finder-enrichment-ai-client 0.3.0

Lightweight AI client package for managing API calls to various AI providers

Finder Enrichment AI Client

A lightweight Python package for managing AI API calls to various providers. Supports Google Gemini AI and OpenRouter (providing access to 100+ models from Anthropic, OpenAI, Meta, Google, and more).

Features

• Multi-Provider Support: Google Gemini and OpenRouter (Claude, GPT, Llama, etc.)
• Lightweight: Minimal dependencies, just HTTP requests and YAML parsing
• Simple: Easy-to-use interface for AI API calls
• Extensible: Designed to support multiple AI providers
• Configurable: Support for both environment variables and config files
• Model Mapping: Friendly model names with config-based ID translation

Installation

pip install finder-enrichment-ai-client

Quick Start

Google Gemini AI

from finder_enrichment_ai_client import FinderEnrichmentGoogleAIClient

# Initialize with environment variable
client = FinderEnrichmentGoogleAIClient()

# Or initialize with direct API key
client = FinderEnrichmentGoogleAIClient(api_key="your-api-key-here", model="gemini-2.0-flash-exp")

# Generate text
response = client.generate_content("Hello! How are you?")
if response['success']:
    print(response['text'])
else:
    print(f"Error: {response['error']}")

# Analyze image (base64 encoded)
response = client.analyze_image(
    image_data="base64_encoded_image_data",
    prompt="Describe this image",
    image_content_type="image/webp"
)
if response['success']:
    print(response['text'])

OpenRouter (Claude, GPT, Llama, etc.)

from finder_enrichment_ai_client import FinderEnrichmentOpenRouterClient

# Initialize with environment variable
# Requires OPENROUTER_API_KEY environment variable
client = FinderEnrichmentOpenRouterClient(model="claude-3-5-sonnet")

# Or initialize with direct API key
client = FinderEnrichmentOpenRouterClient(
    api_key="your-openrouter-api-key",
    model="claude-3-5-sonnet"
)

# Generate text
response = client.generate_content("Explain quantum computing")
if response['success']:
    print(response['text'])

# Analyze image with vision models
response = client.analyze_image(
    image_data="base64_encoded_image_data",
    prompt="What's in this image?",
    image_content_type="image/jpeg"
)
if response['success']:
    print(response['text'])

Model Configuration (config.yaml)

Create a config.yaml file to map friendly model names to provider-specific IDs:

models:
  - name: "claude-3-5-sonnet"
    model_id: "anthropic/claude-3.5-sonnet"
    provider: "openrouter"
    
  - name: "gpt-4o"
    model_id: "openai/gpt-4o"
    provider: "openrouter"

Then use the config loader:

from finder_enrichment_ai_client import get_openrouter_model_id

model_id = get_openrouter_model_id("claude-3-5-sonnet")
# Returns: "anthropic/claude-3.5-sonnet"

Environment Variables

For Google Gemini

export GOOGLE_GEMINI_API_KEY="your-google-api-key"

For OpenRouter

export OPENROUTER_API_KEY="your-openrouter-api-key"
export OPENROUTER_BASE_URL="https://openrouter.ai/api/v1"  # Optional, defaults to this

Get your OpenRouter API key at: https://openrouter.ai/keys

API Reference

FinderEnrichmentGoogleAIClient

Methods

• generate_content(prompt, model=None, temperature=0.7, max_tokens=1000): Generate text content
• analyze_image(image_data, prompt, image_content_type="image/webp", model=None): Analyze images with text prompts
• set_model(model): Change the default model
• set_temperature(temperature): Set the default temperature (0.0 to 1.0)
• get_available_models(page_size=100): Get list of available models

FinderEnrichmentOpenRouterClient

Methods

• generate_content(prompt, model=None, temperature=0.7, max_tokens=1000): Generate text content
• analyze_image(image_data, prompt, image_content_type="image/webp", model=None): Analyze images with vision models
• set_model(model): Change the default model
• set_temperature(temperature): Set the default temperature (0.0 to 2.0)

Config Loader Functions

• load_model_config(): Load config.yaml file
• get_model_config_by_name(model_name): Get full model configuration
• get_openrouter_model_id(model_name): Translate model name to OpenRouter model ID
• get_openrouter_base_url(): Get OpenRouter base URL from config

Response Format

All API methods return a dictionary with:

• success: Boolean indicating if the call was successful
• text: Generated text (if successful)
• raw_response: Full API response
• error: Error message (if not successful)

Development

# Install in development mode
pip install -e .

# Run tests
pytest

# Format code
black .
isort .

License

MIT License