mirror of https://github.com/zenlm/zen-agent.git synced 2026-10-30 14:58:52 +00:00

Files

T

b9e3722ad2 cleanup: refresh SKU links + strip upstream brands (#1 )

Rewrite dead HF links to live successors, refresh stale Zen4 / old Zen-3 chat lineups to current Zen5 canonical + Zen3 specialty, and strip upstream brand mentions from user-facing prose (attribution and base_model chains preserved).

Co-authored-by: zooqueen <[email protected]>

2026-09-11 13:49:59 -07:00

3.1 KiB

Raw Permalink Blame History

Zen Agent — Agentic AI Framework

A Python framework for building AI agents with tool use, planning, memory, and multi-step reasoning capabilities. Powered by Zen models from Zen LM.

Features

Tool Calling: Define custom tools; agents invoke them autonomously
Code Execution: Built-in sandboxed Python interpreter for code tasks
Web Browsing: Agent-driven web search and page retrieval
Multi-step Planning: ReAct and function-calling agent loops
RAG Support: Retrieval-augmented generation over long documents
MCP Integration: Native Model Context Protocol support
GUI: Gradio-based web interface for rapid prototyping

Installation

# Stable release
pip install zen-agent

# With optional extras
pip install "zen-agent[gui,rag,code_interpreter,mcp]"

Quick Start

from zen_agent.agents import Assistant

# Configure the model
llm_cfg = {
    "model": "zenlm/zen-5-coder-gguf",
    "model_server": "http://localhost:8000/v1",
    "api_key": "EMPTY",
    "generate_cfg": {"top_p": 0.8},
}

# Create an agent with tools and files
bot = Assistant(
    llm=llm_cfg,
    system_message="You are a helpful coding assistant.",
    function_list=["code_interpreter"],
    files=["./docs/reference.pdf"],
)

# Run the agent
messages = []
while True:
    query = input("user: ")
    messages.append({"role": "user", "content": query})
    for response in bot.run(messages=messages):
        pass
    messages.extend(response)

Launch Web UI

from zen_agent.gui import WebUI
WebUI(bot).run()

Custom Tool

from zen_agent.tools.base import BaseTool, register_tool

@register_tool("search_docs")
class SearchDocs(BaseTool):
    description = "Search internal documentation for a query."
    parameters = [{"name": "query", "type": "string", "required": True}]

    def call(self, params: str, **kwargs) -> str:
        import json5
        query = json5.loads(params)["query"]
        # ... your search logic ...
        return json5.dumps({"results": []})

Models

Zen Agent works with any OpenAI-compatible endpoint. Recommended models:

Model	Use Case	Size
`zenlm/zen-5-coder-gguf`	Code generation and tool use	4B
`zenlm/zen-5-max-gguf`	Complex multi-step reasoning	671B MoE

Deploy locally with vLLM:

vllm serve zenlm/zen-5-coder-gguf --port 8000

MCP Support

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/files"]
    }
  }
}

See Hanzo MCP for 260+ pre-built MCP tools.

License

Apache 2.0

Zen AI: Clarity Through Intelligence

3.1 KiB Raw Permalink Blame History