API Keys

You can provide API keys in different locations; this allows different pipelines to use different API keys or accounts while sharing common defaults.

Specifying the Path to API Key Files

You can specify the path to API key files containing API keys in several locations. The RAG server will search for your API key in the order these locations are listed below.

The first location searched for an API key is within the pipelines section of your configuration file:

pipelines:
  - name: "production"
    api_keys:
      anthropic: "/etc/pgedge/keys/prod-anthropic.key"
    # ... other pipeline config

Use the following fields within the configuration file:

If the RAG server does not locate an API key in the pipelines section, it searches in the defaults section of the configuration file:

defaults:
  api_keys:
    openai: "/etc/pgedge/keys/default-openai.key"
    anthropic: "/etc/pgedge/keys/default-anthropic.key"

You also have the option of providing key values in a globally accessible api_keys section of the configuration file. Keys provided in the api_keys section apply to all pipelines unless overridden by values in the pipelines or defaults sections.

api_keys:
  anthropic: "/etc/pgedge/keys/anthropic.key"
  voyage: "/etc/pgedge/keys/voyage.key"
  openai: "~/secrets/openai-api-key"

Providing API Values in Environment Variables or Default File Locations

If you don't specify the API key values in the configuration file, the RAG server will check environment variables for key values:

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
export VOYAGE_API_KEY="pa-..."
export GEMINI_API_KEY="your-gemini-key"

If neither configuration paths nor environment variables are set, the server looks for API keys in these default locations:

Gemini Configuration

Google Gemini uses API key authentication. The key is sent as a query parameter with each request. Default models are gemini-2.0-flash for completion and text-embedding-004 for embeddings.

embedding_llm:
  provider: "gemini"
  model: "text-embedding-004"
rag_llm:
  provider: "gemini"
  model: "gemini-2.0-flash"

The default base URL is https://generativelanguage.googleapis.com. To use a different endpoint, set base_url in the LLM configuration:

rag_llm:
  provider: "gemini"
  model: "gemini-2.0-flash"
  base_url: "https://your-gemini-proxy.example.com"

OpenAI-Compatible Local Providers

When using OpenAI-compatible local LLM servers such as LM Studio, Docker Model Runner, or EXO, the API key is optional. Set base_url to point at the local server:

embedding_llm:
  provider: "openai"
  model: "nomic-embed-text"
  base_url: "http://localhost:1234/v1"
rag_llm:
  provider: "openai"
  model: "llama3"
  base_url: "http://localhost:1234/v1"

No API key is required for local servers. If a key is provided (via config, environment variable, or default file location), it will be sent as a Bearer token as usual.

Ollama Configuration

Ollama runs locally and does not require API keys. By default, it connects to http://localhost:11434. To use a different URL, set the base_url in the LLM configuration:

embedding_llm:
  provider: "ollama"
  model: "nomic-embed-text"
  base_url: "http://my-ollama-server:11434"

Alternatively, set the OLLAMA_HOST environment variable:

export OLLAMA_HOST="http://my-ollama-server:11434"