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 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:

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:

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

How to run

Interactive CLI Chat

# 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

# 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

# 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

# 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:

# 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

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:

defaults:
  - /model/base
  - /tools/[email protected]
  - _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:

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:

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

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

Example Output

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