chatixia 0.1.0

Deploy and run AI agents locally

Chatixia SDK

Deploy and run AI agents locally with a single command.

Installation

pip install chatixia

Or from source (development):

pip install -e "./sdk[dev]"

Quick Start

1. Create an agent manifest

chatixia init my-agent

This creates agent.yaml and .env.example in the current directory.

2. Configure credentials

cp .env.example .env
# Edit .env with your API keys

3. Run

# Interactive REPL
chatixia run

# Autonomous mode (goal-driven loop)
chatixia run --daemon

# One-shot (pipe-friendly)
chatixia run --once "Summarize today's new purchase requisitions"

# Validate manifest before running
chatixia validate

Agent Manifest (agent.yaml)

name: procurement-assistant
description: "Monitors and classifies purchase requisitions"

# LLM provider: azure | openai | ollama
provider: azure
model: gpt-5.2

# System prompt — defines the agent's persona
prompt: |
  You are a procurement specialist.
  Use available tools to analyze documents and query databases.

# Skills
skills:
  builtin:                           # Names of built-in skills to enable
    - analyze_document
    - postgres_readonly_query
    - get_ariba_data
  dirs:                              # Additional skill directories
    - ./custom-skills
  disabled:                          # Skills to exclude
    - echo

# Goals (activated with --daemon)
goals:
  - name: monitor_prs
    sensor: "Check Ariba for new purchase requisitions in the last 15 minutes"
    action: "Classify with UNSPSC and post summary to Slack"
    interval: 900                    # seconds
    autonomy: auto                   # auto | notify | approve

# MCP servers
mcp_servers:
  playwright:
    command: npx
    args: [-y, "@playwright/mcp@latest"]

# Environment
env:
  file: .env
  required:
    - AZURE_OPENAI_ENDPOINT
    - AZURE_OPENAI_API_KEY
    - AZURE_OPENAI_DEPLOYMENT

# Runtime tuning
max_turns: 10
context_window: 120000
data_dir: .chatixia

CLI Reference

Command	Description
chatixia run [manifest]	Interactive REPL (default: ./agent.yaml)
chatixia run --daemon	Autonomous goal loop
chatixia run --once "msg"	Single message, then exit
chatixia run --tick 60	Set daemon tick interval (seconds)
chatixia init [name]	Scaffold a new agent.yaml
chatixia validate [manifest]	Validate manifest and check env vars
chatixia -V	Show version

LLM Providers

Provider	Required env vars
azure	AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_API_KEY, AZURE_OPENAI_DEPLOYMENT
openai	OPENAI_API_KEY
ollama	None (defaults to localhost:11434)

Set provider in agent.yaml and configure the corresponding env vars.

Building for Distribution

cd sdk
make build      # Vendors core modules + builds wheel
make clean      # Remove vendored files and build artifacts

License

MIT