Stop faking unavailable chat demos

[konard]

Fixes #10.

Summary

• Replaced fake fallback transcripts for unavailable chat integrations with explicit renderer capability tiers. Non-installed packages and credential-gated hosted SDKs now render a source-only unavailable state and disable the composer.
• Expanded the comparison to 18 React chat UI profiles with refreshed npm/GitHub evidence, package source snippets, and score caps for source-only or hosted integrations.
• Sorted the gallery and comparison by computed score instead of pinning the in-house chat first; Deep Chat now ranks above the built-in chat because its score is higher.
• Changed reply behavior so new messages only render reply blocks after the user selects a reply target.
• Added the issue I don't like the idea, we show integration demos we can't actually show #10 case study, raw issue/PR/CI/package data, local verification logs, and a changeset.

Reproduction

Before this change, selecting SDKs such as assistant-ui, Stream, Sendbird, CometChat, TalkJS, or other unavailable package demos showed a local fake transcript/composer path.

After this change, selecting those profiles shows the real integration source, a clear local-unavailable reason, data-message-count="0", and a disabled composer.

Verification

• npm test — 52/52 unit tests pass
• npm run demo:build — 18 profile pages build successfully
• npm run test:e2e — 12/12 browser tests pass and regenerate the review screenshot
• npm run check — eslint, prettier, and jscpd clean
• git diff --check

Screenshot

Case study

• docs/case-studies/issue-10/README.md
• Raw issue, PR, CI, npm, GitHub, and local verification data are stored under docs/case-studies/issue-10/data/

[konard]

Working session summary

Done. PR #11 is ready for review: #11

Implemented the issue #10 fix set:

• Removed fake fallback chat transcripts for unavailable integrations.
• Added source-only unavailable states with disabled composer.
• Added renderer capability tiers and score caps.
• Expanded to 18 React chat UI profiles with captured npm/GitHub evidence.
• Sorted catalog/comparison by computed score; own chat is no longer pinned first.
• Required explicit reply target selection before rendering new reply blocks.
• Added the issue I don't like the idea, we show integration demos we can't actually show #10 case study, raw data/logs, changeset, and screenshot.

Verification passed locally:

• npm test
• npm run demo:build
• npm run test:e2e
• npm run check
• git diff --check

Pushed commit 81bd818. GitHub CI is green on that commit: Broken Link Checker, Chat demo pages, and Checks and release all passed. Working tree is clean.

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Model: GPT-5.5
• Provider: OpenAI
• Public pricing estimate: $36.581049

📊 Context and tokens usage:

• 810.3K / 1.1M (77%) input tokens, 83.8K / 128K (65%) output tokens

Total: (810.3K + 24.7M cached) input tokens, 83.8K output tokens, $36.581049 cost

🤖 Models used:

• Tool: OpenAI Codex
• Requested: gpt-5.5
• Model: GPT-5.5 (gpt-5.5)

📎 Log file uploaded as Repository (2 chunks) (145081KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

🎉 Auto-merged

This pull request has been automatically merged by hive-mind.

• All CI checks have passed

---

Auto-merged by hive-mind with --auto-merge flag