catalog.json

[
  {
    "id": "cherry-studio",
    "name": "Cherry Studio",
    "category": "chat",
    "mode": "guided",
    "description": "Desktop AI chat with custom model providers.",
    "homepage": "https://www.cherry-ai.com/",
    "sources": [
      "https://docs.cherry-ai.com/cherry-studio-wen-dang/en-us/cherry-studio/preview/settings/providers",
      "https://docs.cherry-ai.com/cherry-studio-wen-dang/en-us/pre-basic/providers/zi-ding-yi-fu-wu-shang",
      "https://github.com/CherryHQ/cherry-studio/blob/main/docs/references/provider-model/provider-registry.md",
      "https://github.com/CherryHQ/cherry-studio/blob/v1.8.1/src/renderer/src/store/index.ts",
      "https://github.com/CherryHQ/cherry-studio/blob/v1.8.1/src/renderer/src/store/llm.ts",
      "https://github.com/CherryHQ/cherry-studio/blob/v1.8.1/src/renderer/src/store/assistants.ts"
    ],
    "detection": {
      "commands": [
        "cherry-studio"
      ],
      "home_paths": [],
      "app_names": [
        "Cherry Studio",
        "cherry-studio"
      ],
      "repo_names": [
        "cherry-studio"
      ]
    },
    "instructions": [
      "For compatible Cherry Studio profiles on macOS and Windows, initialize the app, quit it, and select it in CometAPI Connect.",
      "CometAPI Connect adds a provider and selects the model for the default assistant and new-assistant template. Reopen Cherry Studio to chat."
    ],
    "caveats": [
      "This native LevelDB adapter preserves other providers, assistants, settings and conversations. Restore refuses later edits to the provider or configured assistant-model settings.",
      "Automatic setup requires a compatible Redux provider store on macOS or Windows. Linux and image/video workflows are outside this adapter’s scope."
    ]
  },
  {
    "id": "chatbox",
    "name": "Chatbox",
    "category": "chat",
    "mode": "automatic",
    "description": "Desktop chat with automatic setup for compatible, initialized app profiles.",
    "homepage": "https://chatboxai.app/",
    "sources": [
      "https://docs.chatboxai.app/en/guides/providers",
      "https://docs.chatboxai.app/en/guides/providers/import-config",
      "https://github.com/chatboxai/chatbox/blob/main/src/shared/types/settings.ts",
      "https://github.com/chatboxai/chatbox/blob/main/src/shared/types/provider.ts",
      "https://github.com/chatboxai/chatbox/blob/main/src/main/store-node.ts"
    ],
    "detection": {
      "commands": [
        "chatbox"
      ],
      "home_paths": [
        "Library/Application Support/xyz.chatboxapp.app",
        "Library/Application Support/xyz.chatboxapp.ce",
        "AppData/Roaming/xyz.chatboxapp.app",
        "AppData/Roaming/xyz.chatboxapp.ce",
        ".config/xyz.chatboxapp.app",
        ".config/xyz.chatboxapp.ce"
      ],
      "app_names": [
        "Chatbox",
        "Chatbox AI",
        "chatbox"
      ],
      "repo_names": [
        "chatbox"
      ]
    },
    "instructions": [
      "Open Chatbox once to initialize its settings, then quit it completely before automatic setup.",
      "Automatic setup adds a separate CometAPI custom provider and makes the selected model the default for new conversations.",
      "For a custom user-data location, add that profile folder under additional scan paths. Other schemas or ambiguous profiles use guided setup.",
      "For guided setup, add an OpenAI API compatible provider named CometAPI with API Host https://api.cometapi.com and API Path /v1/chat/completions, enter the key, and add/select the exact model ID."
    ],
    "caveats": [
      "Automatic setup checks the initialized provider structure and preserves configuration version fields. Incompatible structures remain guided.",
      "Chatbox must be closed during setup and restore; the helper checks OS process names. Existing conversations retain their selected provider/model, while new conversations use the configured default.",
      "Other providers, chat history, UI preferences, title-generation and embedding settings are preserved. New models use text chat and a 2048-token output allowance; model-specific capabilities are not assumed.",
      "Keys are stored in the app configuration file and protected backups; avoid putting credentials in provider deep-link URLs."
    ]
  },
  {
    "id": "anythingllm",
    "name": "AnythingLLM",
    "category": "chat",
    "mode": "guided",
    "description": "Desktop and self-hosted document chat and agents.",
    "homepage": "https://anythingllm.com/",
    "sources": [
      "https://raw.githubusercontent.com/Mintplex-Labs/anything-llm/master/server/.env.example"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "AnythingLLM",
        "anythingllm",
        "AnythingLLM Desktop"
      ],
      "repo_names": [
        "anything-llm",
        "anythingllm"
      ]
    },
    "instructions": [
      "Open AI Providers or LLM Preference and select CometAPI if the installed version offers it.",
      "Enter your key and choose the CometAPI chat model. Otherwise choose Generic OpenAI with base URL https://api.cometapi.com/v1.",
      "For self-hosted versions with a native CometAPI provider, configure LLM_PROVIDER=cometapi, COMETAPI_LLM_API_KEY, and COMETAPI_LLM_MODEL_PREF.",
      "Check each workspace's model selection after changing the default provider."
    ],
    "caveats": [
      "Native CometAPI support depends on the installed version.",
      "Generic OpenAI also needs a model name and appropriate context limit.",
      "Embedding and image-generation credentials are separate; changing existing embedding models can require rebuilding document indexes."
    ]
  },
  {
    "id": "open-webui",
    "name": "Open WebUI",
    "category": "chat",
    "mode": "guided",
    "description": "Self-hosted chat interface with OpenAI-compatible connections.",
    "homepage": "https://openwebui.com/",
    "sources": [
      "https://docs.openwebui.com/reference/env-configuration/"
    ],
    "detection": {
      "commands": [
        "open-webui"
      ],
      "home_paths": [],
      "app_names": [
        "Open WebUI"
      ],
      "repo_names": [
        "open-webui"
      ]
    },
    "instructions": [
      "Open Admin Settings → Connections and add an OpenAI-compatible connection.",
      "Set base URL to https://api.cometapi.com/v1, enter your CometAPI key, and select or refresh models.",
      "For a new instance, documented environment variables include OPENAI_API_BASE_URL and OPENAI_API_KEY; use the app's supported connection settings for existing instances."
    ],
    "caveats": [
      "Persisted database settings can override environment variables.",
      "Plural OPENAI_API_BASE_URLS and OPENAI_API_KEYS are paired semicolon-separated lists; preserve existing connection order.",
      "Do not disable all persistent configuration just to force a provider setting.",
      "Requires access to the local or remote instance's administrator settings."
    ]
  },
  {
    "id": "librechat",
    "name": "LibreChat",
    "category": "chat",
    "mode": "guided",
    "description": "Self-hosted chat app with documented custom endpoints.",
    "homepage": "https://www.librechat.ai/",
    "sources": [
      "https://www.librechat.ai/docs/configuration/librechat_yaml/object_structure/custom_endpoint",
      "https://raw.githubusercontent.com/danny-avila/LibreChat/main/librechat.example.yaml",
      "https://raw.githubusercontent.com/danny-avila/LibreChat/main/docker-compose.yml"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "LibreChat"
      ],
      "repo_names": [
        "LibreChat",
        "librechat"
      ]
    },
    "instructions": [
      "Use the LibreChat project folder and its loaded librechat.yaml file.",
      "Add a CometAPI item to endpoints.custom with apiKey referencing ${COMETAPI_API_KEY}, baseURL https://api.cometapi.com/v1, and models.default containing your chosen model ID.",
      "Store COMETAPI_API_KEY in the deployment's loaded .env. Preserve the existing YAML version and other endpoints.",
      "Ensure Docker mounts the YAML into /app/librechat.yaml, restart the service, and select CometAPI in the model picker."
    ],
    "caveats": [
      "A .env file or folder name alone does not identify a LibreChat installation; verify librechat.example.yaml and repository identity.",
      "The official base Docker compose binds .env but does not mount librechat.yaml; writing the YAML alone is insufficient for that deployment.",
      "Respect an existing CONFIG_PATH or custom deployment layout.",
      "Configuration affects the deployment's users; title and tool models may have separate settings."
    ]
  },
  {
    "id": "lobechat",
    "name": "LobeChat / LobeHub",
    "category": "chat",
    "mode": "guided",
    "description": "Chat and agent workspace with multiple model providers.",
    "homepage": "https://lobehub.com/",
    "sources": [
      "https://raw.githubusercontent.com/lobehub/lobehub/main/.env.example",
      "https://www.mintlify.com/lobehub/lobehub/integrations/ai-providers"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "LobeHub",
        "LobeChat",
        "Lobe Chat"
      ],
      "repo_names": [
        "lobe-chat",
        "lobehub",
        "lobechat"
      ]
    },
    "instructions": [
      "Open model-provider settings and use the native Comet API provider when available, or an OpenAI provider with a custom URL.",
      "Enter your CometAPI key, set the compatible provider base URL to https://api.cometapi.com/v1, and select a supported model.",
      "For compatible self-hosted installations, configure OPENAI_API_KEY, OPENAI_PROXY_URL, and OPENAI_MODEL_LIST in the environment loaded by the service."
    ],
    "caveats": [
      "Server defaults and user-specific settings can differ; a repository environment change may not update existing user provider settings.",
      "Confirm the service actually loads the edited environment file and restart it.",
      "Do not patch browser or database provider state using guessed storage keys."
    ]
  },
  {
    "id": "nextchat",
    "name": "NextChat",
    "category": "chat",
    "mode": "guided",
    "description": "Desktop and self-hosted AI chat, formerly ChatGPT-Next-Web.",
    "homepage": "https://nextchat.dev/",
    "sources": [
      "https://raw.githubusercontent.com/ChatGPTNextWeb/NextChat/main/.env.template",
      "https://raw.githubusercontent.com/ChatGPTNextWeb/NextChat/main/app/api/common.ts",
      "https://raw.githubusercontent.com/ChatGPTNextWeb/NextChat/main/package.json",
      "https://raw.githubusercontent.com/ChatGPTNextWeb/NextChat/main/docker-compose.yml",
      "https://raw.githubusercontent.com/ChatGPTNextWeb/NextChat/main/app/utils/model.ts",
      "https://raw.githubusercontent.com/ChatGPTNextWeb/NextChat/main/app/config/server.ts"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "NextChat",
        "ChatGPT Next Web",
        "ChatGPT-Next-Web"
      ],
      "repo_names": [
        "NextChat",
        "nextchat",
        "ChatGPT-Next-Web",
        "ChatGPTNextWeb",
        "chatgpt-next-web"
      ]
    },
    "instructions": [
      "For a source-run project, use the loaded .env.local with OPENAI_API_KEY and BASE_URL=https://api.cometapi.com.",
      "Merge +model-id@openai into CUSTOM_MODELS and set DEFAULT_MODEL=model-id@openai for the chosen CometAPI model, preserving other entries.",
      "Docker deployments use their loaded environment and must forward the model variables to the service.",
      "Restart the source service. For the separately installed desktop app, configure the custom OpenAI URL and key in the app's settings."
    ],
    "caveats": [
      "BASE_URL is the API root without /v1 because NextChat appends the versioned request path.",
      "Source setup requires a nextchat or chatgpt-next-web package with app/api/common.ts.",
      "The official Docker compose forwards key and base URL but omits DEFAULT_MODEL and CUSTOM_MODELS.",
      "Preserve existing access CODE and custom model entries; an installed desktop app is not configured by writing a repository .env.",
      "Use the @openai provider suffix to avoid selecting a preexisting native Claude or Gemini route.",
      "Check your NextChat version’s logging settings: some versions log API keys. CometAPI Connect cannot redact another application’s logs."
    ]
  },
  {
    "id": "jan",
    "name": "Jan",
    "category": "chat",
    "mode": "guided",
    "description": "Desktop local/remote chat with a separate Agent CLI.",
    "homepage": "https://www.jan.ai/",
    "sources": [
      "https://www.jan.ai/docs/desktop/remote-models/custom-endpoint",
      "https://www.jan.ai/docs/agent/providers",
      "https://www.jan.ai/docs/agent/cli"
    ],
    "detection": {
      "commands": [
        "jan"
      ],
      "home_paths": [
        ".jan/config.toml"
      ],
      "app_names": [
        "Jan",
        "jan"
      ],
      "repo_names": [
        "jan"
      ]
    },
    "instructions": [
      "In Jan Desktop's provider settings, add a custom provider and choose the OpenAI-compatible API format.",
      "Use name CometAPI, base URL https://api.cometapi.com/v1, and your CometAPI key.",
      "Fetch or add your chosen model ID and configure its capabilities.",
      "For custom GPT-5/Gemini chat providers, clear optional sampling overrides so Jan uses model defaults.",
      "Jan Agent users can use the documented jan config set provider workflow; its settings are distinct from Desktop."
    ],
    "caveats": [
      "The CLI's ~/.jan/config.toml is not a guaranteed desktop configuration file.",
      "Custom model capabilities are not fully discovered from a model-list endpoint.",
      "Project agent.toml and environment or command-line settings can override Agent defaults."
    ]
  },
  {
    "id": "lm-studio",
    "name": "LM Studio",
    "category": "chat",
    "mode": "guided",
    "description": "Desktop chat with local models and an official OpenAI-compatible remote generator.",
    "homepage": "https://lmstudio.ai/",
    "sources": [
      "https://lmstudio.ai/lmstudio/openai-compat-endpoint",
      "https://lmstudio.ai/lmstudio/openai-compat-endpoint/files/src/config.ts"
    ],
    "detection": {
      "commands": [
        "lms"
      ],
      "home_paths": [],
      "app_names": [
        "LM Studio",
        "LMStudio",
        "lm-studio"
      ],
      "repo_names": []
    },
    "instructions": [
      "Initialize LM Studio and accept its first-run terms, then close it before setup or restore. Automatic setup requires a compatible macOS or Windows profile.",
      "CometAPI Connect configures the official OpenAI-compatible generator and selected chat. If needed, it installs the bundled plugin and creates an initial CometAPI chat."
    ],
    "caveats": [
      "The remote generator supports a fixed model list. The standard Connect model selection uses GPT-4.1 mini; custom models must be supported by the plugin.",
      "The plugin stores its API key in a private plaintext configuration file. Restore preserves chat messages; installed plugin files remain."
    ]
  },
  {
    "id": "sillytavern",
    "name": "SillyTavern",
    "category": "chat",
    "mode": "guided",
    "description": "Character and general chat frontend with custom completion APIs.",
    "homepage": "https://sillytavern.app/",
    "sources": [
      "https://docs.sillytavern.app/usage/api-connections/openai/"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "SillyTavern"
      ],
      "repo_names": [
        "SillyTavern",
        "sillytavern"
      ]
    },
    "instructions": [
      "Open API Connections and select Chat Completion.",
      "Set Chat Completion Source to Custom (OpenAI-compatible), endpoint https://api.cometapi.com/v1, and your CometAPI key.",
      "Fetch or enter the chosen CometAPI model ID and connect."
    ],
    "caveats": [
      "Do not append /chat/completions to the custom endpoint field.",
      "Profiles, secrets, image generation, and extensions have separate settings.",
      "A test message makes a billable model request."
    ]
  },
  {
    "id": "comfyui",
    "name": "ComfyUI",
    "category": "media",
    "mode": "guided",
    "description": "Node-based image and video workflows with dedicated CometAPI cloud nodes.",
    "homepage": "https://www.comfy.org/",
    "sources": [
      "https://docs.comfy.org/custom-nodes/overview"
    ],
    "detection": {
      "commands": [
        "comfy"
      ],
      "home_paths": [],
      "app_names": [
        "ComfyUI",
        "Comfy"
      ],
      "repo_names": [
        "ComfyUI",
        "comfyui",
        "ComfyUI_windows_portable"
      ]
    },
    "instructions": [
      "Select a recognized ComfyUI source installation to install the CometAPI Image and CometAPI Video nodes.",
      "Restart ComfyUI, add a CometAPI node, and select a model and prompt. Generation uses your CometAPI account and may incur charges.",
      "The CometAPI nodes read credentials from a private configuration file outside workflow graphs. Keep that file private."
    ],
    "caveats": [
      "The bundled nodes require ComfyUI’s public node API, comfy_api.latest. Other nodes keep their own provider and credential settings.",
      "Official Comfy partner nodes use Comfy account credentials and credits; COMFY_API_KEY is not a CometAPI key. There is no global setting that redirects every node.",
      "API-key node inputs can appear in workflow JSON, queue history and saved image workflow metadata."
    ]
  },
  {
    "id": "dify",
    "name": "Dify",
    "category": "workflow",
    "mode": "guided",
    "description": "Self-hosted AI application and workflow platform.",
    "homepage": "https://dify.ai/",
    "sources": [
      "https://github.com/langgenius/dify-official-plugins/blob/main/models/openai_api_compatible/provider/openai_api_compatible.yaml",
      "https://github.com/langgenius/dify-official-plugins/blob/main/models/openai_api_compatible/models/llm/llm.py"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "Dify"
      ],
      "repo_names": [
        "dify"
      ]
    },
    "instructions": [
      "Open the Dify workspace's model-provider settings and add the official OpenAI-API-compatible model plugin if needed.",
      "Set endpoint_url to https://api.cometapi.com/v1, enter your CometAPI key, and add the exact model ID with appropriate capabilities.",
      "Choose the new model in the application's LLM nodes and default-model settings."
    ],
    "caveats": [
      "Dify's server .env contains infrastructure settings; provider credentials are workspace/plugin-managed.",
      "Requires access to the workspace's model settings.",
      "Embedding, image, and video tool plugins have separate settings and API formats."
    ]
  },
  {
    "id": "n8n",
    "name": "n8n",
    "category": "workflow",
    "mode": "guided",
    "description": "Workflow automation with OpenAI-compatible credential settings.",
    "homepage": "https://n8n.io/",
    "sources": [
      "https://raw.githubusercontent.com/n8n-io/n8n/master/packages/nodes-base/credentials/OpenAiApi.credentials.ts",
      "https://docs.n8n.io/integrations/builtin/credentials/openai"
    ],
    "detection": {
      "commands": [
        "n8n"
      ],
      "home_paths": [
        ".n8n"
      ],
      "app_names": [
        "n8n"
      ],
      "repo_names": [
        "n8n"
      ]
    },
    "instructions": [
      "Create a new OpenAI credential named CometAPI in your n8n instance.",
      "Enter your CometAPI key and set Base URL to https://api.cometapi.com/v1; leave Organization ID empty unless the chosen gateway specifically requires it.",
      "Attach that credential to selected OpenAI Chat Model or compatible OpenAI nodes and enter the chosen model ID."
    ],
    "caveats": [
      "Credentials are stored by n8n; do not directly rewrite its encrypted database.",
      "A credential does not automatically switch every existing workflow.",
      "Responses, image, video, and embedding nodes require their corresponding CometAPI routes and models."
    ]
  },
  {
    "id": "flowise",
    "name": "Flowise",
    "category": "workflow",
    "mode": "guided",
    "description": "Visual AI workflow builder with a configurable OpenAI chat node.",
    "homepage": "https://flowiseai.com/",
    "sources": [
      "https://docs.flowiseai.com/configuration/environment-variables",
      "https://raw.githubusercontent.com/FlowiseAI/Flowise/flowise%403.1.4/packages/components/nodes/chatmodels/ChatOpenAI/ChatOpenAI.ts",
      "https://raw.githubusercontent.com/FlowiseAI/Flowise/flowise%403.1.4/packages/server/src/utils/index.ts"
    ],
    "detection": {
      "commands": [
        "flowise"
      ],
      "home_paths": [],
      "app_names": [
        "Flowise"
      ],
      "repo_names": [
        "Flowise",
        "flowise"
      ]
    },
    "instructions": [
      "Create an OpenAI credential containing your CometAPI key.",
      "In each selected OpenAI chat node, choose that credential and set Additional Parameters → Base Path to https://api.cometapi.com/v1.",
      "Set Model Name to your chosen CometAPI model and save the flow."
    ],
    "caveats": [
      "Automatic setup supports existing ChatOpenAI nodes with compatible credential and endpoint fields, local SQLite, one active user and one workspace.",
      "The helper updates node base paths and the bound encrypted OpenAI credentials together; it preserves selected models and workflow connections.",
      "Stop Flowise before applying and restart it afterwards. Unsupported shared credentials, custom headers, external databases and secret stores need separate setup."
    ]
  },
  {
    "id": "openmaic",
    "name": "OpenMAIC",
    "category": "workflow",
    "mode": "guided",
    "description": "Interactive multi-agent teaching application.",
    "homepage": "https://github.com/THU-MAIC/OpenMAIC",
    "sources": [
      "https://github.com/THU-MAIC/OpenMAIC",
      "https://raw.githubusercontent.com/THU-MAIC/OpenMAIC/main/.env.example"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [
        "OpenMAIC"
      ],
      "repo_names": [
        "OpenMAIC",
        "openmaic"
      ]
    },
    "instructions": [
      "In a compatible source installation, use .env.local as documented by .env.example.",
      "Set OPENAI_API_KEY to your CometAPI key, OPENAI_BASE_URL to https://api.cometapi.com/v1, and OPENAI_MODELS to selected CometAPI model IDs.",
      "Restart the app and select the configured provider/model in the classroom settings."
    ],
    "caveats": [
      "Image, speech, and video providers use separate IMAGE_*, TTS_*, ASR_*, and VIDEO_* variables.",
      "Do not reuse the chat base URL for native media providers without checking their route format."
    ]
  },
  {
    "id": "openviking",
    "name": "OpenViking",
    "category": "framework",
    "mode": "guided",
    "description": "Context database for AI agents.",
    "homepage": "https://github.com/volcengine/OpenViking",
    "sources": [
      "https://github.com/volcengine/OpenViking",
      "https://raw.githubusercontent.com/volcengine/OpenViking/main/examples/ov.conf.example"
    ],
    "detection": {
      "commands": [
        "openviking-server"
      ],
      "home_paths": [
        ".openviking"
      ],
      "app_names": [],
      "repo_names": [
        "OpenViking",
        "openviking"
      ]
    },
    "instructions": [
      "Open the selected deployment's OpenViking configuration and identify its vision/language, embedding, and reranking providers separately.",
      "Use each provider's documented compatible API mode, supported CometAPI model, key, and exact endpoint.",
      "Validate model capabilities and embedding dimensions before updating the active deployment."
    ],
    "caveats": [
      "No universal all-provider endpoint rewrite is provided.",
      "Changing an embedding model for indexed data may require reindexing.",
      "Native provider paths and OpenAI-compatible paths differ.",
      "The main service API key is distinct from the upstream CometAPI model key."
    ]
  },
  {
    "id": "semantica",
    "name": "Semantica",
    "category": "framework",
    "mode": "guided",
    "description": "Knowledge and context framework for AI applications.",
    "homepage": "https://github.com/semantica-agi/semantica",
    "sources": [
      "https://github.com/semantica-agi/semantica",
      "https://github.com/semantica-agi/semantica/blob/main/docs/index.md"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [],
      "repo_names": [
        "semantica"
      ]
    },
    "instructions": [
      "Inspect the selected application's actual LLM provider construction and extraction configuration.",
      "Use a documented OpenAI-compatible or LiteLLM integration with your CometAPI key and matching model route.",
      "Validate extraction behavior before changing production processing."
    ],
    "caveats": [
      "Framework installation does not imply a single configuration file or one active provider.",
      "Automatic setup supports recognized Python OpenAI constructors; dynamic or custom provider configurations require manual setup.",
      "Embedding, extraction, and generation can use distinct providers."
    ]
  },
  {
    "id": "litellm",
    "name": "LiteLLM",
    "category": "framework",
    "mode": "guided",
    "description": "Model SDK and proxy with a native CometAPI provider.",
    "homepage": "https://www.litellm.ai/",
    "sources": [
      "https://docs.litellm.ai/docs/providers/cometapi",
      "https://docs.litellm.ai/docs/providers/openai_compatible"
    ],
    "detection": {
      "commands": [
        "litellm"
      ],
      "home_paths": [],
      "app_names": [],
      "repo_names": [
        "litellm"
      ]
    },
    "instructions": [
      "For the native provider, supply COMETAPI_KEY to the application process and choose cometapi/<your-model-id>.",
      "Alternatively use an OpenAI-compatible model route openai/<your-model-id> with api_base https://api.cometapi.com/v1 and the CometAPI key.",
      "For a proxy deployment, update the selected model_list entry in the actual loaded configuration and restart the proxy."
    ],
    "caveats": [
      "LiteLLM provider prefixes select routing and are not necessarily part of the upstream model ID.",
      "Preserve virtual keys, access policy, and unrelated routes.",
      "An installed SDK has no universal per-user config that changes every application."
    ]
  },
  {
    "id": "langchain",
    "name": "LangChain",
    "category": "framework",
    "mode": "guided",
    "description": "Application framework with configurable OpenAI-compatible clients.",
    "homepage": "https://www.langchain.com/",
    "sources": [
      "https://docs.langchain.com/oss/python/integrations/chat/openai"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [],
      "repo_names": [
        "langchain",
        "langchainjs"
      ]
    },
    "instructions": [
      "In the selected application, configure ChatOpenAI with the chosen CometAPI model, base_url https://api.cometapi.com/v1, and your key.",
      "When the application relies on environment defaults, current Python docs support OPENAI_API_BASE or OPENAI_BASE_URL and OPENAI_API_KEY.",
      "Verify the application uses this client rather than a separately configured provider."
    ],
    "caveats": [
      "Explicit constructor settings take precedence over environment variables.",
      "A framework dependency is not evidence of a safe global setting.",
      "Tool, streaming-usage, and provider-native features need model-specific validation."
    ]
  },
  {
    "id": "openai-sdk",
    "name": "OpenAI SDK projects",
    "category": "framework",
    "mode": "guided",
    "description": "Custom applications using the OpenAI Python or JavaScript SDK.",
    "homepage": "https://apidoc.cometapi.com/",
    "sources": [
      "https://apidoc.cometapi.com/"
    ],
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [],
      "repo_names": []
    },
    "instructions": [
      "Set the application's OpenAI client base_url (Python) or baseURL (JavaScript) to https://api.cometapi.com/v1.",
      "Pass your CometAPI key through the application's secret configuration and use an exact CometAPI model ID.",
      "Use the API surface required by the application and chosen model."
    ],
    "caveats": [
      "An SDK import does not identify a universal config file; the application's own configuration must be checked.",
      "This helper does not rewrite arbitrary application code.",
      "Chat, Responses, images, audio, and video are distinct API surfaces."
    ]
  },
  {
    "id": "automatic1111",
    "name": "AUTOMATIC1111 Stable Diffusion WebUI",
    "description": "Stable Diffusion WebUI with a separate CometAPI cloud generation extension.",
    "homepage": "https://github.com/AUTOMATIC1111/stable-diffusion-webui",
    "sources": [
      "https://github.com/AUTOMATIC1111/stable-diffusion-webui"
    ],
    "instructions": [
      "Select a recognized AUTOMATIC1111 source installation to add the CometAPI cloud tab, then restart WebUI.",
      "Choose a model and enter a prompt in the CometAPI cloud tab. Cloud generation uses your CometAPI account and may incur charges.",
      "The extension stores its API key in a private configuration file; local txt2img and img2img settings are preserved."
    ],
    "caveats": [
      "The CometAPI tab is separate from local checkpoint generation and requires a compatible Gradio extension interface.",
      "The WebUI's own API serves its local generation engine; it is not a generic upstream-provider setting.",
      "A repository candidate can be checked for webui.py, launch.py, and modules/; its folder name alone is not proof."
    ],
    "category": "media",
    "mode": "guided",
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [],
      "repo_names": [
        "stable-diffusion-webui"
      ]
    }
  },
  {
    "id": "invokeai",
    "name": "InvokeAI",
    "description": "Image canvas with local models and configurable external image providers.",
    "homepage": "https://github.com/invoke-ai/InvokeAI",
    "sources": [
      "https://github.com/invoke-ai/InvokeAI",
      "https://github.com/invoke-ai/InvokeAI/blob/ef832d1aa57641dd7d0c262f4d40fe8b7b420489/invokeai/app/services/config/config_default.py",
      "https://github.com/invoke-ai/InvokeAI/blob/ef832d1aa57641dd7d0c262f4d40fe8b7b420489/invokeai/app/services/external_generation/providers/openai.py"
    ],
    "instructions": [
      "Select the InvokeAI data directory containing invokeai.yaml. CometAPI Connect writes external_openai_api_key and external_openai_base_url to the private api_keys.yaml file.",
      "Restart InvokeAI and select an OpenAI-backed external image model supported by your installed version and CometAPI account.",
      "Check environment variables and app-managed API-key settings if the key is ignored; these can take precedence over invokeai.yaml."
    ],
    "caveats": [
      "Requires an InvokeAI version with the external OpenAI image provider and compatible api_keys.yaml settings. Canvas features and model capabilities depend on the installed version.",
      "Local checkpoint inference cannot be redirected to CometAPI by changing an API endpoint.",
      "These settings replace the external OpenAI image provider credentials and endpoint. Do not replace local model paths or unrelated provider credentials.",
      "A source checkout has pyproject.toml project identity InvokeAI and an invokeai/ directory; launcher-managed installations can live elsewhere."
    ],
    "category": "media",
    "mode": "guided",
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [],
      "repo_names": [
        "InvokeAI"
      ]
    }
  },
  {
    "id": "fooocus",
    "name": "Fooocus",
    "description": "Local image generation with an optional CometAPI cloud panel.",
    "homepage": "https://github.com/lllyasviel/Fooocus",
    "sources": [
      "https://github.com/lllyasviel/Fooocus"
    ],
    "instructions": [
      "Select a compatible Fooocus source installation to add the CometAPI image/video panel, then restart Fooocus.",
      "The cloud panel requires a compatible Gradio launch structure. CometAPI Connect previews and backs up the source change before applying it.",
      "Keep the existing checkpoint paths and local generation configuration."
    ],
    "caveats": [
      "config.txt configures local model paths and generation defaults, not an OpenAI-compatible endpoint.",
      "Automatic setup requires a recognized Fooocus source installation with one compatible top-level Gradio launch statement in webui.py."
    ],
    "category": "media",
    "mode": "guided",
    "detection": {
      "commands": [],
      "home_paths": [],
      "app_names": [],
      "repo_names": [
        "Fooocus"
      ]
    }
  }
]