Job_search_AI

Local pipeline: import job postings from JSON, score matches with Ollama, tailor your CV, store drafts in SQLite. Nothing is marked “sent” or exported until you approve (CLI prompt or API).

Prerequisites

• Python 3.10+
• Ollama running locally (default API: ` )

Install

Optional: create a .env in the repo root if you use env overrides (e.g. OLLAMA_HOST, JOB_SEARCH_API_KEY).

Ollama model

Use the same model name as in data/preferences.json (ollama_model), or change that field to match what you pull:

Optional: point the app at another Ollama URL (overrides ollama_base_url in preferences):

(On Linux/macOS, use export instead of set.)

Run the CLI

Default paths: data/preferences.json, data/jobs_inbox.json, data/my_cv.txt, data/applications.db.

Useful flags: --max-items, --preferences, --jobs-inbox, --cv, --db, --tone, --preview-chars. Example:

Run the API (Uvicorn)

• OpenAPI UI: `
• Health (Ollama reachability): `GET

By default, non-document routes only accept loopback clients. If JOB_SEARCH_API_KEY is set, send header X-API-Key for remote access.

Sample JSON shapes

data/preferences.json (object)

Minimal useful fields for Phase 1 include match_threshold, blacklist_companies, and Ollama settings. Example:

Extra keys are allowed; see the full sample in data/preferences.json in the repo.