IITA cassava breeding data demo (Streamlit + DuckDB + Claude)
# IITA Breeding Data Assistant
Streamlit demo: ask plain-English questions over IITA cassava UYT phenotype data (8 locations). Every number comes from live SQL against DuckDB.
**Live app:**
jambo-antoine-iita-breeding…
**Repo:**
github.com
## Deploy (Streamlit Community Cloud — free)
1. Open the prefilled deploy link:
share.streamlit.io
2. Sign in with GitHub and authorize Streamlit if prompted.
3. Confirm **Main file path** = `app.py`.
4. Open **Advanced settings → Secrets** and paste the contents of your local `.streamlit/secrets.toml`
(generate it with `python _make_secrets_toml.py` from `.env` — never commit secrets).
5. Click **Deploy**.
App URL will look like: `
https://iita-breeding-demo-…`
## Local run
```bash
python -m venv .venv
.venv\Scripts\activate # Windows
pip install -r requirements.txt
# put ANTHROPIC_API_KEY in .env
python ingest.py # only if cassava.duckdb is missing
streamlit run app.py
```
## API spend guardrails
Shared daily call/USD caps, per-session action limits, and cooldowns are configured via `DEMO_*` env vars / Streamlit secrets. See `.streamlit/secrets.toml.example`.