Embeddings

Provider-first embedding architecture — auto-discovered from your model providers.

AgentOven uses a provider-first embedding architecture: when you configure a model provider (OpenAI, Azure OpenAI, Ollama), its embedding capabilities are automatically discovered and made available. No separate embedding configuration required.

How Provider-First Embeddings Work

Instead of requiring separate API keys for embeddings, AgentOven derives embedding capabilities from your existing model providers. If a provider's driver implements the EmbeddingCapableDriver interface, its embedding models become available instantly.

1. Register a model provider (e.g., OpenAI with your API key)
2. Server auto-discovers embedding capabilities from the driver
3. Default embedding model is registered in the embedding registry
4. RAG pipeline, vector stores, and ingestion use it automatically
5. All available embedding models are discoverable via API

Supported Embedding Models

Each provider exposes different embedding models with varying dimensions and batch sizes:

OpenAI

Azure OpenAI

Ollama

Configuring via Provider Registration

Register a provider through the API, and embeddings are available immediately:

$ curl -X POST http://localhost:8080/api/v1/providers \
  -H "Content-Type: application/json" \
  -H "X-Kitchen: default" \
  -d '{
    "name": "my-openai",
    "kind": "openai",
    "config": { "api_key": "sk-..." },
    "models": ["gpt-4o", "gpt-4o-mini"]
  }'

# Embeddings are now auto-available!
$ curl http://localhost:8080/api/v1/embeddings/drivers

Fallback: Environment Variables

For quick local development, you can still use environment variables. These are used as a fallback when no providers with embedding capabilities are configured:

# OpenAI embeddings
$ export OPENAI_API_KEY="sk-..."
$ export AGENTOVEN_EMBEDDING_MODEL="text-embedding-3-small"

# Ollama embeddings
$ export OLLAMA_URL="http://localhost:11434"
$ export AGENTOVEN_OLLAMA_EMBED_MODEL="nomic-embed-text"

API Endpoints

Enterprise Extensions

AgentOven Pro adds embedding support for enterprise cloud providers:

• AWS Bedrock — Amazon Titan embeddings (v1, v2, Image), Cohere English/Multilingual
• Azure AI Foundry — Azure OpenAI deployment-based embeddings
• Google Vertex AI — text-embedding-004/005, multilingual-002, gecko@003