[sagnik11]

In connected mode, data from a repo that no org had connected was uploaded to the login org: notes, commit summaries with prompts, file-change counts and transcripts. A personal or client repo worked on from a work login ended up in the workplace dashboard.

Changes

• route_for_repo (src/api/client.rs) decides per repo: Upload to the org that connected it, NotConnected (keep the data local and check again daily), or Unavailable (retry on the normal schedule). It never falls back to the login org.
• resolve-org now works for device-code logins (bearer token). Answers are cached for 1h and errors are never cached.
• Transcripts carry repo_url in their CAS metadata and are routed per repo (cas_bridge, flush_cas).
• Notes, commit summaries, file changes and CAS park unconnected records, and records the server refused as repo_not_connected, for 24h without incrementing attempts.
• autter sync status (and --json → not_connected) shows how many records are kept on this computer.
• Bumps the version to 3.1.0 and adds the changelog entry. Merging tags v3.1.0, and the Release workflow publishes it.

⚠️ Merge order

The API change has to be deployed first: Autter-dev/autter-monorepo#677. Without it, resolve-org with a bearer token returns 400, so device-login users' uploads would fail and be retried.

Not built or tested locally (no Rust toolchain on this machine, and no tests were added, as requested). CI is the first compile, and cargo fmt hasn't been run.

🤖 Generated with Claude Code

---

^{Open in [code]smith · Need help on this PR? Tag @codesmith-bot with what you need.}