# Code / standard reference corpus (Stage 7)

Drop plain-text `.txt` files here. Stage 7 (`code_review.py`) retrieves the most
relevant excerpts and lets the model cite section numbers **only** when they
appear in a retrieved excerpt. An empty corpus is valid — Stage 7 then runs
reasoning-only and cites nothing.

## File naming → standard

The standard is inferred from the filename prefix:

| prefix          | standard |
|-----------------|----------|
| `ada*.txt`      | ADA      |
| `tas*.txt`      | TAS      |
| `ibc*.txt`      | IBC      |
| `ifc*.txt`      | IFC      |
| `iebc*.txt`     | IEBC     |
| `energy*.txt`   | ENERGY   |

(Only `.txt` files are loaded; this `README.md` is ignored.)

## Format

Plain text, chunks separated by a blank line. If a chunk begins with a section
number (e.g. `404.2.3 Doors ...`), that number is captured as its section id and
is what the model may cite.

```
404.2.3 Door and Gate Opening Force
Fire doors shall have a minimum opening force allowable by the appropriate
administrative authority. The force for pushing or pulling open ...

405.2 Slope
Ramp runs shall have a running slope not steeper than 1:12.
```

## Licensing note

- **ADA 2010 Standards** and **Texas Accessibility Standards (TAS)** are
  government documents and can be included here.
- **IBC / IFC / IECC** are ICC copyrighted works — do **not** paste their text
  unless you have a license. Provide your own licensed copy if needed.