Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.helicone.ai/llms.txt

Use this file to discover all available pages before exploring further.

This integration method is maintained but no longer actively developed. For the best experience and latest features, use our new AI Gateway with unified API access to 100+ models.
1

Create an account + Generate an API Key

Log into Helicone or create an account. Once you have an account, you can generate an API key.
2

Create Google Generative AI API Key

Visit the Google Generative AI API Key page. Follow the instructions to create a new API key. Make sure to save the key as you will need it for the next steps.
3

Set API keys as environment variables

export HELICONE_API_KEY=<your Helicone API key>
export GOOGLE_API_KEY=<your Google Generative AI API key>
4

Install the Google Generative AI SDK

Ensure you have the necessary packages installed in your Python environment:
pip install -U -q "google-genai"
5

Import and configure the client

from google import genai
import os

client = genai.Client(
    api_key=os.environ.get('GOOGLE_API_KEY'),
    http_options={
        "base_url": 'https://gateway.helicone.ai',
        "headers": {
            "helicone-auth": f'Bearer {os.environ.get("HELICONE_API_KEY")}',
            "helicone-target-url": 'https://generativelanguage.googleapis.com'
        }
    }
)
If you’re using Vertex AI integration (with GOOGLE_GENAI_USE_VERTEXAI=True), you need to modify the target URL to point to the Vertex AI endpoint:
# Use this target URL when GOOGLE_GENAI_USE_VERTEXAI=True
"helicone-target-url": f'https://{os.environ.get("GOOGLE_CLOUD_LOCATION")}-aiplatform.googleapis.com'
Make sure the GOOGLE_CLOUD_LOCATION environment variable is set to your Google Cloud region (e.g., us-central1).
6

Generate content using the model

response = client.models.generate_content(
    model='gemini-2.0-flash',
    contents='Tell me a story in 300 words.'
)
print(response.text)

# Optional: Print the full response details
print(response.model_dump_json(
    exclude_none=True, indent=4))

​
Adding User-Specific Headers with Client Factory Method

Currently, the Gemini Python SDK doesn’t support setting headers at request time, unlike OpenAI and Anthropic SDKs (see GitHub issue #698). As a workaround, you can create a factory function that dynamically generates client instances with specific headers.

​
Client Factory Method for User-Specific Headers

from typing import Optional, List, Tuple, Any
from google import genai
import os
import instructor

def setup_gemini_client(
    user_id: Optional[str] = None,
    session_id: Optional[str] = None,
    session_name: Optional[str] = None,
    session_path: Optional[str] = None,
    custom_properties: Optional[dict] = None,
    other_metadata: Optional[dict] = None
) -> Any:
    """
    Factory function to create a Gemini client with user-specific Helicone headers.

    Args:
        user_id: Optional user identifier for Helicone tracking
        session_id: Optional session identifier for Helicone tracking
        session_name: Optional name for the session in Helicone
        session_path: Optional path for the session in Helicone
        custom_properties: Optional dictionary of custom properties for Helicone
        other_metadata: Optional dictionary of additional metadata headers

    Returns:
        Configured Gemini client with Instructor integration
    """
    # Base headers required for Helicone
    metadata: List[Tuple[str, str]] = [
        ("helicone-auth", f"Bearer {os.environ.get('HELICONE_API_KEY')}"),
        ("helicone-target-url", "https://generativelanguage.googleapis.com"),
    ]

    # Add user_id if provided
    if user_id:
        metadata.append(("helicone-user-id", user_id))

    # Add session_id if provided
    if session_id:
        metadata.append(("helicone-session-id", session_id))

    # Add session_name if provided
    if session_name:
        metadata.append(("helicone-session-name", session_name))

    # Add session_path if provided
    if session_path:
        metadata.append(("helicone-session-path", session_path))

    # Add custom properties if provided
    if custom_properties:
        for key, value in custom_properties.items():
            metadata.append((f"helicone-property-{key}", value))

    # Add other metadata if provided
    if other_metadata:
        for key, value in other_metadata.items():
            metadata.append((key, value))

    # Configure the client with metadata
    genai.configure(
        api_key=os.environ.get('GOOGLE_API_KEY'),
        client_options={
            "api_endpoint": "gateway.helicone.ai",
        },
        default_metadata=metadata,
        transport="rest",
    )

    # Create Gemini client with Instructor integration
    gemini_client = instructor.from_gemini(
        genai.GenerativeModel(
            model_name="gemini-2.0-flash",
        ),
        mode=instructor.Mode.GEMINI_JSON,
    )

    return gemini_client

session_client = setup_gemini_client(
    user_id="user123",
    session_id="session456",
    session_name="Customer Support Chat",
    session_path="/support/tickets/123",
    custom_properties={
        "job_id": "1234567890",
        "job_name": "Customer Support Chat",
    }
)

response = session_client.chat.completions.create(
    messages=[
        {"role": "user", "content": "What are black holes?"}
    ],
    response_model=Response
)
This factory method approach allows you to create clients with different user identifiers, session tracking, and custom properties for each request, working around the current limitation in the Gemini SDK while providing comprehensive tracking capabilities in Helicone.