# Latent: FDA Ingestor (Tier 0)

This is the "Ground Truth" ingestor for the Latent system. It fetches the latest Structured Product Labels (SPL) from the OpenFDA API for target molecules and converts them into **StemeDB Assertions**.

## Scope (Week 1)
- **Source:** OpenFDA API (`api.fda.gov`)
- **Target Molecules:** Semaglutide, Tirzepatide, Liraglutide
- **Sections Extracted:** 
  - `boxed_warning` (Black box warnings)
  - `adverse_reactions` (Side effects list)
  - `warnings_and_precautions` (General safety)

## Usage

1. Install dependencies:
   ```bash
   pip install -r requirements.txt
   ```

2. Run the ingestor:
   ```bash
   python main.py
   ```

3. Output:
   - Creates `tier0_regulatory_graph.jsonl`.
   - Each line is a JSON object representing a StemeDB Assertion with `source_class: 0`.

## Next Steps
- Implement NLP entity extraction to break the large text blocks into granular assertions (e.g., "causes nausea" instead of the full text block).
- Connect directly to the StemeDB Rust bindings instead of outputting JSONL.