Commit
Β·
83fb1b5
1
Parent(s):
ca75c2d
api migration v1
Browse files- DOCKER_SDK_SETUP.md +233 -0
DOCKER_SDK_SETUP.md
ADDED
|
@@ -0,0 +1,233 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Docker SDK Setup Guide
|
| 2 |
+
|
| 3 |
+
## Overview
|
| 4 |
+
|
| 5 |
+
The Research AI Assistant now uses **Docker SDK** instead of Gradio SDK for Hugging Face Spaces deployment. This provides more control and flexibility.
|
| 6 |
+
|
| 7 |
+
## Changes Made
|
| 8 |
+
|
| 9 |
+
### 1. README.md Configuration
|
| 10 |
+
|
| 11 |
+
Changed from:
|
| 12 |
+
```yaml
|
| 13 |
+
sdk: gradio
|
| 14 |
+
app_file: app.py
|
| 15 |
+
```
|
| 16 |
+
|
| 17 |
+
To:
|
| 18 |
+
```yaml
|
| 19 |
+
sdk: docker
|
| 20 |
+
app_port: 7860
|
| 21 |
+
```
|
| 22 |
+
|
| 23 |
+
### 2. Dockerfile Configuration
|
| 24 |
+
|
| 25 |
+
**Key Features:**
|
| 26 |
+
- **Base Image**: Python 3.10-slim
|
| 27 |
+
- **Entry Point**: `main.py` (starts both Gradio and Flask API)
|
| 28 |
+
- **Ports**:
|
| 29 |
+
- 7860: Gradio UI (primary, exposed by HF Spaces)
|
| 30 |
+
- 5001: Flask API (background thread)
|
| 31 |
+
- **Health Check**: Monitors Gradio endpoint
|
| 32 |
+
|
| 33 |
+
## How It Works
|
| 34 |
+
|
| 35 |
+
### Startup Sequence
|
| 36 |
+
|
| 37 |
+
1. **Docker Container Starts**
|
| 38 |
+
- Runs `python main.py`
|
| 39 |
+
|
| 40 |
+
2. **main.py Initializes**
|
| 41 |
+
- Starts Flask API in background thread (port 5001)
|
| 42 |
+
- Creates Gradio interface from `app.py`
|
| 43 |
+
- Launches Gradio on port 7860
|
| 44 |
+
|
| 45 |
+
3. **Both Services Available**
|
| 46 |
+
- Gradio UI: `https://your-space.hf.space` (port 7860)
|
| 47 |
+
- Flask API: Available internally (port 5001)
|
| 48 |
+
|
| 49 |
+
## File Structure
|
| 50 |
+
|
| 51 |
+
```
|
| 52 |
+
Research_AI_Assistant/
|
| 53 |
+
βββ Dockerfile # Main Dockerfile (used by HF Spaces)
|
| 54 |
+
βββ Dockerfile.hf # Alternative/backup Dockerfile
|
| 55 |
+
βββ README.md # HF Spaces config (sdk: docker)
|
| 56 |
+
βββ main.py # Entry point (starts Flask + Gradio)
|
| 57 |
+
βββ app.py # Gradio interface definition
|
| 58 |
+
βββ flask_api.py # Flask API (started by main.py)
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
## Benefits of Docker SDK
|
| 62 |
+
|
| 63 |
+
### 1. **More Control**
|
| 64 |
+
- Full control over container environment
|
| 65 |
+
- Can install system dependencies
|
| 66 |
+
- Custom startup scripts
|
| 67 |
+
|
| 68 |
+
### 2. **Multiple Services**
|
| 69 |
+
- Can run multiple processes (Gradio + Flask)
|
| 70 |
+
- Better resource management
|
| 71 |
+
- Health checks
|
| 72 |
+
|
| 73 |
+
### 3. **Flexibility**
|
| 74 |
+
- Not limited to Gradio's constraints
|
| 75 |
+
- Can add additional services
|
| 76 |
+
- Better debugging
|
| 77 |
+
|
| 78 |
+
### 4. **Production Ready**
|
| 79 |
+
- Standard Docker patterns
|
| 80 |
+
- Easier to migrate to other platforms
|
| 81 |
+
- Better for CI/CD
|
| 82 |
+
|
| 83 |
+
## Deployment
|
| 84 |
+
|
| 85 |
+
### On Hugging Face Spaces
|
| 86 |
+
|
| 87 |
+
1. **Push to Repository**
|
| 88 |
+
```bash
|
| 89 |
+
git push origin main
|
| 90 |
+
```
|
| 91 |
+
|
| 92 |
+
2. **HF Spaces Auto-Builds**
|
| 93 |
+
- Detects `sdk: docker` in README.md
|
| 94 |
+
- Builds Docker image from Dockerfile
|
| 95 |
+
- Exposes port 7860
|
| 96 |
+
|
| 97 |
+
3. **Services Start**
|
| 98 |
+
- Flask API starts in background
|
| 99 |
+
- Gradio UI launches on port 7860
|
| 100 |
+
- Both share orchestrator instance
|
| 101 |
+
|
| 102 |
+
### Local Testing
|
| 103 |
+
|
| 104 |
+
Test the Docker build locally:
|
| 105 |
+
|
| 106 |
+
```bash
|
| 107 |
+
# Build image
|
| 108 |
+
docker build -t research-assistant .
|
| 109 |
+
|
| 110 |
+
# Run container
|
| 111 |
+
docker run -p 7860:7860 -p 5001:5001 \
|
| 112 |
+
-e HF_TOKEN=your_token \
|
| 113 |
+
research-assistant
|
| 114 |
+
```
|
| 115 |
+
|
| 116 |
+
## Port Configuration
|
| 117 |
+
|
| 118 |
+
### Exposed Ports
|
| 119 |
+
|
| 120 |
+
- **7860**: Gradio UI (primary, exposed by HF Spaces)
|
| 121 |
+
- Accessible at: `https://your-space.hf.space`
|
| 122 |
+
|
| 123 |
+
- **5001**: Flask API (internal, background thread)
|
| 124 |
+
- Not directly exposed by HF Spaces
|
| 125 |
+
- Can be accessed via port forwarding if needed
|
| 126 |
+
|
| 127 |
+
### Internal Communication
|
| 128 |
+
|
| 129 |
+
- Flask API runs in same container
|
| 130 |
+
- Both services share Python process
|
| 131 |
+
- Shared orchestrator instance
|
| 132 |
+
- No network overhead
|
| 133 |
+
|
| 134 |
+
## Environment Variables
|
| 135 |
+
|
| 136 |
+
Set in HF Spaces Secrets:
|
| 137 |
+
|
| 138 |
+
```bash
|
| 139 |
+
HF_TOKEN=your_huggingface_token
|
| 140 |
+
ENABLE_FLASK_API=true # Default: true
|
| 141 |
+
FLASK_PORT=5001 # Default: 5001
|
| 142 |
+
FLASK_HOST=0.0.0.0 # Default: 0.0.0.0
|
| 143 |
+
```
|
| 144 |
+
|
| 145 |
+
## Health Checks
|
| 146 |
+
|
| 147 |
+
The Dockerfile includes a health check:
|
| 148 |
+
|
| 149 |
+
```dockerfile
|
| 150 |
+
HEALTHCHECK --interval=30s --timeout=30s --start-period=60s --retries=3 \
|
| 151 |
+
CMD curl -f http://localhost:7860/ || exit 1
|
| 152 |
+
```
|
| 153 |
+
|
| 154 |
+
- Checks Gradio endpoint every 30 seconds
|
| 155 |
+
- 60-second startup period (allows initialization)
|
| 156 |
+
- 3 retries before marking unhealthy
|
| 157 |
+
|
| 158 |
+
## Troubleshooting
|
| 159 |
+
|
| 160 |
+
### Issue: Build Fails
|
| 161 |
+
|
| 162 |
+
**Check:**
|
| 163 |
+
1. Dockerfile syntax is correct
|
| 164 |
+
2. Requirements.txt has all dependencies
|
| 165 |
+
3. Python 3.10 is available
|
| 166 |
+
|
| 167 |
+
**Solution:**
|
| 168 |
+
```bash
|
| 169 |
+
# Test build locally
|
| 170 |
+
docker build -t test-build .
|
| 171 |
+
```
|
| 172 |
+
|
| 173 |
+
### Issue: Port Not Accessible
|
| 174 |
+
|
| 175 |
+
**Check:**
|
| 176 |
+
1. `app_port: 7860` in README.md
|
| 177 |
+
2. Gradio is listening on 0.0.0.0:7860
|
| 178 |
+
3. No firewall blocking
|
| 179 |
+
|
| 180 |
+
**Solution:**
|
| 181 |
+
- Verify in logs: "Running on local URL: http://0.0.0.0:7860"
|
| 182 |
+
|
| 183 |
+
### Issue: Flask API Not Starting
|
| 184 |
+
|
| 185 |
+
**Check:**
|
| 186 |
+
1. Logs show Flask initialization
|
| 187 |
+
2. `ENABLE_FLASK_API=true` (default)
|
| 188 |
+
3. Port 5001 not conflicting
|
| 189 |
+
|
| 190 |
+
**Solution:**
|
| 191 |
+
- Check logs for "Starting Flask API in background thread..."
|
| 192 |
+
- Verify Flask thread starts successfully
|
| 193 |
+
|
| 194 |
+
## Comparison: Gradio SDK vs Docker SDK
|
| 195 |
+
|
| 196 |
+
| Feature | Gradio SDK | Docker SDK |
|
| 197 |
+
|---------|------------|------------|
|
| 198 |
+
| **Setup** | Simple | More control |
|
| 199 |
+
| **Multiple Services** | β Limited | β
Full control |
|
| 200 |
+
| **System Dependencies** | β οΈ Limited | β
Full access |
|
| 201 |
+
| **Customization** | β οΈ Limited | β
Complete |
|
| 202 |
+
| **Debugging** | β οΈ Harder | β
Easier |
|
| 203 |
+
| **Port Configuration** | β οΈ Fixed | β
Flexible |
|
| 204 |
+
| **Health Checks** | β No | β
Yes |
|
| 205 |
+
|
| 206 |
+
## Migration Notes
|
| 207 |
+
|
| 208 |
+
### From Gradio SDK
|
| 209 |
+
|
| 210 |
+
If migrating from Gradio SDK:
|
| 211 |
+
|
| 212 |
+
1. β
**README.md**: Changed `sdk: gradio` β `sdk: docker`
|
| 213 |
+
2. β
**Dockerfile**: Created/updated to use `main.py`
|
| 214 |
+
3. β
**Entry Point**: Changed from `app.py` β `main.py`
|
| 215 |
+
4. β
**Ports**: Explicitly configured (7860, 5001)
|
| 216 |
+
|
| 217 |
+
### What Still Works
|
| 218 |
+
|
| 219 |
+
- β
All Gradio functionality preserved
|
| 220 |
+
- β
All API endpoints work
|
| 221 |
+
- β
Flask API integration works
|
| 222 |
+
- β
Orchestrator initialization unchanged
|
| 223 |
+
|
| 224 |
+
## Summary
|
| 225 |
+
|
| 226 |
+
β
**Docker SDK**: More control and flexibility
|
| 227 |
+
β
**Dual Services**: Gradio + Flask API
|
| 228 |
+
β
**Production Ready**: Standard Docker patterns
|
| 229 |
+
β
**Health Checks**: Built-in monitoring
|
| 230 |
+
β
**Flexible**: Easy to extend and customize
|
| 231 |
+
|
| 232 |
+
The application now uses Docker SDK for better control and flexibility while maintaining all existing functionality.
|
| 233 |
+
|