API reference
Signatures and docstrings, generated from the code. The guide pages explain how the pieces fit together; come here for exact parameters and defaults.
| Page | Covers |
|---|---|
| Native runtime | Tactic, Agent, Prompt, parsers and handlers, tool, Dialog, Message, config and package loaders |
| Protocol | The runtime-neutral Tactic, CallContext, events, as_tactic, the Pydantic AI and native adapters, refs, errors |
| Services | create_tactic_app, custom endpoints, RemoteTactic, request and response models |
| lllm.hub | Manifests, validation, local hubs, catalogs, cards |
| Proxies and sandbox | BaseProxy, proxy tactics, sandbox_tactic, SandboxPolicy |
| lllm.testing | ScriptedModel, FunctionModel, ModelRequest, ToolCall |
| CLI | Every lllm command and flag |
| Service API | The HTTP endpoints of a served tactic, with request and error envelopes |
| Hub metadata | Helpers that export what a tactic or channel publishes to the hub |
Each page says where to import its names from. Most of what the guides use
comes straight from lllm, as in from lllm import Agent, Prompt, Tactic, tool.