Add url parameter to ai::extract_roaster(), extract_roast(), and
extract_bag_scan() instead of using a hardcoded constant. Wire through
AppState.openrouter_url (defaulting to the production URL), matching
the existing pattern used for foursquare_url.
Add test infrastructure: session creation helper, page_url() on
TestApp, and spawn_app_with_openrouter_mock() for wiremock-based
AI extraction testing.
Add Usage struct to parse prompt_tokens, completion_tokens,
total_tokens, and cost from every OpenRouter response. Extraction
functions now return (result, Option<Usage>) tuples. New ai_usage
table stores per-call records with a repository trait and SQL
implementation.
- Add migration 0016 to drop notes column from roasters table
- Remove notes from domain structs, repositories, views, CLI, routes
- Remove notes from AI extraction prompts and ExtractedRoaster struct
- Remove notes from scan page form and submission handler
- Update templates, bootstrap script, backup module, and tests
Add /scan page that lets authenticated users photograph or describe a
coffee bag, extracts both roaster and roast data via a single AI call,
and creates both entities on submit. Existing roasters are matched by
slug to avoid duplicates. Camera icon added to nav bar.
Users can upload a photo of a coffee bag or type a text prompt to
auto-fill roaster/roast form fields via OpenRouter's chat completions
API. The feature is opt-in: hidden when BREWLOG_OPENROUTER_API_KEY is
not set.
- Add src/infrastructure/ai.rs with OpenRouter client, prompt
templates, JSON extraction from model responses, and unit tests
- Add extract-roaster and extract-roast API endpoints behind auth
- Add extraction UI (photo button + text input) to both form templates
- Add openrouter_api_key and openrouter_model to CLI args and AppState
- Enable reqwest gzip feature for compressed API responses
- Default model set to openrouter/free (auto-selects available models)
- Prompts instruct model to use web search for missing details
