All authors
Claude Skills by quangdang46
github.com/quangdang4616 skills0 installs6 views
- Openproxy ChatSend chat completions through OpenProxy's OpenAI-compatible API — POST /v1/chat/completions, the `provider/model` addressing convention, combos as bare model names, `stream: true` SSE framing, tools / tool_calls passthrough, and reading combo fallback out of a response. Use whenever the user asks to call a model through the proxy, pick a model from GET /v1/models, wire an agent loop to the router, or debug a chat call that 404s with `No active credentials for provider: …`, 400s with `Unknown ...Votes: 0GitHub stars: 26
- Openproxy Cli ToolsWire AI coding CLIs into the local OpenProxy router at 127.0.0.1:4623 — Claude Code, Codex CLI, OpenCode, Cline, Roo, Kilo Code, Continue, Copilot, Cursor, Factory Droid, OpenClaw, Hermes, jcode, Grok Build, Devin CLI, Claude Cowork. Covers /api/cli-tools and the per-tool *-settings routes, which model each tool should use, and how to verify a tool is really talking to the proxy. Use when a CLI should route through OpenProxy, when a tool's settings file needs the proxy base URL or key, or whe...Votes: 0GitHub stars: 26
- Openproxy CombosCreate, edit, test, and dispatch OpenProxy combos — ordered fallback chains across providers and models. Use whenever the user asks to build a combo, set up failover or round-robin across models, list or delete combos, check why a combo 400s or 503s, or use a combo as the `model` field in /v1/chat/completions.Votes: 0GitHub stars: 26
- Openproxy EmbeddingsGenerate embedding vectors through OpenProxy's `POST /v1/embeddings` — the `input` / `dimensions` / `encoding_format` contract, the built-in embedding-provider adapters, the `custom-embedding` providerNode type for self-hosted and OpenAI-compatible embedding servers, and the checks that prove an embedding provider is actually wired. Use whenever the user asks to embed text, set up RAG / semantic search / vector search, add an embedding model, or debug an embeddings call that 400s, 401s, or re...Votes: 0GitHub stars: 26
- Openproxy ImageGenerate images through the OpenProxy router on 127.0.0.1:4623 — POST /v1/images/generations (and /v1/images/edits), which providers serve them, how the model string is addressed, and how to add a new image endpoint with a custom openai-compatible provider node. Use whenever the user wants to generate, edit, or text-to-image an image through openproxy, wants to know which image models are available, or wants to wire a new image vendor into the router.Votes: 0GitHub stars: 26
- Openproxy ProvidersManage OpenProxy providers from the CLI or the REST API — add/edit/delete provider connections, register custom provider nodes (vLLM, Ollama, any OpenAI/Anthropic-compatible endpoint), set priority, run connectivity probes, and import a provider's model catalog. Use whenever the user asks to add a provider key, connect an OAuth subscription (Claude Code, Codex, Copilot, Cursor), wire up a self-hosted or third-party endpoint, troubleshoot a failing provider connection, or reorder providers for...Votes: 0GitHub stars: 26
- Openproxy RtkOperate OpenProxy's RTK request-compression stack — the token saver that rewrites tool-call results before they reach a model, the `x-9router-token-saver` per-request opt-out header, the `rtkEnabled` / `cavemanEnabled` / `ponytailEnabled` / `headroomEnabled` settings toggles, and the `GET /api/compression/stats` endpoint that reports real savings. Use whenever the user asks to reduce input tokens, cut cost or latency, measure compression savings, debug why tool output is not being compressed,...Votes: 0GitHub stars: 26
- Openproxy SttTranscribe audio through OpenProxy's OpenAI-compatible POST /v1/audio/transcriptions endpoint — multipart and base64-JSON request shapes, the seven STT providers it dispatches to, per-provider credential fallback, and the `openproxy media stt transcribe` CLI. Use whenever the user needs speech-to-text, Whisper, Deepgram, AssemblyAI, or Gemini audio transcription through the local router on 127.0.0.1:4623.Votes: 0GitHub stars: 26
- Openproxy TtsGenerate speech through OpenProxy's OpenAI-compatible text-to-speech API — POST /v1/audio/speech, the ?response_format=json envelope, how the model/voice string addresses each provider (OpenAI, Gemini, Fish Audio, ElevenLabs, OpenRouter, MiniMax, AWS Polly, self-hosted), combo fallback, voice listing, and the openproxy media tts CLI. Use whenever the user asks to synthesize audio, convert text to speech, pick a TTS voice, or debug a TTS provider returning the wrong container or the wrong voice.Votes: 0GitHub stars: 26
- Openproxy VideoGenerate video through OpenProxy's async `/v1/videos/*` routes — submit a text-to-video or image-to-video job, poll the returned `request_id` to completion, and download the MP4. Covers the xAI / OpenRouter / Vertex video providers, the per-account cooldown and rotation rules, and the `openproxy media video` CLI that drives the whole flow. Use whenever the user asks to create a video, animate an image, wire a video model into OpenProxy, or debug a stuck or failed video job.Votes: 0GitHub stars: 26
- Openproxy Web FetchFetch a web page through OpenProxy and get back extracted text. Use whenever the user wants to read a URL through the proxy, route page fetches to Firecrawl / Jina Reader / Tavily / Exa / Ollama, fall back across fetch accounts, or debug `POST /v1/web/fetch` — including the `/v1/fetch` path that the model kind map advertises but the router does not serve.Votes: 0GitHub stars: 26
- Openproxy Web SearchRun web search and web fetch through OpenProxy's local router on 127.0.0.1:4623 — pick between the raw `/v1/search` endpoint and the chat-completion-shaped `/v1/chat/search` endpoint, configure the 14 dedicated search providers and the 10 search-via-chat providers, and read back the response envelopes. Use whenever the user asks to search the web through openproxy, wire a search tool into the router, add a search credential (tavily, serper, exa, brave, …), make a chat model do grounded web se...Votes: 0GitHub stars: 26
- OpenproxyInstall, initialize, and operate OpenProxy from the CLI — either guiding a human through setup or driving it fully autonomously as an agent. Use whenever the user asks to install the openproxy binary, start the local AI router on 127.0.0.1:4623, configure providers / combos / keys, or wire an AI coding CLI (Claude Code, Codex, Cursor, Cline, OpenClaw, Copilot, …) into OpenProxy.Votes: 0GitHub stars: 26
- Loop Test FixEnd-to-end self-healing loop for openproxy — build+install locally, wire up a free opencode model, point codex at openproxy, smoke-test via `codex exec`, sweep every CLI subcommand for real coverage, run the repo's unit/integration tests, and if anything fails go back to research → fix → test → rebuild → reinstall until everything passes. Use when the user asks to "loop test fix openproxy", validate a local openproxy build end-to-end with codex/opencode, exercise "all CLI features/flags", or ...Votes: 0GitHub stars: 26
- Lessonsskills: - ms - agents - mcp --- AI agents like Claude Code are powerful, but they don't know your specific project conventions or unique workflows out of the box. **Skills** are reusable prompt templates that give agents new capabilities. They are like "plugins" for your AI. We use **meta_skill (ms)** to manage these skills locally. It's a local-first knowledge management tool with hybrid semantic search (BM25 + hash embeddings) and Git-backed audit trails. Your skills live in your project re...Votes: 0GitHub stars: 5
- ToolsLocal-first knowledge management with hybrid semantic search for AI coding agents.Votes: 0GitHub stars: 5