skills.rest
topoteretes
topoteretesOFFICIAL
Explore all skills by this author →
Website𝕏@cognee_GitHub

cognee-integrations

Configure cognee LLM providers, embedding backends, databases, storage, and MCP server via environment variables.

30.4k|3.0k|Updated Aug 16, 2023
Search Tags:#vector database#environment variables#Configuration#neo4j#LLM providers#MCP Server#Cognee
One-click install
npx skills add https://github.com/topoteretes/cognee --skill cognee-integrations
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cognee-integrations
Source: https://github.com/topoteretes/cognee/tree/main/.claude/skills/cognee-integrations
Command: npx skills add https://github.com/topoteretes/cognee --skill cognee-integrations

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Switching cognee between LLM providers, embedding models, databases, or storage backends requires knowing the correct environment variables and matching pip extras, and misconfiguring one side (LLM vs. embeddings) silently leaves the other on OpenAI defaults.

Core Features & Use Cases

  • LLM & Embedding Provider Setup: Configure OpenAI, Azure, Gemini, Anthropic, Ollama, OpenRouter, or AWS Bedrock via LLM_PROVIDER and related variables, with independent EMBEDDING_* configuration.
  • Database & Storage Switching: Set relational (sqlite/postgres), vector (lancedb/pgvector/neptune/turso), graph (ladybug/neo4j/neptune) backends, plus S3 storage and session cache options.
  • MCP Server for IDEs: Start the MCP server with docker compose --profile mcp up to connect Cursor, Claude Desktop, or Claude Code to cognee memory.
  • Use Case: You want to move a project from OpenAI to a local Ollama model with Neo4j as the graph store — the Skill tells you exactly which env vars, extras (cognee[ollama], cognee[neo4j]), and post-switch reset steps are required.

Quick Start

Ask the assistant to switch cognee from OpenAI to Gemini for the LLM and from lancedb to pgvector for the vector store, and have it list the exact .env variables and pip extras needed.

Frequently Asked Questions about cognee-integrations

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I switch cognee from OpenAI to another LLM provider?▼

Set LLM_PROVIDER, LLM_MODEL, LLM_API_KEY, and where relevant LLM_ENDPOINT and LLM_API_VERSION in your .env file. For example, use LLM_PROVIDER=azure with an endpoint and API version for Azure OpenAI, or LLM_PROVIDER=gemini with a Gemini model name.

How do I configure a different vector database in cognee?▼

Set VECTOR_DB_PROVIDER to pgvector, neptune_analytics, or turso and install the matching extra such as cognee[postgres]. Other databases like ChromaDB or Qdrant require community adapters registered with use_vector_adapter before use.

Why does cognee still call OpenAI after I changed the LLM provider?▼

LLM and embedding settings are configured independently. Changing only LLM_PROVIDER leaves EMBEDDING_PROVIDER on OpenAI, so you must also set EMBEDDING_PROVIDER, EMBEDDING_MODEL, and related variables or keep a valid OpenAI key.

Can I use cognee with Neo4j as the graph database?▼

Yes, set GRAPH_DATABASE_PROVIDER=neo4j with a bolt URL and credentials, and install cognee[neo4j]. The repo's docker-compose.yml includes a ready-to-use neo4j profile with matching default credentials.

What happens to existing data when I change the embedding model?▼

Embeddings from different models are not comparable, so you must reset local state with cognee-cli forget --all or await cognee.forget(everything=True) and re-ingest. Use memory_only=True to drop just the graph and vectors while keeping ingested files.

How do I connect cognee memory to Cursor or Claude Desktop?▼

Run docker compose --profile mcp up to start the MCP server on port 8001 using SSE transport, then point your IDE at it. Configure its DB_* environment variables to match the main service so both share the same data.

Related Skills

Explore More Software Engineering Skills→
alexmaruccialexmarucci

agentless-scanning

Manage Datadog Agentless Scanning for AWS and Azure accounts.

Community
Advanced
topoteretestopoteretes
29.9k

cognee

Converts inputs into a persistent, searchable graph-based memory for AI agents via asynchronous Python API.

Official
Advanced
PostergullyPostergully

setup

Automate GBrain provisioning, AGENTS.md injection, and initial data import.

Community
Advanced
bykoleksii-hardobykoleksii-hardo

memory-systems

Integrate persistent semantic memory with vector stores and knowledge graphs.

Community
Advanced