Feat: Added kimi support

[buriy]

No description provided.

[vakovalskii]

Conflicts after merging recent PRs (#165 Kilo, #177 Linux focus, #172 WSL). Please rebase on latest main.

[vakovalskii]

Still conflicts. Please rebase on latest main.

[buriy]

@vakovalskii rebased

[buriy]

also pls see/check active sessions search speedup w/ caching : using /proc instead of lsof + 5 seconds caching

[NovakPAai]

Code Review

⚠️ Installation command may be incorrect — README and app.js both show:

pip install kimi-cli

But Kimi CLI (MoonshotAI) appears to be a Python package, so pip is correct per the GitHub repo. However, the app.js install entry also adds uv tool install kimi-cli as an alternative — that's a good addition. Please double-check the package name on PyPI to make sure it matches exactly.

⚠️ Agent count mismatch across docs — the PR updates different files with different counts:

• ARCHITECTURE.md: "7 agents" (adds Kimi)
• docs/README_RU.md: "6 агентов"
• docs/README_ZH.md: "6 个代理"
• README.md: table header unchanged

All four files should agree on the same count.

⚠️ ARCHITECTURE.md diagram still shows "reads from 6 locations" but the PR adds ~/.kimi/sessions/ as a 7th location. The count in the ASCII diagram needs to be updated too.

ℹ️ The kimi -r <sessionId> resume command — please confirm this is the correct CLI flag for Kimi Code. The similar Codex uses codex resume, Claude uses --resume. If the flag is wrong, the "Resume" button in the UI and the terminal integration will silently do nothing useful.

[vakovalskii]

Thanks for the contribution! Closing for now — we're keeping the agent roster focused (Claude Code, Codex, Cursor, OpenCode, Kiro + adding Qwen and GitHub Copilot Chat this round). Kimi is on the radar but we don't want to grow past ~7-8 agents without a pluggable agent API. Happy to revisit if there's strong user demand — feel free to reopen or discuss in an issue.