts: `outputs/` (avoid editing these unless explicitly requested)

## LangGraph / agent conventions

- State must be explicit and serializable.
- External calls (web search, HTTP) must have timeouts and clear error messages.
- Log enough to debug runs, but never log secrets.

## Rule conflicts

- Prefer **project reality** (Python/LangGraph) over any legacy rules/prompts that mention .NET/XML.
- Prefer `.cursor/rules/` for detailed standards; this file is only the high-level overlay.