Spaces:
Running on CPU Upgrade
Running on CPU Upgrade
File size: 10,767 Bytes
a56db97 bdbcdab a56db97 bdbcdab a56db97 914a44a a1a2253 914a44a 0127ea2 4d92a9f 7b4e2da 914a44a 7b4e2da 914a44a d13cac7 914a44a d13cac7 914a44a d13cac7 e83f59a d13cac7 914a44a d13cac7 914a44a d13cac7 914a44a d13cac7 914a44a d13cac7 914a44a d13cac7 914a44a d13cac7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 | ---
title: HF Agent
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
hf_oauth: true
hf_oauth_scopes:
- read-repos
- write-repos
- contribute-repos
- manage-repos
- inference-api
- jobs
- write-discussions
---
# HF Agent
An MLE agent CLI with MCP (Model Context Protocol) integration and built-in tool support.
## Quick Start
### Installation
```bash
# Clone the repository
git clone git@github.com:huggingface/hf_agent.git
cd hf_agent
```
#### Install recommended dependencies
```bash
uv sync --extra agent # or uv sync --extra all
```
### Interactive CLI
```bash
uv run python -m agent.main
```
This starts an interactive chat session with the agent. Type your messages and the agent will respond, using tools as needed.
The agent will automatically discover and register all tools from configured MCP servers.
### Env Setup
```bash
ANTHROPIC_API_KEY=<one-key-to-rule-them-all>
HF_TOKEN=<hf-token-to-access-the-hub>
GITHUB_TOKEN=<gh-pat-key-for-not-reinventing-the-wheel>
HF_NAMESPACE=<hf-namespace-to-use>
```
## Architecture
### Component Overview
```
┌─────────────────────────────────────────────────────────────┐
│ User/CLI │
└────────────┬─────────────────────────────────────┬───────────┘
│ User request │ Events
↓ ↑
submission_queue event_queue
│ │
↓ │
┌────────────────────────────────────────────────────┐ │
│ submission_loop (agent_loop.py) │ │
│ ┌──────────────────────────────────────────────┐ │ │
│ │ 1. Receive Operation from queue │ │ │
│ │ 2. Route to Handler (run_agent/compact/...) │ │ │
│ └──────────────────────────────────────────────┘ │ │
│ ↓ │ │
│ ┌──────────────────────────────────────────────┐ │ │
│ │ Handlers.run_agent() │ ├─────────┤
│ │ │ │ Emit │
│ │ ┌────────────────────────────────────────┐ │ │ Events │
│ │ │ Agentic Loop (max 10 iterations) │ │ │ │
│ │ │ │ │ │ │
│ │ │ ┌──────────────────────────────────┐ │ │ │ │
│ │ │ │ Session │ │ │ │ │
│ │ │ │ ┌────────────────────────────┐ │ │ │ │ │
│ │ │ │ │ ContextManager │ │ │ │ │ │
│ │ │ │ │ • Message history │ │ │ │ │ │
│ │ │ │ │ (litellm.Message[]) │ │ │ │ │ │
│ │ │ │ │ • Auto-compaction (180k) │ │ │ │ │ │
│ │ │ │ └────────────────────────────┘ │ │ │ │ │
│ │ │ │ │ │ │ │ │
│ │ │ │ ┌────────────────────────────┐ │ │ │ │ │
│ │ │ │ │ ToolRouter │ │ │ │ │ │
│ │ │ │ │ ├─ explore_hf_docs │ │ │ │ │ │
│ │ │ │ │ ├─ fetch_hf_docs │ │ │ │ │ │
│ │ │ │ │ ├─ find_hf_api │ │ │ │ │ │
│ │ │ │ │ ├─ plan_tool │ │ │ │ │ │
│ │ │ │ │ ├─ hf_jobs* │ │ │ │ │ │
│ │ │ │ │ ├─ hf_private_repos* │ │ │ │ │ │
│ │ │ │ │ ├─ github_* (3 tools) │ │ │ │ │ │
│ │ │ │ │ └─ MCP tools (e.g., │ │ │ │ │ │
│ │ │ │ │ model_search, etc.) │ │ │ │ │ │
│ │ │ │ └────────────────────────────┘ │ │ │ │ │
│ │ │ └──────────────────────────────────┘ │ │ │ │
│ │ │ │ │ │ │
│ │ │ Loop: │ │ │ │
│ │ │ 1. LLM call (litellm.acompletion) │ │ │ │
│ │ │ ↓ │ │ │ │
│ │ │ 2. Parse tool_calls[] │ │ │ │
│ │ │ ↓ │ │ │ │
│ │ │ 3. Execute via ToolRouter │ │ │ │
│ │ │ ↓ │ │ │ │
│ │ │ 4. Add results to ContextManager │ │ │ │
│ │ │ ↓ │ │ │ │
│ │ │ 5. Repeat if tool_calls exist │ │ │ │
│ │ └────────────────────────────────────────┘ │ │ │
│ └──────────────────────────────────────────────┘ │ │
└────────────────────────────────────────────────────┴─────────┘
```
### Agentic Loop Flow
```
User Message
↓
[Add to ContextManager]
↓
╔═══════════════════════════════════════╗
║ Iteration Loop (max 10) ║
║ ║
║ Get messages + tool specs ║
║ ↓ ║
║ litellm.acompletion() ║
║ ↓ ║
║ Has tool_calls? ──No──> Done ║
║ │ ║
║ Yes ║
║ ↓ ║
║ Add assistant msg (with tool_calls) ║
║ ↓ ║
║ For each tool_call: ║
║ • ToolRouter.execute_tool() ║
║ • Add result to ContextManager ║
║ ↓ ║
║ Continue loop ─────────────────┐ ║
║ ↑ │ ║
╚═════════╧═══════════════════════╧═════╝
```
## Project Structure
```
agent/
├── config.py # Configuration models
├── main.py # Interactive CLI entry point
├── prompts/
│ └── system_prompt.yaml # Agent behavior and personality
├── context_manager/
│ └── manager.py # Message history & auto-compaction
└── core/
├── agent_loop.py # Main agent loop and handlers
├── session.py # Session management
├── mcp_client.py # MCP SDK integration
└── tools.py # ToolRouter and built-in tools
configs/
└── main_agent_config.json # Model and MCP server configuration
tests/ # Integration and unit tests
eval/ # Evaluation suite (see eval/README.md)
```
## Events
The agent emits the following events via `event_queue`:
- `processing` - Starting to process user input
- `assistant_message` - LLM response text
- `tool_call` - Tool being called with arguments
- `tool_output` - Tool execution result
- `approval_request` - Requesting user approval for sensitive operations
- `turn_complete` - Agent finished processing
- `error` - Error occurred during processing
- `interrupted` - Agent was interrupted
- `compacted` - Context was compacted
- `undo_complete` - Undo operation completed
- `shutdown` - Agent shutting down
## Development
### Adding Built-in Tools
Edit `agent/core/tools.py`:
```python
def create_builtin_tools() -> list[ToolSpec]:
return [
ToolSpec(
name="your_tool",
description="What your tool does",
parameters={
"type": "object",
"properties": {
"param": {"type": "string", "description": "Parameter description"}
},
"required": ["param"]
},
handler=your_async_handler
),
# ... existing tools
]
```
### Adding MCP Servers
Edit `configs/main_agent_config.json`:
```json
{
"model_name": "anthropic/claude-sonnet-4-5-20250929",
"mcpServers": {
"your-server-name": {
"transport": "http",
"url": "https://example.com/mcp",
"headers": {
"Authorization": "Bearer ${YOUR_TOKEN}"
}
}
}
}
```
Note: Environment variables like `${YOUR_TOKEN}` are auto-substituted from `.env`.
|