agenai

Agenai

A modular, extensible AI agent system built around a core loop of perceive -> reason -> act.

Features

• Modular Architecture: Build specialized agents with custom tools
• Multiple LLM Providers: Anthropic Claude, Ollama (local, free)
• Web Interface: FastAPI backend with WebSocket chat
• Built-in Agents: Teaching Assistant for teaching Python programming
• Hooks System: Full observability into agent behavior

Installation

# From PyPI
pip install agenai

# With web server support (for chat UI)
pip install agenai[server]

# For development (editable install)
pip install -e ".[dev]"

Quick Start

Option 1: Using Ollama (Free, Local)

No API key required! Uses local LLMs.

# 1. Install Ollama from https://ollama.ai

# 2. Pull a model
ollama pull llama3.1

# 3. Start the web server
agenai-server

# 4. Open http://localhost:8000 in your browser

Option 2: Using Claude API

# Set your API key
export ANTHROPIC_API_KEY="your-key"

# Run CLI
agenai run "Explain Python for loops" -v

# Or interactive chat
agenai chat

Web Chat Interface

Start the server with WebSocket support:

# Install server dependencies
pip install -e ".[server]"

# Start server
agenai-server --port 8000

# Or with auto-reload for development
agenai-server --reload

Open http://localhost:8000 for the chat UI.

API Endpoints

Creating Custom Agents

from agenai.agent import SpecializedAgent, OllamaProvider
from agenai.agent.tools import Tool, ToolResult

class MyTool(Tool):
    @property
    def name(self) -> str:
        return "my_tool"

    @property
    def description(self) -> str:
        return "Does something useful"

    @property
    def parameters(self) -> dict:
        return {"type": "object", "properties": {}}

    def execute(self) -> ToolResult:
        return ToolResult.ok("Done!")

class MyAgent(SpecializedAgent):
    name = "my_agent"
    description = "My custom agent"

    def get_system_prompt(self) -> str:
        return "You are a helpful assistant..."

    def get_tools(self) -> list:
        return [MyTool()]

# Use with Ollama (free, local)
agent = MyAgent(llm=OllamaProvider(model="llama3.1"))
result = agent.run("Hello!")

Built-in Agents

Teaching Assistant

Teaching assistant for Python programming:

from agenai.agents.teaching_assistant import TeachingAssistantAgent

agent = TeachingAssistantAgent(difficulty_level="beginner")
result = agent.run("Explain what a for loop is")

Tools: run_code, review_code, create_exercise, create_quiz

Architecture

┌─────────────┐     WebSocket      ┌─────────────┐              ┌─────────────┐
│   Frontend  │ <================> │   Backend   │ <==========> │   Ollama    │
│  (Web Chat) │                    │  (FastAPI)  │              │ (Local LLM) │
└─────────────┘                    └──────┬──────┘              └─────────────┘
                                          │
                                          ▼
                                   ┌─────────────┐
                                   │   Agenai    │
                                   │   Agents    │
                                   └─────────────┘

See AI-AGENT-ARCHITECTURE.md for full details.

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest tests/ -v

# Format code
ruff check --fix .

Publishing to PyPI

Local Publishing

# Build only
./scripts/publish.sh --build-only

# Publish to TestPyPI
TEST_PYPI_TOKEN=xxx ./scripts/publish.sh --test

# Publish to PyPI
PYPI_TOKEN=xxx ./scripts/publish.sh

Automated Publishing (GitHub Actions)

The repo includes a GitHub Actions workflow that auto-publishes using Trusted Publisher (OIDC - no API tokens needed).

How it works:

• Create a GitHub Release → auto-publishes to PyPI
• Manual dispatch → choose testpypi or pypi target

Step 1: Setup Trusted Publisher on PyPI

1. Go to https://pypi.org/manage/project/agenai/settings/publishing/

2. Click "Add a new publisher"

3. Fill in:

   Field	Value
   Owner	tautobot
   Repository	agenai
   Workflow name	publish.yml
   Environment name	pypi

Step 2: Setup Trusted Publisher on TestPyPI

1. Go to https://test.pypi.org/manage/project/agenai/settings/publishing/
2. Same fields as above, but Environment name: testpypi

Step 3: Setup GitHub Environments

1. Go to https://github.com/tautobot/agenai/settings/environments
2. Create environment pypi
3. Create environment testpypi
4. (Optional) Add protection rules like required reviewers

Releasing a New Version

1. Update version in pyproject.toml
2. Commit and push
3. Create a GitHub Release with tag v0.x.x
4. GitHub Actions auto-publishes to PyPI

License

MIT