Add agent memory benchmark

[ichko]

Note

Low Risk
Adds an isolated benchmark package and documentation/CI; it does not change V0 server code. Operational risk is limited to optional external API credentials when running provider adapters locally or in CI beyond unit tests.

Overview
Introduces agent-memory-benchmark/, a new Python package with the memory-bench CLI for reproducible LongMemEval v1 runs: load public splits, ingest sessions per example with isolated user_ids, wait for provider extraction where needed, answer with a shared OpenAI step, and write resumable artifacts (answers.jsonl, metadata.json, errors.jsonl). A separate judge command applies task-specific LLM rubrics and writes metrics.json (including task-averaged accuracy).

The harness ships nine MemoryStore adapters (Redis Agent Memory REST, Mem0, LangMem, Zep, Graphiti, Supermemory, Vertex Memory Bank, Bedrock AgentCore, Oracle Agent Memory), optional uv extras, provider recipes under docs/providers/, and pytest coverage for CLI, datasets, judge, and runner resume/redaction behavior. GitHub Actions runs ruff, pytest, and uv build in that directory on push/PR.

Root landing narrative is updated so the repo is explicitly two projects: V0/ as the preserved research server and the benchmark as an open audit surface (not a leaderboard). README.md, AGENTS.md, CLAUDE.md, and V0/README.md emphasize Redis Agent Memory in Iris as the supported product path with clearer quickstart links.

^{Reviewed by Cursor Bugbot for commit 92b7be7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[tylerhutcherson]

Overall looks promising. A few of the cursor bug notes might be worth addressing. I also commented on some small readme improvements. Seems like we are still missing the redis implementation afaik

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{Reviewed by Cursor Bugbot for commit af09623. Configure here.}