Memory and context persistence

[pavelanni]

Summary

Add memory and context persistence so agents can retain information across conversations.

Motivation

Stateless agents lose all context between requests. For business use cases (ongoing document review, multi-session projects), agents need persistent memory.

Evaluation targets

• Letta — open source, agent memory management framework
• Mem0 — open source, structured memory layer for LLM apps
• Graphiti — open source, temporal knowledge graph for agent memory

All candidates should be evaluated for OpenShift compatibility, resource requirements, and integration complexity.

Acceptance criteria

• Evaluate candidates against OpenShift deployment constraints
• Prototype integration with at least one memory backend
• Memory accessible via tool interface in the agentic loop
• Configuration via agent-config.yaml

From roadmap: docs/slides/docsclaw-intro.html, slide 6