Skip to main contentModule llm
Source - ChatMessage
- Chat message for /api/chat (multi-turn with tool support).
- ChatToolResponse
- Result of chat_with_tools call.
- LlmCallStats
- Stats captured from an LLM call for pipeline reports.
- LlmClient
- Unified LLM client — routes to Claude API, Grok API, or Ollama based on model name.
Supports both blocking and streaming generation.
- OllamaTool
- Ollama tool definition (JSON Schema).
- OllamaToolCall
- Tool call returned by the model.
- OllamaToolCallFunction
- OllamaToolFunction
- StreamEvent
- Events emitted during streaming generation.
- extract_code
- Extract clean code from an LLM response that may contain markdown fences.
- ollama_url
- Get the Ollama base URL from OLLAMA_HOST env var, or default to localhost.
Supports: “host:port”, “http://host:port”, or just “host”.