eklavya 1.72.0 → 1.74.0

package/README.md

@@ -122,7 +122,7 @@ npx eklavya --version
 
 The [CLI reference](https://eklavya-run.web.app/docs/cli/) lists all commands and
 flags. `feedback generate` reviews one of your earlier prompts, using what your later prompts in that session had to add, when `feedback.enabled`,
-memory and `providers.observer` are set. The dashboard reads the database and has five guarded writes: a setting
+memory and `providers.observer` are set. The dashboard reads the database (its Concepts page shows each concept's tags and filters by one with `?tag=`) and has five guarded writes: a setting
 change, a correction of a missed answer (which can also complete a level), and acknowledging, deleting or counting the opening of a
 feedback item (`POST /api/feedback/acknowledge`, `/delete`, `/opened`; the reads are `GET /api/feedback` and `/api/feedback/list`). An open page listens to
 `GET /api/events`, a stream of the change cursor (a counter that moves on every
@@ -135,7 +135,7 @@ it cannot parse gets a 400 without stopping the server
 ([what it is, underneath](https://eklavya-run.web.app/docs/dashboard/#what-it-is-underneath)).
 Its Learning page shows your streak and its calendar under the summary tiles; a
 **To be corrected** tile there and on the Artifacts page counts missed answers still
-waiting for a correction. The
+waiting for a correction; correcting one closes its dialog after three seconds unless you choose **Stay open**. The
 review queue gives each project a command to copy that starts Claude Code on its due
 concepts, in the project's folder when that folder is available.
 Its feature tips show the archer companion and use Driver.js, bundled in `dist/` at build and served by the
@@ -147,8 +147,11 @@ covers retrieval, processing, providers, privacy and sync.
 State is local under `~/.eklavya/`, including `knowledge.db` (SQLite, WAL).
 `EKLAVYA_HOME` and `EKLAVYA_DB` override those paths. Questions and memory have
 independent switches. Set `providers.observer` to `null` for local memory
-summaries and no prompt feedback. Each feedback item shows the project, host,
-session and time it came from, read from memory when you open it;
+summaries and no prompt feedback. Each feedback item leads with a verdict that
+says, in words and never a score, whether the prompt left each area clear, left
+it out or did not need it (an older item says "Not flagged" for an area it
+did not name), and shows the project, host, session and time it came
+from, read from memory when you open it;
 see [your data](https://eklavya-run.web.app/docs/your-data/). Anonymous daily
 usage counts are on by default and never include paths, names or text;
 [usage analytics](https://eklavya-run.web.app/docs/usage-analytics/) lists every