Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.wolffi.sh/llms.txt

Use this file to discover all available pages before exploring further.

​
Providers

Wolffish communicates with LLMs via three providers using pure fetch() — no SDKs. Each provider has its own streaming format and tool-calling convention, which wernicke.ts normalizes into a single interface.

​
Anthropic (Claude)

POST https://api.anthropic.com/v1/messages
Uses SSE streaming. Tool calls arrive as tool_use content blocks. Configure your API key in Settings or directly in config.json. Best for: Complex reasoning, detailed instructions following, nuanced tool use.

​
OpenAI (GPT)

POST https://api.openai.com/v1/chat/completions
Uses SSE streaming. Tool calls arrive as function_call objects. Configure your API key in Settings or directly in config.json. Best for: General-purpose tasks, broad knowledge, fast responses.

​
Ollama (Local)

POST http://localhost:11434/api/chat
Uses NDJSON streaming. Tool calls arrive as structured JSON in the response. No API key needed — runs entirely on your machine. Best for: Privacy, offline use, zero-cost experimentation, always-available fallback.

​
Health Tracking

The thalamus tracks each provider’s health independently:
  • Failure count — Incremented on each failed request
  • Backoff cooldown — Exponential backoff after failures
  • Online check — net.isOnline() for instant offline detection
When a provider fails, it enters a cooldown period before being retried. The cascade skips unhealthy providers and goes directly to the next available one.

​
Choosing a Primary Provider

Set your primary provider in Settings or config.json. The cascade always falls back in order: Claude → OpenAI → Ollama. Your primary provider is tried first, then the cascade takes over on failure.
For the best experience, keep Ollama running with a pulled model as your safety net. Even if you primarily use Claude, having a local fallback means you’re never stuck without a response.