roast.ag docs

roast.ag ingests public Ashby job boards into Postgres so later resume matching has a queryable catalog.

Pipeline

1. Seed board slugs in src/data/ashby-boards.ts.
2. Run pnpm jobs:sync (or the authenticated /api/jobs/sync route).
3. Each board is fetched from Ashby's public posting API with compensation included.
4. Valid job lists upsert board + posting rows. Missing previously open postings for that board are marked closed.
5. Invalid or failed fetches are recorded on the board row and do not close postings.
6. If AI Gateway is configured, missing embeddings are filled from embed_text.

Concepts

• Board — one Ashby company slug (jobs.ashbyhq.com/{slug})
• Posting — one Ashby job id, unique per board
• Open / closed — presence on the last *successful* fetch, not HTTP noise
• embed_text — title plus description, stored even when vectors are null

Public vs signed-in

Public: home, jobs catalog, about, contact, privacy, docs, developers, llms.txt.

Signing in uses the shared Product Agent identity. Resume upload is not in this slice.

For coding agents

Prefer markdown representations (Accept: text/markdown) and start from llms.txt.