fix(ai): accept local session chat replies

[scotej]

What changed and why

Fixes #212.

The attached Windows and macOS diagnostics show the in-session AI requests reaching the pinned llama-server, generating tokens, and returning HTTP 200. The failure was at the client contract: StudyVis requested generic JSON, then rejected any response that was not exactly a one-key reply_text object.

This change supplies llama.cpp b9095's supported json_object + schema contract, bounds the generated reply, and defensively accepts reply_text while ignoring non-executable extra fields. It also keeps fetch and response-body reads inside the request deadline and records privacy-safe outcome/shape diagnostics without prompts, topics, or replies.

Manual test

• npm run tauri dev launched and the changed surface behaves as described
  — not run: this environment has no Rust/desktop model runtime; the exact request, response, timeout, and cancellation paths are covered by unit tests
• Checked in both themes and at reduced-motion — n-a: no rendered UI or motion changed; the existing Storybook interaction and accessibility suite passed

Compatibility surfaces

None. This does not change persistence, migrations, peer messages, identity derivation, settings, or release artifacts.

Gates

• npm run build && npm run lint && npm run test — 1,252 tests passed
• npm run check-tokens && npm run check-strings && npm run check-contrast
• npm run check-migrations && npm run check-stories
• npm run build-storybook && npm run check-a11y — 335 tests passed
• cd src-tauri && cargo fmt --check && cargo clippy && cargo test
  — no Rust changes / CI is the first compiler for this box
• cd src-tauri && cargo deny check — no dependency changes; CI will run the supply-chain gate

Also verified Prettier, git diff --check, and version lockstep locally. The old diagnostics intentionally omitted model response bodies, so they prove the healthy transport/engine path but not the exact rejected extra key; the schema mismatch is the high-confidence code-level cause.

Merge style

• Squash
• Merge commit — because:

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: d130b66c-3067-4f2d-aa48-789205d92895

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}