OmniMemEval Documentation
July 24, 2026 · View on GitHub
OmniMemEval has two complementary evaluation tracks. Choose the track that matches the system you want to evaluate:
| Track | Evaluation target | Guide | Results |
|---|---|---|---|
| User Memory Evaluation | Memory backend APIs exposed through add() and search() | Guide | Results |
| Agent Memory Evaluation | Agent runtimes equipped with memory plugins | Guide | Results |
Agent Memory contributors can also read the current architecture and extension contracts.
Related documentation remains next to the files it describes:
- Dataset preparation and licenses:
data/<benchmark>/README.md - Environment templates and parameters: env_examples/README.md and env_examples/PARAMETERS.md
- Test commands: scripts/tests/README.md
- Third-party attribution: THIRD_PARTY_NOTICES.md