Tune cognee performance
Speed up, throttle, or estimate the cost of cognee memory ingestion and queries
- Repository
- topoteretes/cognee
- Stars
- 33,002
- Licence
- Apache-2.0
- Files
- 1 file
What it does
This skill teaches an AI agent how to make the cognee memory system faster, cheaper, or more controlled. It explains that most of the time and cost comes from LLM calls, and gives the agent the settings for batching, chunk size, rate limits, per-stage models, background runs, and cost estimates. It also covers read latency and how to scale a self-hosted deployment.
Great for
- Developers running cognee who see slow ingestion, high LLM bills, or rate-limit (429) errors
- Teams planning how many documents to process and what it will cost before a large import
- People deploying cognee who need to understand its single-process scaling limits
Skip it if
- Users who just want to store or search documents with cognee for the first time; the skill assumes an existing working setup
- Anyone expecting horizontal, multi-server scaling; the file says open-source cognee scales vertically in one process and horizontal scaling is not covered
- Situations where the estimate must be exact: the dry-run covers only graph extraction and summarization, not improve() or embeddings
Try asking
- “Estimate the token count and cost of remembering my ./docs folder before I ingest it”
- “My cognee ingestion keeps hitting 429 rate limit errors. How should I configure the LLM rate limiter?”
- “Make chat answers faster by reducing the extra LLM calls per turn”
- “Use a cheaper model just for graph extraction and run ingestion in the background”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.
cognee-performance.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip cognee-performance.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip cognee-performance.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Run the dry_run estimate first, then tune data_per_batch and chunks_per_batch before raising concurrency
- Set LLM_RATE_LIMIT_REQUESTS yourself when routing a stage to a local model, since the local default of 10 is not applied automatically
- Keep the default safety features such as the dataset queue cap rather than disabling them to gain speed
What's inside
- SKILL.md
Redistributed unmodified from topoteretes/cognee under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.