Skip to content
Back to search

llm-ui

Gradio chat UI for KoboldCpp with MCP servers and skills

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/lavenderhorse--llm-ui/badge.svg)](/listings/lavenderhorse--llm-ui)

HTML

<a href="/listings/lavenderhorse--llm-ui"><img src="/api/listings/lavenderhorse--llm-ui/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 32 lines
  1. # llm-ui
  2. Gradio chat UI with MCP servers and Anthropic-style skills. Providers: [KoboldCpp](https://github.com/LostRuins/koboldcpp) (local), OpenAI, DeepSeek, Google Gemini, and Anthropic Claude.
  3. ```bash
  4. uv sync
  5. uv run python -m llmui # http://127.0.0.1:7860
  6. ```
  7. - **Settings**: pick a provider, base URL, API key and model (type any id, or 🔄 refresh the list from the
  8. provider). Keys left blank are read from `OPENAI_API_KEY`, `DEEPSEEK_API_KEY`, `GEMINI_API_KEY`/`GOOGLE_API_KEY`
  9. or `ANTHROPIC_API_KEY` (Claude also works with `ant auth login`). Keys typed in the UI are saved in
  10. `config.json` (gitignored, mode 600). Also: sampling, max tool iterations.
  11. - **MCP**: edit a Claude-Desktop-style `mcp.json` (see `mcp.example.json`). stdio (`command`/`args`/`env`)
  12. and remote (`url`/`headers`; streamable HTTP, or SSE if the URL ends in `/sse` or `"type": "sse"`).
  13. Tools are exposed to the model as `server__tool`; calls auto-run and show as collapsible blocks.
  14. - **Skills**: `./skills/<name>/SKILL.md` (frontmatter `name`, `description`) or `~/.config/llmui/skills/`.
  15. Descriptions go in the system prompt; the model calls `load_skill` / `read_skill_file` on demand.
  16. Provider notes:
  17. - OpenAI / DeepSeek / Gemini use their OpenAI-compatible endpoints. If a model rejects `temperature`/`top_p`
  18. the request is retried without them. DeepSeek's `reasoning_content` is kept and echoed back as it requires.
  19. - Claude uses the native Anthropic SDK (streaming, tool use, thinking blocks replayed unchanged). Sampling
  20. controls are not sent (current Claude models reject them). For models that support it, the server-side
  21. refusal-fallback beta is on by default; untick it in Settings if your account can't use it.
  22. - Switching provider mid-conversation works; provider-specific data (thinking blocks, reasoning) is only
  23. replayed to the provider that produced it.
  24. Tool calling on KoboldCpp uses its native `tools` support (run it with a tool-capable model; `--jinja` is
  25. recommended). If the backend rejects `tools`, the request is retried without them.
  26. Tests: `uv run pytest`.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.