agent-base

内容来源：README.md（说明文档） · 原始地址 · 查看安装指南

原始内容

# Agent Base

A functional agent base for building AI agents with multi-provider LLM support and built-in observability.

[![Python 3.12+](https://img.shields.io/badge/python-3.12+-blue.svg)](https://www.python.org/downloads/)
[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)

## Overview

Build conversational AI agents with enterprise-grade features: session persistence, conversation memory, observability, and extensible toolsets.

Supports Local (Docker Models), GitHub Models, OpenAI, Anthropic, Google Gemini, Azure OpenAI, and Azure AI Foundry.

```bash
agent

Agent - Conversational Assistant
Version 0.1.0 • Local/ai/phi4

> Say hello to Alice

● Thinking...

Hello, Alice! ◉‿◉

> What was the name I just mentioned?

● Thinking...

You mentioned "Alice."

> exit
Saved 4 memories
Session auto-saved as '2025-11-10-11-38-07'

Goodbye!
```

## Prerequisites

### Required

- Python 3.12+
- [uv](https://docs.astral.sh/uv/getting-started/installation/) package manager


### LLM Providers

**Local (Docker Models)**  
Requires [Docker Desktop](https://www.docker.com/products/docker-desktop/) for local model serving.

> **Note:** Docker-based local models work on Windows, but performance is significantly slower due to Windows virtualization overhead. For best results, use a hosted provider or run Docker only on Linux/macOS.

**Hosted Providers**

| Provider | Auth Method |
|----------|-------------|
| GitHub Models | GitHub CLI (`gh auth login`) |
| OpenAI | API Key |
| Anthropic | API Key |
| Google Gemini | API Key |
| Azure OpenAI | Azure CLI (`az login`) |
| Azure AI Foundry | Azure CLI (`az login`) |

#### Azure Provider Prerequisites

Use these when selecting Azure OpenAI or Azure AI Foundry providers:

- [Azure CLI](https://learn.microsoft.com/cli/azure/install-azure-cli) – authentication (`az login`)
- [Azure Application Insights](https://learn.microsoft.com/azure/azure-monitor/app/app-insights-overview) – observability (optional but recommended) 


## Quick Setup

```bash
# 1. Install agent
uv tool install --prerelease=allow git+https://github.com/danielscholl/agent-base.git

# 2. Start agent
agent

# Get help
agent --help
```

### Configuration

Agent uses a JSON configuration file at `~/.agent/settings.json` for managing providers, memory, and observability settings.

**Configuration Commands:**

```bash
# Interactive setup wizard
agent config init

# View current configuration
agent config show

# Manage providers
agent config provider local       # Enable/configure local (Docker)
agent config provider github      # Enable/configure GitHub Models
agent config provider openai      # Enable/configure OpenAI

# Configure memory backend
agent config memory               # Switch between in_memory and mem0
```

See [docs/design/config.md](https://github.com/danielscholl/agent-base/blob/HEAD/docs/design/config.md) for complete configuration.

## Usage

```bash
# Interactive chat mode
agent

# Check the agent configuration
agent --check

# Check the tools being exposed to the agent
agent --tools

# Single query (clean output for scripting)
agent -p "Say hello to Alice"

# Single query with verbose execution details
agent -p "Analyze this text" --verbose

# Switch providers on the fly
agent --provider openai -p "Hello"

# Switch models on the fly
agent --provider anthropic --model claude-sonnet-4-5-20250929 -p "Hello"
```

**Note:** Single prompt mode (`-p`) outputs clean text by default, perfect for piping or scripting. Use `--verbose` to see execution details.

### Observability

Monitor your agent's performance with Telemetry:

```bash
# Start local dashboard (requires Docker)
# View at http://localhost:18888
agent --telemetry start
```

See [docs/design/usage.md](https://github.com/danielscholl/agent-base/blob/HEAD/docs/design/usage.md) for complete examples.

## Skills

Skills are lightweight extensions that add domain-specific capabilities to an agent without increasing the core footprint. They’re automatically discovered at runtime and come in two forms: bundled skills that ship with the agent, and plugin skills that can install from any git repository. Each skill loads only its minimal metadata by default, with additional tools or scripts activated on demand to keep token usage low.

### Installing Skills

Agent Base supports installing skills directly from any git repository:

```bash
agent skill install <git-url>
```

Once installed, skills are automatically discovered and integrated into the agent runtime.

### Managing Skills

Use the built-in management commands to view, enable/disable, or update installed skills:

```bash
agent skill show
agent skill manage
```

See [SKILLS.md](https://github.com/danielscholl/agent-base/blob/HEAD/docs/design/skills.md) for additional details and information on constructing skills.


## Contributing

See [CONTRIBUTING.md](https://github.com/danielscholl/agent-base/blob/HEAD/CONTRIBUTING.md) for development setup, code quality guidelines, and contribution workflow.

## License

This project is licensed under the MIT License - see the [LICENSE](https://github.com/danielscholl/agent-base/tree/HEAD/LICENSE) file for details.

## Acknowledgments

- Built with [Microsoft Agent Framework](https://github.com/microsoft/agent-framework)
- Powered by [Azure AI Foundry](https://azure.microsoft.com/en-us/products/ai-services/ai-studio)

Agent Base

A functional agent base for building AI agents with multi-provider LLM support and built-in observability.

Overview

Build conversational AI agents with enterprise-grade features: session persistence, conversation memory, observability, and extensible toolsets.

Supports Local (Docker Models), GitHub Models, OpenAI, Anthropic, Google Gemini, Azure OpenAI, and Azure AI Foundry.

agent

Agent - Conversational Assistant
Version 0.1.0 • Local/ai/phi4

> Say hello to Alice

● Thinking...

Hello, Alice! ◉‿◉

> What was the name I just mentioned?

● Thinking...

You mentioned "Alice."

> exit
Saved 4 memories
Session auto-saved as '2025-11-10-11-38-07'

Goodbye!

Prerequisites

Required

Python 3.12+
uv package manager

LLM Providers

Local (Docker Models)
Requires Docker Desktop for local model serving.

Note: Docker-based local models work on Windows, but performance is significantly slower due to Windows virtualization overhead. For best results, use a hosted provider or run Docker only on Linux/macOS.

Hosted Providers

Provider	Auth Method
GitHub Models	GitHub CLI (`gh auth login`)
OpenAI	API Key
Anthropic	API Key
Google Gemini	API Key
Azure OpenAI	Azure CLI (`az login`)
Azure AI Foundry	Azure CLI (`az login`)

Azure Provider Prerequisites

Use these when selecting Azure OpenAI or Azure AI Foundry providers:

Azure CLI – authentication (az login)
Azure Application Insights – observability (optional but recommended)

Quick Setup

# 1. Install agent
uv tool install --prerelease=allow git+https://github.com/danielscholl/agent-base.git

# 2. Start agent
agent

# Get help
agent --help

Configuration

Agent uses a JSON configuration file at ~/.agent/settings.json for managing providers, memory, and observability settings.

Configuration Commands:

# Interactive setup wizard
agent config init

# View current configuration
agent config show

# Manage providers
agent config provider local       # Enable/configure local (Docker)
agent config provider github      # Enable/configure GitHub Models
agent config provider openai      # Enable/configure OpenAI

# Configure memory backend
agent config memory               # Switch between in_memory and mem0

See docs/design/config.md for complete configuration.

Usage

# Interactive chat mode
agent

# Check the agent configuration
agent --check

# Check the tools being exposed to the agent
agent --tools

# Single query (clean output for scripting)
agent -p "Say hello to Alice"

# Single query with verbose execution details
agent -p "Analyze this text" --verbose

# Switch providers on the fly
agent --provider openai -p "Hello"

# Switch models on the fly
agent --provider anthropic --model claude-sonnet-4-5-20250929 -p "Hello"

Note: Single prompt mode (-p) outputs clean text by default, perfect for piping or scripting. Use --verbose to see execution details.

Observability

Monitor your agent's performance with Telemetry:

# Start local dashboard (requires Docker)
# View at http://localhost:18888
agent --telemetry start

See docs/design/usage.md for complete examples.

Skills

Skills are lightweight extensions that add domain-specific capabilities to an agent without increasing the core footprint. They’re automatically discovered at runtime and come in two forms: bundled skills that ship with the agent, and plugin skills that can install from any git repository. Each skill loads only its minimal metadata by default, with additional tools or scripts activated on demand to keep token usage low.

Installing Skills

Agent Base supports installing skills directly from any git repository:

agent skill install <git-url>

Once installed, skills are automatically discovered and integrated into the agent runtime.

Managing Skills

Use the built-in management commands to view, enable/disable, or update installed skills:

agent skill show
agent skill manage

See SKILLS.md for additional details and information on constructing skills.

Contributing

See CONTRIBUTING.md for development setup, code quality guidelines, and contribution workflow.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Built with Microsoft Agent Framework
Powered by Azure AI Foundry