Intelligence
tnsai-intelligence adds planning, multi-step reasoning, state machines, context management and learning to agents. Add the module to use any of these pages; see Install.
| Page | Covers |
|---|---|
| Planning | Planners that choose actions from goals and state, and LLM-written plans with review and replanning |
| Reasoning | ReAct, Tree of Thoughts, Graph of Thoughts, self-consistency and other executors that spend several LLM calls on one question |
| State machines | Fixed states and transitions driven by events and guards |
| Context | Compaction, the context graph, sessions, knowledge extraction and memory consolidation |
| Learning | Feedback, norms, the refinement loop and the prompt optimizer |
For typed LLM output, see Structured output.