Installation
Requirements
- Python 3.11+
- An API key for your chosen LLM provider
Core install
pip install agentix-agent-sdk
The core package bundles all supported LLM provider SDKs along with pydantic, httpx, tiktoken, and MCP support. No extra install step is needed to use any provider.
Optional extras
| Extra | Installs | Use case |
|---|---|---|
gateway | aiohttp + adapter deps | Messaging gateway — HTTP webhooks, Slack, WhatsApp, Email |
gateway-slack | slack-sdk | Slack adapter only (lighter than full gateway) |
gateway-whatsapp | gateway-core | WhatsApp Business API adapter |
gateway-email | aiosmtplib | Email adapter (SMTP/IMAP) |
dev | pytest, mypy, ruff, fakeredis | Development and testing tools |
all | redis + all gateway adapters + dev | Everything |
pip install agentix-agent-sdk[gateway] # All messaging gateway adapters
pip install agentix-agent-sdk[all] # Everything
API key
Pass your LLM provider API key via the AGENTIX_API_KEY environment variable:
export AGENTIX_API_KEY=your-api-key-here
Or supply it inline via llm_options:
options = AgentixAgentOptions(
provider="anthropic",
llm_options={"api_key": "your-api-key-here"},
)
Agentix automatically redacts
api_keyfrom logs and__repr__to prevent accidental leakage.
OpenAI-compatible endpoints
For any OpenAI-compatible endpoint (Cerebras, Ollama, vLLM, LiteLLM, etc.) use provider="openai_compatible" and set base_url:
options = AgentixAgentOptions(
provider="openai_compatible",
model="your-model-name",
llm_options={
"base_url": "http://localhost:11434/v1", # e.g. Ollama
"api_key": "ollama",
},
)
Or via environment variables:
export AGENTIX_PROVIDER=openai_compatible
export AGENTIX_BASE_URL=http://localhost:11434/v1
export AGENTIX_MODEL=llama3.2
Web search keys (optional)
WebSearch auto-detects which backend to use based on available keys:
export TAVILY_API_KEY=tvly-... # recommended
export BRAVE_API_KEY=BSA...
export SERPAPI_API_KEY=...
# (no key) → falls back to DuckDuckGo
Verify installation
import agentix
print(agentix.__version__)