An agent that remembers across conversations. It loads conversation history and facts from memory, thinks using relevant facts to plan a response, updates memory with the new turn, and responds with a confidence score and token count.
userId, userMessage -> am_load_memory -> am_agent_think -> am_update_memory -> am_agent_respond
LoadMemoryWorker (am_load_memory) -- Loads prior conversation entries (e.g., {role: "user", content: "Hello, I'm interested in machine learning."}).
AgentThinkWorker (am_agent_think) -- Retrieves relevant facts from memory and produces structured thoughts.
UpdateMemoryWorker (am_update_memory) -- Updates the memory snapshot: {totalEntries: 6, factsStored: 5}.
AgentRespondWorker (am_agent_respond) -- Generates the final response with confidence score and token count.
33 tests cover memory loading, reasoning, memory updates, and response generation.
- RUNNING.md -- how to build and run this example