File size: 13,961 Bytes
5b6f681
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
"""Production-ready FastAPI server for sentiment analysis."""

import os
import asyncio
from typing import List, Dict, Any, Optional
from contextlib import asynccontextmanager

from fastapi import FastAPI, HTTPException, BackgroundTasks, File, UploadFile
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel, Field
import uvicorn
import json

from src.inference import SentimentInference
from src.data_utils import load_config
from src.interpretability import InterpretabilityPipeline, AttentionVisualizer
import base64
import io


# Global model instance
inference_pipeline: Optional[SentimentInference] = None
interpretability_pipeline: Optional[InterpretabilityPipeline] = None


@asynccontextmanager
async def lifespan(app: FastAPI):
    """Manage application lifespan - load model on startup."""
    global inference_pipeline, interpretability_pipeline
    
    # Load configuration
    config = load_config()
    
    # Determine model path
    model_path = os.environ.get("MODEL_PATH", "./results")
    if not os.path.exists(model_path):
        model_path = config["model"]["name"]  # Fall back to base model
    
    print(f"🚀 Loading model: {model_path}")
    
    # Initialize inference pipeline
    inference_pipeline = SentimentInference(
        model_path=model_path,
        batch_size=config["api"]["max_batch_size"]
    )
    
    # Initialize interpretability pipeline
    try:
        interpretability_pipeline = InterpretabilityPipeline(model_path)
        print("🔍 Interpretability pipeline loaded!")
    except Exception as e:
        print(f"⚠️  Could not load interpretability pipeline: {e}")
        interpretability_pipeline = None
    
    print("✅ Model loaded successfully!")
    yield
    
    # Cleanup
    print("🧹 Shutting down...")


app = FastAPI(
    title="Sentiment Analysis API",
    description="Production-ready sentiment analysis using Transformer models",
    version="1.0.0",
    lifespan=lifespan
)

# Add CORS middleware
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)


# Pydantic models
class TextInput(BaseModel):
    text: str = Field(..., description="Text to analyze", min_length=1, max_length=10000)


class BatchTextInput(BaseModel):
    texts: List[str] = Field(..., description="List of texts to analyze", min_items=1, max_items=100)


class PredictionResponse(BaseModel):
    text: str
    predicted_label: str
    confidence: float
    model_path: str


class BatchPredictionResponse(BaseModel):
    predictions: List[PredictionResponse]
    total_processed: int


class ProbabilityResponse(BaseModel):
    text: str
    predicted_label: str
    confidence: float
    probability_distribution: Dict[str, float]
    model_path: str


class ModelInfo(BaseModel):
    model_path: str
    device: str
    total_parameters: int
    trainable_parameters: int


class HealthResponse(BaseModel):
    status: str
    model_loaded: bool
    device: str


class InterpretabilityResponse(BaseModel):
    text: str
    predicted_class: int
    confidence: float
    attention_summary_plot: str  # base64 encoded image
    attention_heatmap_plot: str  # base64 encoded image
    shap_explanation: Optional[str] = None  # base64 encoded image if available


class AttentionWeightsResponse(BaseModel):
    text: str
    tokens: List[str]
    attention_weights: List[List[List[List[float]]]]  # [layer][head][seq][seq]
    predicted_class: int
    confidence: float


@app.get("/", response_model=Dict[str, str])
async def root():
    """Root endpoint with API information."""
    return {
        "message": "Sentiment Analysis API",
        "version": "1.0.0",
        "docs": "/docs",
        "health": "/health"
    }


@app.get("/health", response_model=HealthResponse)
async def health_check():
    """Health check endpoint."""
    global inference_pipeline
    
    return HealthResponse(
        status="healthy" if inference_pipeline is not None else "unhealthy",
        model_loaded=inference_pipeline is not None,
        device=inference_pipeline.device if inference_pipeline else "unknown"
    )


@app.post("/predict", response_model=PredictionResponse)
async def predict_sentiment(input_data: TextInput):
    """Predict sentiment for a single text."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        result = inference_pipeline.predict_single(input_data.text)
        return PredictionResponse(**result)
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Prediction failed: {str(e)}")


@app.post("/predict/batch", response_model=BatchPredictionResponse)
async def predict_batch_sentiment(input_data: BatchTextInput):
    """Predict sentiment for multiple texts."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        results = inference_pipeline.predict_batch(input_data.texts)
        predictions = [PredictionResponse(**result) for result in results]
        
        return BatchPredictionResponse(
            predictions=predictions,
            total_processed=len(predictions)
        )
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Batch prediction failed: {str(e)}")


@app.post("/predict/probabilities", response_model=ProbabilityResponse)
async def predict_with_probabilities(input_data: TextInput):
    """Predict sentiment with full probability distribution."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        result = inference_pipeline.predict_with_probabilities(input_data.text)
        return ProbabilityResponse(**result)
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Probability prediction failed: {str(e)}")


@app.post("/predict/file")
async def predict_from_file(file: UploadFile = File(...)):
    """Predict sentiment for texts in uploaded file (one text per line)."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    if not file.filename.endswith(('.txt', '.csv')):
        raise HTTPException(status_code=400, detail="Only .txt and .csv files are supported")
    
    try:
        content = await file.read()
        text_content = content.decode('utf-8')
        
        # Split by lines and filter empty lines
        texts = [line.strip() for line in text_content.split('\n') if line.strip()]
        
        if len(texts) > 1000:
            raise HTTPException(status_code=400, detail="File contains too many texts (max 1000)")
        
        results = inference_pipeline.predict_batch(texts)
        predictions = [PredictionResponse(**result) for result in results]
        
        return BatchPredictionResponse(
            predictions=predictions,
            total_processed=len(predictions)
        )
    except UnicodeDecodeError:
        raise HTTPException(status_code=400, detail="File encoding not supported (use UTF-8)")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"File processing failed: {str(e)}")


@app.get("/model/info", response_model=ModelInfo)
async def get_model_info():
    """Get model information."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        summary = inference_pipeline.get_model_summary()
        return ModelInfo(
            model_path=summary["model_path"],
            device=summary["device"],
            total_parameters=summary["total_parameters"],
            trainable_parameters=summary["trainable_parameters"]
        )
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get model info: {str(e)}")


@app.post("/model/benchmark")
async def benchmark_model(input_data: BatchTextInput, background_tasks: BackgroundTasks):
    """Benchmark model performance."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        benchmark_result = inference_pipeline.benchmark_inference(input_data.texts)
        return benchmark_result
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Benchmark failed: {str(e)}")


@app.get("/model/attention")
async def get_attention_weights(text: str):
    """Get attention weights for interpretability (for debugging/research)."""
    global inference_pipeline
    
    if inference_pipeline is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        result = inference_pipeline.get_attention_weights(text)
        # Convert numpy arrays to lists for JSON serialization
        result["attention_weights"] = [layer.tolist() for layer in result["attention_weights"]]
        return result
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Attention extraction failed: {str(e)}")


@app.post("/interpret", response_model=InterpretabilityResponse)
async def interpret_text(input_data: TextInput):
    """Provide full interpretability analysis for a text."""
    global interpretability_pipeline
    
    if interpretability_pipeline is None:
        raise HTTPException(status_code=503, detail="Interpretability pipeline not available")
    
    try:
        import matplotlib.pyplot as plt
        import tempfile
        import os
        
        # Create temporary directory for plots
        with tempfile.TemporaryDirectory() as temp_dir:
            # Run analysis
            report = interpretability_pipeline.full_analysis(input_data.text, temp_dir)
            
            # Read and encode plots as base64
            def encode_plot(filename):
                plot_path = os.path.join(temp_dir, filename)
                if os.path.exists(plot_path):
                    with open(plot_path, 'rb') as f:
                        plot_data = f.read()
                    return base64.b64encode(plot_data).decode('utf-8')
                return ""
            
            attention_summary = encode_plot("attention_summary.png")
            attention_heatmap = encode_plot("attention_heatmap.png")
            shap_explanation = encode_plot("shap_explanation.png") if os.path.exists(os.path.join(temp_dir, "shap_explanation.png")) else None
            
            return InterpretabilityResponse(
                text=input_data.text,
                predicted_class=report["predicted_class"],
                confidence=report["confidence"],
                attention_summary_plot=attention_summary,
                attention_heatmap_plot=attention_heatmap,
                shap_explanation=shap_explanation
            )
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Interpretability analysis failed: {str(e)}")


@app.post("/interpret/attention", response_model=AttentionWeightsResponse)
async def get_detailed_attention(input_data: TextInput):
    """Get detailed attention weights for visualization."""
    global interpretability_pipeline
    
    if interpretability_pipeline is None:
        raise HTTPException(status_code=503, detail="Interpretability pipeline not available")
    
    try:
        # Get attention weights
        attention_data = interpretability_pipeline.attention_viz.get_attention_weights(input_data.text)
        
        # Get prediction
        import torch
        inputs = interpretability_pipeline.tokenizer(input_data.text, return_tensors="pt", padding=True, truncation=True)
        with torch.no_grad():
            outputs = interpretability_pipeline.model(**inputs)
            predictions = torch.softmax(outputs.logits, dim=-1)
            predicted_class = torch.argmax(predictions, dim=-1).item()
            confidence = predictions[0, predicted_class].item()
        
        # Convert attention weights to lists for JSON serialization
        attention_weights_list = [layer.tolist() for layer in attention_data["attention_weights"]]
        
        return AttentionWeightsResponse(
            text=input_data.text,
            tokens=attention_data["tokens"],
            attention_weights=attention_weights_list,
            predicted_class=predicted_class,
            confidence=confidence
        )
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Attention analysis failed: {str(e)}")


def create_app(model_path: Optional[str] = None) -> FastAPI:
    """Factory function to create FastAPI app with custom model path."""
    if model_path:
        os.environ["MODEL_PATH"] = model_path
    return app


def main():
    """Run the FastAPI server."""
    import argparse
    
    parser = argparse.ArgumentParser(description="Run sentiment analysis API server")
    parser.add_argument("--host", type=str, default="0.0.0.0", help="Host to bind to")
    parser.add_argument("--port", type=int, default=8000, help="Port to bind to")
    parser.add_argument("--model", type=str, help="Path to model")
    parser.add_argument("--reload", action="store_true", help="Enable auto-reload for development")
    parser.add_argument("--workers", type=int, default=1, help="Number of worker processes")
    
    args = parser.parse_args()
    
    # Set model path if provided
    if args.model:
        os.environ["MODEL_PATH"] = args.model
    
    # Run server
    uvicorn.run(
        "src.api:app",
        host=args.host,
        port=args.port,
        reload=args.reload,
        workers=args.workers if not args.reload else 1
    )


if __name__ == "__main__":
    main()