Installation¶
AgentFlow requires Python 3.11 or later.
Core Package¶
The core package includes config loading, routing, workflow execution, session management, and the tool registry. It does not include any LLM provider SDKs.
Core dependencies are minimal: pydantic, python-frontmatter, pyyaml, and httpx.
Provider Extras¶
Install the extras for the LLM providers you plan to use:
Installs anthropic>=0.42.0. Use with AnthropicProvider.
Installs openai>=1.0.0. Use with OpenAICompatProvider. Works with any OpenAI-compatible API (OpenAI, Azure OpenAI, Ollama, vLLM, etc.).
Optional Extras¶
| Extra | Install command | What it adds |
|---|---|---|
s3 |
pip install "gittielabs-agentflow[s3]" |
S3 storage backend (boto3) |
vector |
pip install "gittielabs-agentflow[vector]" |
Vector memory with Qdrant (qdrant-client) |
orchestration |
pip install "gittielabs-agentflow[orchestration]" |
Multi-step orchestration primitives |
telemetry |
pip install "gittielabs-agentflow[telemetry]" |
Langfuse event handler (langfuse) |
Install Everything¶
This installs all provider SDKs and optional extras.
Development Setup¶
Clone the repository and install in editable mode with dev dependencies:
Dev dependencies include pytest, pytest-asyncio, and ruff for linting.
Run the test suite:
Run the linter:
Environment Variables¶
Each provider reads its API key from standard environment variables:
| Provider | Environment Variable |
|---|---|
| Anthropic | ANTHROPIC_API_KEY |
| OpenAI | OPENAI_API_KEY |
| Google Gemini | GOOGLE_API_KEY |
| Langfuse | LANGFUSE_PUBLIC_KEY, LANGFUSE_SECRET_KEY |
Set these before running your agents: