[enyst]

This notebook explains how Insider Cat uses one saved OpenHands conversation for text, Voice, tools, and history. It separates Agent Server's responsibility for saved work and credentials from Canvas's companion surface and the App's controls and audio.

The pages describe the current Cat interface, direct realtime handoff, normal tools and enabled skills, and backend inspection through the terminal. Source-derived drawings illustrate the design; they are not screenshots or proof of a live provider call. Earlier relay experiments remain dated history.

Linked work

SDK issue #5515 tracks shared acceptance criteria and current verification; Canvas issue #17557 tracks host behavior. This repository has issues disabled.

Implementation PRs: SDK #5180, Canvas #17558, and Insider #2.

Current update and validation

Commit 38b90d1800287f11da13b17b28ab33e29d8d9c4c adds a 5 October maintenance note and links the acceptance criteria from the notebook's status page. It explains that the failed macOS job stopped at a PyPI dependency download before build/test execution, and distinguishes the new regression checks from September's real-provider observations.

The final merge, 1e3ab9aa9d41db412f70bf6630a9591aa0d857c9, also brings in current main (024ccdcb18485a84280a909ce4f95100284ec0d6). It resolves one conflict in the documentation index, keeping both the October Voice evidence and main's newer reset implementation guide. The Voice pages and newer architecture pages retain their source contents, with no layout changes.

All nine HTML pages and 535 local links, assets, and anchors passed validation, including the status page links; the PR delta passed git diff --check. No credentials, private conversation IDs, or local machine paths were added.

Fresh browser screenshots and verification notes now show the paired implementation with a real isolated backend and scripted model responses; they are separate from the source-derived drawings and from real-provider audio evidence.

Physical-device audio and fresh real-provider calls are still separate validation work. The linked issues and implementation PRs hold current test/CI results so this dated notebook does not pretend to be live runtime telemetry.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: a430ccfa-124c-4c52-b105-25b59d858667

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}