Environment Configuration Examples

This directory contains product-specific environment templates for the public OmniMemEval adapter layer. The open-source benchmark pipelines currently include LoCoMo, LongMemEval, BEAM, PersonaMem v2, and HaluMem.

Agent Memory Evaluation uses a separate template, .env.agent, for OpenClaw, judge, evaluation, and embedding credentials. This keeps agent-runtime settings separate from User Memory backend credentials such as .env.memos.

Track guides:

• User Memory Evaluation
• Agent Memory Evaluation

Quick Start

cp env_examples/.env.memos .env.memos

Fill in the required fields:

• Memory product credentials, for example MEMOS_API_KEY or MEM0_API_KEY.
• ANSWER_MODEL, ANSWER_API_KEY, ANSWER_BASE_URL.
• EVAL_MODEL, EVAL_API_KEY, EVAL_BASE_URL.

Then run one of the benchmark entrypoints:

./scripts/run_locomo_eval.sh --lib memos --env .env.memos
./scripts/run_lme_eval.sh --lib memos --env .env.memos
./scripts/run_beam_eval.sh --lib memos --env .env.memos
./scripts/run_pmv2_eval.sh --lib memos --env .env.memos
./scripts/run_halumem_eval.sh --lib memos --env .env.memos

See PARAMETERS.md for shared runner, benchmark-specific, and LLM settings.

Available Templates

File	Purpose
.env.memos	MemOS
.env.mem0	Mem0
.env.zep	Zep
.env.supermemory	Supermemory
.env.everos	EverOS
.env.letta	Letta
.env.hindsight	Hindsight
.env.graphiti	Graphiti local/self-hosted
.env.cognee	Cognee
.env.viking	Viking Memory
.env.memori	Memori
.env.memmachine	MemMachine
.env.memorylake	MemoryLake
.env.backboard	Backboard.io
.env.mem9	mem9
.env.agent	Agent Memory Evaluation / AgentBench

Templates include only placeholders and public defaults. Do not commit real .env.* files at the repository root.