FakeSeek

FakeSeek is a browser-first AI workspace for chat, search-assisted reasoning, image generation, and fine-tuning dataset workflows. The frontend stores conversations and presets locally, while the Cloudflare Worker provides optional hosted API routes for search, MCP tools, OpenAI-compatible proxying, logs, and long-running research workflows.

Features

• Local chat history backed by IndexedDB
• Model provider presets with API key, base URL, default model, API mode, tool provider, temperature, image quality, and fine-tuning endpoint settings
• OpenAI Responses API mode and Chat Completions compatibility mode
• Optional OpenAI built-in tools for web search, code interpreter, and image generation
• Default local tool mode for Google search and remote Python execution
• Image search and image generation/editing from the chat input
• Fine-tuning dataset editing, evaluation, and job management UI
• PWA-style static asset caching with a settings action to refresh deployed assets
• Cloudflare Worker routes for /api, /logs, /mcp, and OpenAI-compatible /api/v1
• Hosted default chat fallback: if the user has not created a preset, normal chat requests use the deployed /api/v1 proxy and the default OpenRouter model

Deployment

Build the frontend with:

npm run build

Deploy with Wrangler after configuring the Worker bindings in wrangler.toml and the required secrets.

Worker Variables

• GOOGLE_API_KEY: API key for Google Search API
• GOOGLE_CSE_CX: Custom search engine ID for Google Search API
• OPENROUTER_API_KEY: API key used by the hosted /api/v1/chat/completions and /api/v1/responses proxy routes
• OPENROUTER_BASE_URL: Optional OpenAI-compatible base URL for the hosted proxy routes. Defaults to https://openrouter.ai/api/v1
• OPENROUTER_MODEL: Optional fallback model for hosted proxy routes when a request does not provide model
• OPENAI_API_KEY: API key used by Cloudflare Workflows research tasks
• OPENAI_BASE_URL: Optional OpenAI-compatible base URL for Cloudflare Workflows research tasks
• OPENAI_MODEL: Optional model for Cloudflare Workflows research tasks
• FALLBACK_API_KEY: Optional API key for the workflow fallback model provider
• FALLBACK_BASE_URL: Optional base URL for the workflow fallback model provider
• FALLBACK_MODEL: Optional fallback model for workflow retries

Client-side presets can also point directly at a user-provided OpenAI-compatible API base URL. When no preset base URL is configured, the app uses the deployed Worker proxy at /api/v1.

If no preset has been created or selected, normal chat sends a non-secret placeholder API key and an empty model value to the hosted Worker proxy. The placeholder only satisfies the browser SDK credential check; the Worker still uses OPENROUTER_API_KEY server-side and chooses the model from OPENROUTER_MODEL or its server-side fallback. This keeps the first-run chat path usable without asking the user for an API key. Search, image search, deep research, image generation, and fine-tuning still depend on their corresponding Worker variables or preset settings.

Development

npm install
npm run dev

Useful checks:

npm run build
npm run lint