Connect this agent to a running Guaardvark (self-hosted AI studio) and check what it can do right now. Use first when the user mentions Guaardvark, asks to generate images/video/music/voice locally, run a Film Crew, train a LoRA, launch a swarm, or when a guaardvark MCP tool is missing or a call fails with "not reachable" / "plugin offline".
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Guaardvark runs on the user's own machine. Everything below is local; nothing leaves the box.
${GUAARDVARK_URL:-http://localhost:5000}. The macOS default port is 5055
(AirPlay owns 5000). The web UI is on port 5173 in dev, or the same port as the backend in Docker.curl -sf ${GUAARDVARK_URL:-http://localhost:5000}/api/health || curl -sf http://localhost:5055/api/health || echo "backend not reachable on 5000 or 5055"curl -sf ${GUAARDVARK_URL:-http://localhost:5000}/api/plugins/status || echo "plugin status unavailable"If the backend is not reachable, tell the user to start it from the Guaardvark checkout
(./start.sh, or docker compose up), then retry. Do not try to start it yourself.
guaardvark MCP server exposes RAG, memory,
code intelligence, file processing, web fetch, image/video/animation/music-video/film-crew
generation, songs and speech (generate_music, generate_speech),
get_generation_status for any queued batch or song, outreach drafting and GPU/log
inspection. Generation tools queue by default over MCP and return a batch id. In Claude Code they appear as
mcp__guaardvark__<tool> after python -m backend.mcp install, or as
mcp__plugin_guaardvark_guaardvark__<tool> when the plugin was installed from the marketplace. Install once from the checkout:
python -m backend.mcp install # writes the server entry into every detected client: Claude Code, Codex,
# Cursor, Grok, Antigravity, opencode, Claude Desktop, Zed, Gemini
python -m backend.mcp doctor # self-test + stale-config scan
python -m backend.mcp list-tools # what is exposed right now
curl against the backend URL.
Responses are wrapped as {"success": bool, "data": {...}, "message": str} on most routes; a few
older routes return the bare object. Read data when it is present.GET /api/plugins/status; a generation route answers 503 when
its plugin is not running. Start one with POST /api/plugins/<id>/start where <id> is
comfyui (image + video), audio_foundry (voice, music, FX), upscaling, lora_trainer,
swarm. Only one heavy model owns the GPU at a time; the orchestrator evicts Ollama for video
and vice versa, so a first call after a switch is slow. inspect_gpu (MCP) shows who holds it.GET /api/batch-video/models and GET /api/batch-image/models list every
registry entry with capabilities; check is_ready before naming a model. Nothing downloads
without an explicit Install, so if a model is missing say so and offer
POST /api/batch-video/models/download {"model_id": "..."} (or the image route with
{"model_path": "..."}).data/outputs/ in the checkout and are also served read-only over MCP as
guaardvark://outputs/... resources.| Skill | Use for |
|---|---|
| image | one image, edits, cast characters, batch image runs |
| video | one clip, image-to-video, first/last frame, batch video runs, MiniMax H3 with sound |
| music-video | a song in, a beat-cut music video out, with the approval gate |
| film-crew | screenplay to finished short: writer, casting, storyboards, render, edit |
| voice | narration, TTS, consent-gated voice cloning |
| music | full songs with lyrics, instrumentals, sound effects |
| upscale | 2x to 8K upscaling of images and video |
| cast | Cast Library subjects and LoRA training for consistent characters |
| models | add any Hugging Face model or LoRA from a URL |
| swarm | parallel coding agents in git worktrees from a plan file |
| knowledge | the user's indexed documents, memory, web fetch |
| code | code search, repository map, self-improvement status |
| outreach | supervised social drafts (never posts) |
| ops | GPU, logs, Celery, plugins, Interconnector sync, autoresearch, infographics |
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Build consistent characters, environments and props in Guaardvark's Cast Library and train LoRAs for them locally (reference photos → vision bible → sample plan → approved samples → training). Use when the user wants the same face or object across images, videos, a music video, or a Film Crew production, or asks to "train a LoRA".
中文简介正在准备中。当前显示原文说明。
Code intelligence over repositories Guaardvark has indexed: semantic code search, exact-text search, file listing, PageRank repository map, import dependency graph, reading one class or function, analysis and generated rewrites, plus the System Mapper and self-improvement status. Use when the user asks where something is implemented, how modules depend on each other, or wants an architecture map of an indexed repo.
中文简介正在准备中。当前显示原文说明。
Run Guaardvark's five-role Film Crew locally: screenwriter, casting (Cast Library LoRAs), cinematographer storyboards, keyframe renders, editor. Turns a logline or screenplay into a finished short video with human gates at casting and storyboard. Use when the user says "film this", "make a short film", "storyboard this script", or "start the film crew".
中文简介正在准备中。当前显示原文说明。
Generate or edit images on the user's own GPU through Guaardvark: single images, instruction edits, background cut-outs, inpaint and outpaint, consistent characters from the Cast Library, and batch runs of many prompts. Use when the user asks to create, draw, render, visualize, edit, or batch-generate images locally.
中文简介正在准备中。当前显示原文说明。
Answer from the user's own indexed documents with Guaardvark's local RAG, browse and read indexed files section by section, remember facts across sessions, and process or fetch new material. Use when the user asks what their documents say, wants a source-cited answer from their files, says "remember this", or asks to index a PDF, DOCX, spreadsheet, image or URL.
中文简介正在准备中。当前显示原文说明。
Add any Hugging Face image or video model, checkpoint or LoRA to Guaardvark from a URL, list what is installed, and download registry models on request. Use when the user pastes a huggingface.co link or a .safetensors URL, asks "can it run <model>", or wants a new LoRA or checkpoint available in the Studio.
中文简介正在准备中。当前显示原文说明。