Skip to content

INSTALLATION.md

Latest commit

 

History

History
315 lines (238 loc) · 7.12 KB

INSTALLATION.md

File metadata and controls

315 lines (238 loc) · 7.12 KB

cascadeflow Installation Guide

📦 Files

requirements.txt       → Core dependencies only (2 packages)
requirements-dev.txt   → Development + all providers + testing tools

🎯 Installation Options

Production Users

# Minimal install (just core)
pip install cascadeflow

# With specific provider
pip install cascadeflow[openai]
pip install cascadeflow[anthropic]

# With common providers (OpenAI + Anthropic + Groq)
pip install cascadeflow[providers]

# With everything
pip install cascadeflow[all]

Developers/Contributors

# Method 1: Using requirements-dev.txt
pip install -r requirements-dev.txt

# Method 2: Using pyproject.toml extras (recommended)
pip install -e ".[dev]"

📊 What's Included

requirements.txt (Core Only)

pydantic>=2.0.0      # Data validation
httpx>=0.25.0        # HTTP client

That's it! Just 2 core dependencies. Providers are optional extras.

requirements-dev.txt (Everything)

Core dependencies       ✅
All provider SDKs       ✅ (openai, anthropic, groq, huggingface, together, vllm)
Testing tools           ✅ (pytest, pytest-asyncio, pytest-cov, pytest-mock)
Code quality tools      ✅ (black, ruff, mypy, isort, pre-commit)
Development utilities   ✅ (rich for terminal output)

🆓 Free/Local Providers

Ollama (Recommended for Development)

# No Python package needed!

# 1. Install Ollama from https://ollama.ai
curl -fsSL https://ollama.com/install.sh | sh

# 2. Pull a model
ollama pull llama3.2:1b

# 3. Use with cascadeflow
pip install cascadeflow  # Core only, no extras needed!

Cost: $0/month 💰

vLLM

# Option 1: HTTP server (no Python package)
# Run vLLM server, connect via HTTP

# Option 2: Python package
pip install cascadeflow[vllm]

🔑 Provider Setup

API Keys (.env file)

# Only add keys for providers you want to use

# OpenAI
OPENAI_API_KEY=sk-proj-...

# Anthropic
ANTHROPIC_API_KEY=sk-ant-...

# Groq (free tier available!)
GROQ_API_KEY=gsk_...

# HuggingFace
HF_TOKEN=hf_...

# Together.ai
TOGETHER_API_KEY=...

# Ollama - no API key needed! (local)
# vLLM - no API key needed! (local)

🚀 Quick Start

For Production

# Install with common providers
pip install cascadeflow[providers]

# Set API keys in .env
echo "OPENAI_API_KEY=sk-..." >> .env
echo "ANTHROPIC_API_KEY=sk-ant-..." >> .env

# Start using
python your_app.py

For Development

# Clone repo
git clone https://github.com/lemony-ai/cascadeflow.git
cd cascadeflow

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # Linux/Mac
# or: .venv\Scripts\activate  # Windows

# Install in dev mode
pip install -e ".[dev]"

# Run tests
pytest

🧪 Testing Without API Keys

Use Ollama for free local testing:

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a small model
ollama pull llama3.2:1b

# Install just core cascadeflow
pip install cascadeflow

# Test
python examples/multi_provider.py

📋 Dependencies Summary

Package Used By Required? Install Via
pydantic Core validation ✅ Always requirements.txt
httpx Core HTTP client ✅ Always requirements.txt
openai OpenAIProvider ❌ Optional [openai] or [providers] or [all]
anthropic AnthropicProvider ❌ Optional [anthropic] or [providers] or [all]
groq GroqProvider ❌ Optional [groq] or [providers] or [all]
huggingface-hub HuggingFaceProvider ❌ Optional [huggingface] or [all]
together TogetherProvider ❌ Optional [together] or [all]
vllm VLLMProvider ❌ Optional [vllm] or [all]
rich Dev/Debug ❌ Dev only requirements-dev.txt
pytest Testing ❌ Dev only requirements-dev.txt
black Formatting ❌ Dev only requirements-dev.txt
ruff Linting ❌ Dev only requirements-dev.txt
mypy Type checking ❌ Dev only requirements-dev.txt

💡 Best Practices

For Production

  1. Start minimal: pip install cascadeflow
  2. Add providers as needed: pip install cascadeflow[openai]
  3. Use Ollama for free local inference

For Development

  1. Always use virtual environment
  2. Install in editable mode: pip install -e ".[dev]"
  3. Run tests before committing: pytest
  4. Format code: black . && isort .
  5. Check types: mypy cascadeflow/

For Testing

  1. Use Ollama for free testing (no API costs!)
  2. Mock API calls when testing without keys
  3. Use pytest fixtures for provider initialization

🔧 Troubleshooting

"No module named 'openai'"

# Install OpenAI extra
pip install cascadeflow[openai]

"No module named 'anthropic'"

# Install Anthropic extra
pip install cascadeflow[anthropic]

"Connection refused" with Ollama

# Make sure Ollama is running
ollama serve

# Check if running
curl http://localhost:11434/api/tags

Version conflicts

# Create fresh virtual environment
python -m venv .venv
source .venv/bin/activate
pip install cascadeflow[providers]

IDE shows missing dependencies

# Sync your IDE with virtual environment
# In PyCharm/IntelliJ: File → Project Structure → SDK
# Select .venv/bin/python

# Or reinstall
pip install -r requirements-dev.txt

🎯 Installation Examples by Use Case

"I want to try cascadeflow with OpenAI"

pip install cascadeflow[openai]

"I want OpenAI + Anthropic"

pip install cascadeflow[openai,anthropic]
# or
pip install cascadeflow[providers]

"I want everything"

pip install cascadeflow[all]

"I want free local models only"

pip install cascadeflow
# Then install Ollama from https://ollama.ai

"I'm contributing to cascadeflow"

git clone https://github.com/lemony-ai/cascadeflow.git
cd cascadeflow
pip install -e ".[dev]"
pre-commit install
pytest

✅ Verification

Test your installation:

# Core
python -c "import cascadeflow; print('✅ Core OK')"

# OpenAI (if installed)
python -c "import openai; print('✅ OpenAI OK')"

# Anthropic (if installed)
python -c "import anthropic; print('✅ Anthropic OK')"

# Full test
python -c "
from cascadeflow import CascadeAgent, ModelConfig
print('✅ All imports working!')
"

📊 Provider Comparison

Provider Cost Speed Quality Setup API Key
OpenAI $$$ Medium High Easy Yes
Anthropic $$$ Medium High Easy Yes
Groq $ Fast Medium Easy Yes
Ollama Free Fast Medium Medium No
vLLM Free Very Fast Medium Hard No

🔗 Links

📝 Notes

  • tiktoken removed: Not used in current implementation
  • Ollama: No Python package needed - uses HTTP directly
  • Core is minimal: Only 2 dependencies for maximum flexibility