# Youtu Agent

> Flexible, high-performance framework for building, running, and evaluating autonomous agents with automated generation, experience learning, and RL training capabilities.

- Skill: `lamm-mit/youtu-agent` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add lamm-mit/youtu-agent`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lamm-mit/youtu-agent/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: lamm-mit (https://skillmd.com/u/lamm-mit)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lamm-mit/youtu-agent

---


## youtu_agent

Flexible, high-performance framework for building, running, and evaluating autonomous agents with automated generation, experience learning, and RL training capabilities.

### Code repository

<https://github.com/TencentCloudADP/youtu-agent>

**Use this as the implementation source:** clone the repo and follow its README for install, dependencies, and how to run code or experiments. The generated client prints JSON with a suggested ``git clone`` command.

### Paper (arXiv — explanation)

<https://arxiv.org/abs/2512.24615>

This is the **paper** reference. The client can optionally fetch live Atom metadata (title, abstract) for agents; it does **not** run training or upstream research code by itself.

### What “running” this client does

The `*_client.py` script prints **JSON** that combines a **GitHub repository** (clone URL + suggested ``git clone``) with **optional paper context** from arXiv (live Atom metadata when **reference_url** is arXiv). Run the real code by cloning the repo and following its README — the skill is your agent-facing entrypoint, not a substitute for the repo’s install steps.

To call a **REST API** instead, set ``BASE_URL`` in `scripts/youtu_agent_client.py` or wrap the upstream CLI with ``subprocess`` after clone.

### How to run the method (from the source)

Extracted for **operators and agents**. Confirm against the upstream repository or paper before relying on it in production.

## Prerequisites

- Python 3.12 or higher
- [uv](https://github.com/astral-sh/uv) package manager (recommended) or pip
- API keys for LLM providers (DeepSeek, OpenAI, etc.)
- Optional: API keys for tools (Serper for web search, Jina for web reading)

## Installation

Clone and set up the repository:

```bash
git clone https://github.com/TencentCloudADP/youtu-agent.git
cd youtu-agent
uv sync  # or, `make sync`
source ./.venv/bin/activate
cp .env.example .env
```

Alternatively, use Docker:

```bash
# Refer to docker/README.md for Docker-based setup with interactive frontend
```

## How to run

### Interactive CLI Chat

```bash
# Basic agent (no internet search)
python scripts/cli_chat.py --config simple/base

# Agent with web search capabilities
python scripts/cli_chat.py --config simple/base_search
```

### Generate Agent Automatically

```bash
# Interactively clarify requirements and auto-generate a config
python scripts/gen_simple_agent.py

# Run the generated config
python scripts/cli_chat.py --config generated/xxx
```

### Run Examples

```bash
# SVG generator (requires SERPER_API_KEY and JINA_API_KEY)
python examples/svg_generator/main.py

# SVG generator with web UI
python examples/svg_generator/main_web.py
```

### Run Evaluations

```bash
# Prepare WebWalkerQA dataset
python scripts/data/process_web_walker_qa.py

# Run evaluation
python scripts/run_eval.py --config_name ww --exp_id <your_exp_id> --dataset WebWalkerQA_15 --concurrency 5
```

## Configuration

### Environment Variables

Edit `.env` file with required API keys:

```bash
# LLM Configuration (OpenAI API format compatible)
UTU_LLM_TYPE=chat.completions
UTU_LLM_MODEL=deepseek-chat
UTU_LLM_BASE_URL=https://api.deepseek.com/v1
UTU_LLM_API_KEY=replace-to-your-api-key

# Optional: Judge LLM (for evaluation)
JUDGE_LLM_TYPE=chat.completions
JUDGE_LLM_MODEL=deepseek-chat
JUDGE_LLM_BASE_URL=https://api.deepseek.com/v1
JUDGE_LLM_API_KEY=replace-to-your-api-key

# Tool APIs (optional, for web search)
SERPER_API_KEY=your-serper-api-key
JINA_API_KEY=your-jina-api-key
```

### Alternative: Tencent Cloud DeepSeek

```bash
UTU_LLM_TYPE=chat.completions
UTU_LLM_MODEL=deepseek-v3
UTU_LLM_BASE_URL=https://api.lkeap.cloud.tencent.com/v1
UTU_LLM_API_KEY=replace-with-your-api-key
```

### Agent Configuration Files

Agent configurations are YAML files in `configs/agents/`. Example structure:

```yaml
defaults:
  - /model/base
  - /tools/search@toolkits.search
  - _self_

agent:
  name: simple-tool-agent
  instructions: "You are a helpful assistant that can search the web."
```

### Web UI Frontend

Download and install the frontend package:

```bash
curl -LO https://github.com/Tencent/Youtu-agent/releases/download/frontend%2Fv0.2.0/utu_agent_ui-0.2.0-py3-none-any.whl
uv pip install utu_agent_ui-0.2.0-py3-none-any.whl
```

Then run web-enabled examples:

```bash
python examples/svg_generator/main_web.py
# Access at http://127.0.0.1:8848/
```

*The same text lives in* ``scripts/USAGE.md`` *for tools that prefer reading files under* ``scripts/``*.*

### Parameters

  --api-key  (str)  [required]  API key for authentication
  --config  (str)  [required]  Path or name of the agent configuration file (YAML). Examples: 'simple/base_search', 'simple/base', 'ww'
  --exp-id  (str)  [optional, default=None]  Experiment identifier for evaluation runs
  --dataset  (str)  [optional, default=None]  Dataset name for evaluation. Examples: 'WebWalkerQA_15', 'GAIA'
  --concurrency  (int)  [optional, default=5]  Number of concurrent evaluation tasks
  --config-name  (str)  [optional, default=None]  Configuration name for evaluation (e.g., 'ww' for WebWalkerQA)

### Usage

```bash
python3 scripts/youtu_agent_client.py --config simple/base_search
```

### Example Output

```json
{"response": "agent_output", "trajectory": [...]}
```