BYOK (Bring Your Own Key)

EvidenceLens runs at $0 server-side LLM cost. To get conversational answers and summaries, you supply your own API key for the provider of your choice. We proxy the request, stream the response back, and never store the key.

Supported providers

• Anthropic — Claude models, prompt caching enabled by default.
• OpenAI-compatible — OpenAI, Groq, OpenRouter, Together, DeepInfra, and any other endpoint that speaks the OpenAI Chat Completions API. You configure thebase_url alongside the key.
• Ollama — your own local model. The "key" field carries your OLLAMA_BASE_URL instead of a credential.
• WebLLM — fully in-browser inference, no key required. Slowest, but private by construction.

How it works

1. You paste your key in the settings panel. It is stored only in your browser's localStorage.
2. Each request sends the key over TLS to POST /api/agent on the agent service, which forwards to the upstream provider and streams Server-Sent Events back.
3. The server validates the key on first use (cached for 10 minutes by SHA-256(key + provider)) and writes only token counts, latency, and error codes to telemetry. The key itself is never logged or persisted.

Why BYOK

This is a free public service run on a small budget. Server-side LLM costs would scale with traffic and force us to either rate-limit aggressively or shut down. BYOK keeps the search engine itself permanently free while letting power users opt into whatever model and budget they prefer.

Privacy

Keys live in your browser and the in-memory request lifecycle of the proxy. They are not written to disk, not logged, not shared with third parties. If you'd rather not trust the proxy at all, switch to WebLLM and the entire interaction stays on your machine.