Using MetaMemory with Ollama (Self-Hosted)

Ollama enables fully self-hosted embedding generation, and MetaMemory's integration brings production-grade memory management to local deployments. For teams with strict data privacy requirements — healthcare, finance, government, or any environment where data cannot leave the network — Ollama eliminates the need to send content to third-party APIs entirely. The nomic-embed-text model is the recommended default: it produces 768-dimensional vectors with quality that rivals commercial embedding APIs, runs efficiently on consumer hardware, and is completely open-source. For higher-dimensional representations, mxbai-embed-large offers 1024-dimensional vectors with stronger performance on complex content. The all-minilm model provides a lightweight alternative at 384 dimensions for resource-constrained environments. MetaMemory's Ollama integration requires a base URL pointing to your Ollama instance, which can run on the same machine, a local server, or any network-accessible host. The system automatically detects available models and their dimensions. Because Ollama runs locally, there are no API costs, no rate limits, and no network latency for embedding generation — which makes it the fastest option for high-throughput memory encoding. MetaMemory handles all the specifics of Ollama's API format and provides the same multi-vector encoding, adaptive retrieval, and consolidation capabilities as cloud-hosted providers. For teams that want the full MetaMemory experience with zero external dependencies, Ollama is the clear choice.

Setup Guide

Configuration Example

curl -X POST https://api.metamemory.tech/v1/providers \
  -H "Authorization: Bearer YOUR_METAMEMORY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "ollama",
    "base_url": "http://localhost:11434",
    "default_model": "nomic-embed-text",
    "settings": {
      "keep_alive": "5m",
      "num_ctx": 8192
    }
  }'

Supported Models

Capabilities