bc-observatory 0.2.0 → 0.3.0

package/README.md

@@ -16,11 +16,26 @@ The server downloads the site's search index (cached by hash in `~/.cache/bc-obs
 markdown. `BC_OBSERVATORY_LOCAL=<checkout of waldo1001/waldo.BCObservatory>` reads everything from disk instead;
 `BC_OBSERVATORY_SITE` points at another deployment.
 
-`search` is hybrid: keywords plus meaning, so "who is allowed to see what" finds the permission sets. Meaning comes from
-static embeddings ([model2vec](https://github.com/MinishLab/model2vec) `minishlab/potion-base-8M`, MIT) computed locally in
+`search` ranks with the site's own scorer (`packages/search`, bundled into `dist/server.js`), so an agent and a reader get the
+same order: references (`t36`, `cu 80`, `table 11300 BE`) and bare ids (`36`) go straight to the object; per query word
+the title and an AL object's caption weigh 3, tags 2, summary 1, through the exact word, a prefix, the plural, a synonym
+(`client` is Customer, `G/L` is general ledger) or one typo; a name equal to the query adds 10; base app before
+first-party app before country layer, then what references, calls and documents an object; topic hubs by their size and
+review state, so "subscription" returns the Subscription billing hub before the codeunits that share the word. Fields,
+events, global procedures and enum values are searchable too (`OnAfterPostSalesDoc`, `field: Posting Date`); their path
+ends in an anchor (`objects/codeunit/80#event-OnAfterPostSalesDoc`). Filters: `type`, `tier`, `system`, `country`,
+`app`, `object_type`, `kind`. Each result says where it sits (`in:` the Learn TOC path, the app, the channel) and, for a
+hub, its size and review state.
+
+The search is hybrid by default: the exact band (references and exact names) stays on top and the rest is fused with a
+meaning ranking, so "who is allowed to see what" finds the permission sets; each hit says `[keyword]`, `[meaning]` or
+`[both]`. Meaning comes from static embeddings ([model2vec](https://github.com/MinishLab/model2vec) `minishlab/potion-base-8M`, MIT) computed locally in
 plain JavaScript; the 30 MB model is downloaded once on the first search, pinned by revision and checked by SHA-256. Pass
-`mode: "keyword"` for exact terms such as an object name. `BC_OBSERVATORY_EMBEDDINGS=0` turns it off, and
+`mode: "keyword"` for the scorer alone. `BC_OBSERVATORY_EMBEDDINGS=0` turns it off, and
 `BC_OBSERVATORY_MODEL_DIR=<dir with config.json, tokenizer.json, model.safetensors>` uses a model on disk (offline). Without a
 model, search is keyword-only and says so.
 
+`get_object` takes an id, a reference (`t36`, `36-be`), a name or a caption, and a `country`; `whats_new` lists nothing
+dated after today unless `include_future: true`.
+
 MIT. Content: see the repository's CONTENT-NOTICE.md (Microsoft content attributed; community content derived only).