Skip to main content

​
Installation

Fenic is a Python package that requires Python 3.10 or higher. It can be installed using pip or uv.

​
Quick Install

pip install fenic

​
Python Version Requirements

Fenic requires Python 3.10, 3.11, or 3.12. Verify your Python version: 
python --version
Python 3.13 is not yet supported. Make sure you’re using Python 3.10-3.12.

​
Optional Provider Packages

Fenic supports multiple AI providers. Install additional packages based on the providers you plan to use: 
pip install "fenic[anthropic]"
Enables Claude models (Haiku, Sonnet, Opus)

​
Install Multiple Providers

You can install multiple provider packages at once: 
pip install "fenic[anthropic,google,mcp]"

​
API Keys Configuration

Fenic requires API keys for the AI providers you use. Set them as environment variables: 
export OPENAI_API_KEY="sk-..."
Never commit API keys to version control. Use environment variables or a secrets management system.

​
Alternative: Use .env Files

For local development, you can create a .env file:
.env
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AIza...
Then load it in your Python code: 
from dotenv import load_dotenv
load_dotenv()

import fenic as fc
# API keys are now available to Fenic
You’ll need to install python-dotenv: pip install python-dotenv

​
Verify Installation

Verify that Fenic is installed correctly: 
import fenic as fc

print(fc.__version__)
# Should print: 0.7.0 (or your installed version)

​
Supported AI Providers

Fenic integrates with multiple AI providers out of the box:

​
Language Models

OpenAI

  • GPT-4, GPT-4o, GPT-5 series
  • o1, o3 series (reasoning models)
  • Fast and reliable

Anthropic

  • Claude 3.5 (Haiku, Sonnet, Opus)
  • Claude 4 series
  • Excellent for long contexts

Google Gemini

  • Gemini 2.0/2.5 Flash
  • Thinking modes support
  • Multimodal capabilities

OpenRouter

  • Access to 200+ models
  • Aggregator for multiple providers
  • Flexible model selection

​
Embedding Models

ProviderModelsUse Case
OpenAItext-embedding-3-small/largeGeneral-purpose embeddings
GoogleGemini embedding modelsMultimodal embeddings
Cohereembed-v4.0Multi-language support

​
Core Dependencies

Fenic automatically installs these core dependencies:
  • polars (1.20.0-1.30.0) — DataFrame operations
  • duckdb (≥1.1.3) — SQL engine
  • lancedb (0.22.0-0.25.0) — Vector storage
  • openai (≥1.101.0) — OpenAI API client
  • tiktoken (≥0.11.0) — Token counting
  • pydantic — Data validation
  • jinja2 (≥3.1.6) — Template rendering
  • pymupdf (≥1.26.4) — PDF parsing
All dependencies are automatically managed. You don’t need to install them separately.

​
Data Source Support

Fenic can read from various data sources without additional installation:
  • Local files: PDF, Markdown, CSV, JSON, Parquet
  • Cloud storage: S3 (via boto3)
  • In-memory: Polars, Pandas, PyArrow DataFrames
  • Datasets: Hugging Face Datasets
  • Databases: DuckDB

​
Development Installation

For contributors or those wanting to modify Fenic: 
# Clone the repository
git clone https://github.com/typedef-ai/fenic.git
cd fenic

# Install with development dependencies
just setup
See the Contributing Guide for more details.

​
Troubleshooting

​
Import Error

If you get an import error: 
ImportError: No module named 'fenic'
Make sure Fenic is installed in your active Python environment: 
pip list | grep fenic

​
API Key Not Found

If you get an API key error: 
Error: OPENAI_API_KEY not found
Verify the environment variable is set: 
echo $OPENAI_API_KEY

​
Version Conflicts

If you experience dependency conflicts, create a fresh virtual environment: 
python -m venv fenic-env
source fenic-env/bin/activate  # On Windows: fenic-env\Scripts\activate
pip install fenic

​
Next Steps

Quickstart Guide

Build your first context pipeline with Fenic

Build docs developers (and LLMs) love

Get started for freeTalk to us