Configure a model provider

akasha selects providers through model aliases. Put credentials in your shell environment or a local .env file. Never commit API keys.

GEMINI_API_KEY=your_key
OPENAI_API_KEY=your_key
ANTHROPIC_API_KEY=your_key

Gemini backend modes

For the Gemini Developer API, set only an API key:

GEMINI_API_KEY=your_key

To use Vertex AI Express Mode with an API key instead of Application Default Credentials (ADC), use the same key and explicitly opt in:

GEMINI_API_KEY=your_key
GOOGLE_GENAI_USE_VERTEXAI=true

Do not set GOOGLE_CLOUD_PROJECT or GOOGLE_CLOUD_LOCATION for Express Mode. They select the full Vertex AI project/location path, which requires ADC and takes precedence over an API key. Akasha's Vertex API-key support is Express Mode; it does not configure the full ADC-based Vertex AI path.

Examples of model aliases:

gemini:gemini-2.5-flash
openai:gpt-4o
anthropic:claude-3-5-sonnet-latest
ollama:qwen3:8b

The provider name before : selects the integration. The model name after : selects the model.

A safe configuration check

Do not print the key. Check only that the variable exists:

import os

if not os.getenv("GEMINI_API_KEY"):
    raise RuntimeError("Set GEMINI_API_KEY before running this example")

print("Provider configuration is present")

Warning

A configured key may still incur provider charges. Check the provider's model access and billing settings before running live examples.

Next: Run your first chat.