Back to Agno

xAI Cookbook

cookbook/90_models/xai/README.md

3.0.0a32.5 KB
Original Source

xAI Cookbook

Note: Fork and clone this repository if needed

1. Create and activate a virtual environment

shell
python3 -m venv ~/.venvs/aienv
source ~/.venvs/aienv/bin/activate

2. Export your XAI_API_KEY

shell
export XAI_API_KEY=***

3. Install libraries

shell
uv pip install -U openai ddgs duckdb yfinance agno

4. Run basic Agent

  • Streaming on
shell
python cookbook/92_models/xai/basic_stream.py
  • Streaming off
shell
python cookbook/92_models/xai/basic.py

5. Run with Tools

  • DuckDuckGo Search
shell
python cookbook/92_models/xai/tool_use.py

6. Run Agent with Image URL Input

shell
python cookbook/92_models/xai/image_agent.py

7. Run Agent with Image Input

shell
python cookbook/92_models/xai/image_agent_bytes.py

8. Run Agent with Image Input and Memory

shell
python cookbook/92_models/xai/image_agent_with_memory.py

9. Run Agent with SuperGrok sign-in (no API key)

Sign in with a SuperGrok subscription through the OAuth device flow instead of setting XAI_API_KEY. The stored token is encrypted with a dedicated key:

shell
export XAI_TOKEN_ENCRYPTION_KEY=***

Generate a key with python -c "from agno.utils.encryption import generate_encryption_key; print(generate_encryption_key())"

shell
python cookbook/90_models/xai/oauth_device_login.py

10. Run Agent with SuperGrok sign-in from chat

Sign in from inside the conversation instead of the terminal: a sign-in agent hands the user an approval link on one turn and finishes the sign-in on the next, and a Grok agent then answers on the subscription. Two agents, because an agent cannot sign in to the model it is already running on - so the sign-in agent runs on a model that does not need the SuperGrok session. Use this for chatbots and web UIs.

shell
export OPENAI_API_KEY=***
export XAI_TOKEN_ENCRYPTION_KEY=***
shell
python cookbook/90_models/xai/oauth_chat_signin.py

11. Run Agents with per-user SuperGrok sign-in

Several people share one deployment and each spends their own subscription. The token is stored under the user_id the run carries, and the model resolves that user's token per request. A user who has not signed in falls back to the deployment's own session; require_user_token=True on the model refuses that fallback and requires everyone to sign in first. Per-user tokens need a database - one token file cannot hold a session each. Same two keys as above:

shell
python cookbook/90_models/xai/oauth_multi_user.py