code-kb Telemetry

Summarize code-kb tool usage, token savings, and recent errors from ~/.code-kb/telemetry.db.

Steps

1. Read the arguments. The window defaults to all. The word workspace sets workspace_only=true; otherwise report all workspaces. telemetry_summary takes no project_root: workspace_only scopes to the project of the most recent code-kb call.
2. Call telemetry_summary(time_window=<window>, workspace_only=<bool>).
3. Reply in this shape:
   • First line: scope, window, total calls, success rate, tokens served, tokens saved with its coverage.
   • Then only what needs attention: tools under 95% success, errors that repeat, calls to tool names that do not exist.
   • Skip tools with nothing notable. Do not paste the whole table.
4. Report tokens saved with the coverage figure beside it. The header reads Est. Tokens Saved: ~N (baseline known for K of M calls), and each tool row reads ~N (K/M). Never quote the number without the coverage.
5. End with the terminal twin so the user can verify:

code-kb stats --since <window> [--workspace]

What "saved" means

The saved figure is the size, in estimated tokens, of the files the answer points into, minus the tokens served. A skeleton, body, or context read is measured against its own file. A lookup, search, references, blast-radius, or facts answer is measured against the distinct files its rows name, at most 20 files. A call with no file to point at, such as an outline or a telemetry summary, records no baseline and is not counted as saved. Calls recorded before this measurement existed keep no baseline either, so old windows show a lower coverage.

It's working if

• The reply fits on one screen and names the window it covers.
• The saved number carries its coverage figure.
• Every flagged item points at one tool or one repeated error.