Awesome-AI-Memory — Papers & Projects for LLM Memory

Main

• Use it as a reading queue: start from the scope/goal section, then jump to the linked Papers and Projects lists (README badges show counts).

• When designing memory, keep components explicit: storage (vector/graph/sql), write policy, retrieval policy, and compression/forgetting.

• Track evaluation early: use the repo’s benchmark/evaluation sections to choose measurable tasks (long-context, personalization, multi-session).

• Keep your own “memory design doc” next to your codebase and cite entries from this repo as references for tradeoffs.

Source-backed notes

• README includes badges with quantitative counts: 399 papers and 104 open source projects.
• README describes the motivation (context window limits) and positions memory systems as external/persistent structures for agents.
• README links to separate Chinese README and outlines scope and exclusions.

FAQ

• Is this an implementation?: No — it’s a curated list of papers/projects; use it to choose architectures and references.
• How do I keep it actionable?: Pick one memory pattern, one benchmark, and one open-source project to replicate as a baseline.
• Where are the numbers from?: The counts (papers/projects) are shown in README badges at the top of the repo.