README.md

---
title: deep_research
app_file: deep_research.py
sdk: gradio
sdk_version: 6.10.0
---

Multi-agent deep research: plan web searches, run them, evaluate coverage, optionally replan, then write a markdown report.

## Secrets (Settings → Repository secrets)

| Name | Required | Purpose |
|------|----------|---------|
| `OPENAI_API_KEY` | Yes | OpenAI API for agents, models, and web search tool |
| `SENDGRID_API_KEY` | No | If set, emails the report via SendGrid |

## Deploy from this repo folder

Create a **new Gradio Space** on Hugging Face, then upload or connect git with **this directory** (`2_openai/deep_research`) as the Space root so `deep_research.py`, `requirements.txt`, and all `*_agent.py` / `research_manager.py` files sit at the top level of the Space repository.

If you keep the project inside the full course monorepo, either use a **subdirectory deploy** (if your workflow supports it) or copy these files into a dedicated Space repo.

## Local run

```bash
pip install -r requirements.txt
python deep_research.py
```

The same file is the Space entry point (`app_file` above); Hugging Face runs it as `python deep_research.py`.