Skip to content

Commit 2d4ae87

Browse files
ZahinTapadarZahinTapadar
committed
winidows specific fix
1 parent e9e7eb4 commit 2d4ae87

7 files changed

Lines changed: 895 additions & 9 deletions

File tree

‎.github/workflows/build-release.yml‎

Whitespace-only changes.

‎.github/workflows/release.yml‎

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -59,13 +59,13 @@ jobs:
5959
- name: Install frontend dependencies
6060
run: npm install
6161

62-
- name: Build the app
63-
run: |
64-
if [ "${{ matrix.platform }}" = "macos-latest" ]; then
65-
npm run tauri build -- --target universal-apple-darwin
66-
else
67-
npm run tauri build
68-
fi
62+
- name: Build the app (macOS)
63+
if: matrix.platform == 'macos-latest'
64+
run: npm run tauri build -- --target universal-apple-darwin
65+
66+
- name: Build the app (Windows/Linux)
67+
if: matrix.platform != 'macos-latest'
68+
run: npm run tauri build
6969

7070
- name: Upload artifacts (windows only)
7171
if: matrix.platform == 'windows-latest'

‎README.md‎

Lines changed: 22 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -66,6 +66,9 @@ https://github.com/user-attachments/assets/8ed11232-de9c-469b-b332-143ca41daf15
6666
## ✨ Features
6767

6868
### Current Features
69+
- **🔍 Intelligent Web Search** - Real-time internet search with SearxNG integration
70+
- **🧠 Thinking Mode Control** - Toggle AI reasoning traces on/off
71+
- **🌐 Multi-Engine Fallback** - Multiple SearxNG instances for reliability
6972
- **🎬 Animated Shine Borders** - Eye-catching animated message borders with color cycling
7073
- **📱 Responsive Design** - Mobile-first approach with seamless cross-device compatibility
7174
- **🌙 Theme System** - Dark/light mode with system preference detection
@@ -116,11 +119,28 @@ ollama pull llama2
116119
ollama pull codellama
117120
ollama pull mistral
118121

122+
# For web search feature, also pull:
123+
ollama pull qwen3:0.6b
124+
119125
# Verify installation
120126
ollama list
121127
```
122128

123-
### Step 3: Install BeautifyOllama
129+
### Step 3: Setup Web Search (Optional)
130+
131+
For enhanced web search capabilities, set up a local SearxNG instance:
132+
133+
```bash
134+
# Quick setup with provided script
135+
./setup-searxng.sh
136+
137+
# Or manually install Python dependencies
138+
pip install ollama requests
139+
```
140+
141+
For detailed web search setup, see [Web Search Integration Guide](WEB_SEARCH_INTEGRATION.md).
142+
143+
### Step 4: Install BeautifyOllama
124144

125145
```bash
126146
# Clone the repository
@@ -142,7 +162,7 @@ yarn dev
142162
pnpm dev
143163
```
144164

145-
### Step 4: Access the Application
165+
### Step 5: Access the Application
146166

147167
Open your browser and navigate to [http://localhost:3000](http://localhost:3000)
148168

‎SEARCH_API_DOCS.md‎

Lines changed: 300 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,300 @@
1+
# Search API Documentation
2+
3+
## Overview
4+
5+
The `search_api.py` module provides a command-line interface for the Ollama Web Search functionality. This script is called by the Rust backend to perform web searches and return structured JSON results.
6+
7+
## Usage
8+
9+
### Basic Search
10+
11+
```bash
12+
python3 ollama-web-search/main.py --query "What is the capital of France?" --json
13+
```
14+
15+
### With Thinking Mode
16+
17+
```bash
18+
python3 ollama-web-search/main.py --query "Latest AI developments" --json --thinking
19+
```
20+
21+
### Interactive Mode
22+
23+
```bash
24+
python3 ollama-web-search/main.py
25+
```
26+
27+
## Command Line Arguments
28+
29+
| Argument | Description | Default |
30+
|----------|-------------|---------|
31+
| `--query` | Search query string | Required in JSON mode |
32+
| `--json` | Output results in JSON format | False |
33+
| `--thinking` | Enable AI thinking mode | False |
34+
| `--model` | Ollama model to use | From config |
35+
| `--history` | Show search history | False |
36+
| `--config` | Show configuration | False |
37+
38+
## JSON Response Format
39+
40+
### Successful Response
41+
42+
```json
43+
{
44+
"success": true,
45+
"user_query": "What is the capital of France?",
46+
"search_queries": ["capital France"],
47+
"summaries": [
48+
{
49+
"query": "capital France",
50+
"summary": "Paris is the capital and largest city of France...",
51+
"sources": ["https://en.wikipedia.org/wiki/Paris"],
52+
"title": "Paris - Wikipedia"
53+
}
54+
],
55+
"sources": ["https://en.wikipedia.org/wiki/Paris"],
56+
"error": null
57+
}
58+
```
59+
60+
### Error Response
61+
62+
```json
63+
{
64+
"success": false,
65+
"user_query": "search query",
66+
"search_queries": [],
67+
"summaries": [],
68+
"sources": [],
69+
"error": "Error description"
70+
}
71+
```
72+
73+
## Configuration
74+
75+
### Configuration File (`config.json`)
76+
77+
```json
78+
{
79+
"model": "qwen3:0.6b",
80+
"searxng_instances": [
81+
"http://localhost:32768",
82+
"https://search.inetol.net/search",
83+
"https://searx.be/search",
84+
"https://search.brave4u.com/search",
85+
"https://priv.au/search"
86+
],
87+
"max_results": 8,
88+
"timeout": 10,
89+
"max_retries": 3,
90+
"history_file": "search_history.json",
91+
"enable_colors": true,
92+
"streaming_delay": 0.02
93+
}
94+
```
95+
96+
### Environment Variables
97+
98+
| Variable | Description | Default |
99+
|----------|-------------|---------|
100+
| `OLLAMA_HOST` | Ollama server host | `localhost:11434` |
101+
102+
## API Classes and Methods
103+
104+
### WebSearchAssistant
105+
106+
Main class that handles web search operations.
107+
108+
#### Methods
109+
110+
##### `generate_search_query(question: str, thinking: bool = False) -> Optional[str]`
111+
112+
Generates an optimized search query from a user question.
113+
114+
**Parameters:**
115+
- `question`: User's original question
116+
- `thinking`: Enable thinking mode for query generation
117+
118+
**Returns:** Optimized search query string or None if generation fails
119+
120+
##### `browse_web(query: str) -> Optional[List[Dict]]`
121+
122+
Searches the web using configured SearxNG instances.
123+
124+
**Parameters:**
125+
- `query`: Search query string
126+
127+
**Returns:** List of search results or None if all instances fail
128+
129+
##### `select_best_result(question: str, query: str, results: List[Dict], thinking: bool = False) -> Optional[Tuple[str, str]]`
130+
131+
Uses AI to select the most relevant search result.
132+
133+
**Parameters:**
134+
- `question`: Original user question
135+
- `query`: Generated search query
136+
- `results`: List of search results
137+
- `thinking`: Enable thinking mode
138+
139+
**Returns:** Tuple of (title, url) or None if selection fails
140+
141+
##### `retrieve_page_information(url: str) -> Optional[str]`
142+
143+
Retrieves and cleans webpage content using Jina Reader API.
144+
145+
**Parameters:**
146+
- `url`: URL to extract content from
147+
148+
**Returns:** Cleaned webpage content or None if extraction fails
149+
150+
##### `model_response(model: str, message: str, max_retries: int = 3, thinking: bool = False) -> Optional[str]`
151+
152+
Gets response from Ollama model with thinking mode control.
153+
154+
**Parameters:**
155+
- `model`: Ollama model name
156+
- `message`: Message to send to model
157+
- `max_retries`: Maximum retry attempts
158+
- `thinking`: Enable thinking mode
159+
160+
**Returns:** Model response or None if all retries fail
161+
162+
## Search Flow
163+
164+
1. **Query Generation**: Convert user question to optimized search query
165+
2. **Web Search**: Search using SearxNG instances with fallback
166+
3. **Result Selection**: AI selects most relevant result
167+
4. **Content Extraction**: Retrieve webpage content via Jina Reader
168+
5. **Summarization**: Generate AI summary of extracted content
169+
6. **Response Formatting**: Return structured JSON response
170+
171+
## Error Handling
172+
173+
### Common Error Types
174+
175+
- **Ollama Connection Error**: Cannot connect to Ollama server
176+
- **Search Instance Failure**: All SearxNG instances failed
177+
- **Content Extraction Timeout**: Jina Reader API timeout
178+
- **Model Response Error**: Ollama model response failure
179+
180+
### Retry Logic
181+
182+
- **Search Instances**: Automatic fallback to next instance
183+
- **Model Responses**: Exponential backoff retry (up to 3 attempts)
184+
- **Content Extraction**: Single attempt with timeout
185+
186+
## Integration with Rust Backend
187+
188+
### Rust Command Execution
189+
190+
```rust
191+
let mut cmd = Command::new("python3");
192+
cmd.arg("ollama-web-search/main.py")
193+
.arg("--query")
194+
.arg(&query)
195+
.arg("--json");
196+
197+
if thinking_mode {
198+
cmd.arg("--thinking");
199+
}
200+
201+
let output = cmd.output()?;
202+
```
203+
204+
### Response Parsing
205+
206+
```rust
207+
let response = String::from_utf8_lossy(&output.stdout);
208+
let json: serde_json::Value = serde_json::from_str(&response)?;
209+
210+
if json.get("success").and_then(|v| v.as_bool()).unwrap_or(false) {
211+
// Handle successful response
212+
format_search_results_from_python(&json, &query)
213+
} else {
214+
// Handle error
215+
let error = json.get("error").and_then(|v| v.as_str()).unwrap_or("Unknown error");
216+
Err(format!("Python search failed: {}", error))
217+
}
218+
```
219+
220+
## Testing
221+
222+
### Unit Tests
223+
224+
```bash
225+
# Test search query generation
226+
python3 -c "
227+
from main import WebSearchAssistant
228+
assistant = WebSearchAssistant()
229+
query = assistant.generate_search_query('What is Python?')
230+
print(f'Generated query: {query}')
231+
"
232+
```
233+
234+
### Integration Tests
235+
236+
```bash
237+
# Test full search flow
238+
python3 ollama-web-search/main.py --query "test query" --json
239+
240+
# Test thinking mode
241+
python3 ollama-web-search/main.py --query "test query" --json --thinking
242+
```
243+
244+
### SearxNG Connectivity Test
245+
246+
```bash
247+
# Test local instance
248+
curl "http://localhost:32768?q=test&format=json"
249+
250+
# Test public instance
251+
curl "https://search.inetol.net/search?q=test&format=json"
252+
```
253+
254+
## Performance Considerations
255+
256+
### Optimization Tips
257+
258+
1. **Use Local SearxNG**: Significantly faster than public instances
259+
2. **Adjust Timeouts**: Balance between speed and reliability
260+
3. **Limit Results**: Reduce `max_results` for faster processing
261+
4. **Model Selection**: Smaller models respond faster
262+
263+
### Resource Usage
264+
265+
- **Memory**: ~50MB for Python process
266+
- **Network**: Dependent on search instances and content extraction
267+
- **CPU**: Model inference for query generation and summarization
268+
269+
## Security Notes
270+
271+
### Data Flow
272+
273+
- User queries are sent to configured SearxNG instances
274+
- Webpage content is extracted via Jina Reader API
275+
- All processing happens locally except for web requests
276+
277+
### Privacy
278+
279+
- No logging of user queries (with proper SearxNG configuration)
280+
- Search history stored locally only
281+
- Can be configured to use local SearxNG instance only
282+
283+
## Contributing
284+
285+
### Adding Features
286+
287+
1. Fork the repository
288+
2. Add new methods to `WebSearchAssistant` class
289+
3. Update JSON response format if needed
290+
4. Add corresponding tests
291+
5. Update documentation
292+
293+
### Reporting Issues
294+
295+
Please include:
296+
- Python version
297+
- Ollama version and models
298+
- SearxNG instance details
299+
- Error messages and logs
300+
- Steps to reproduce

0 commit comments

Comments
 (0)