File size: 26,967 Bytes
ab2012f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ce3e778
 
 
ab2012f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ce3e778
ab2012f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ce3e778
ab2012f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ce3e778
ab2012f
 
 
 
 
 
 
 
 
 
 
 
ce3e778
 
 
 
 
 
 
 
 
ab2012f
 
 
ce3e778
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ab2012f
 
 
 
 
 
 
ce3e778
ab2012f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
import json
import base64
from typing import List, Optional
from fastapi import HTTPException, UploadFile, File, Form
from fastapi.responses import StreamingResponse
from models import ChatRequest, ChatResponse, WardrobeItem, TextRequest
from query_processing import (
    extract_clothing_info, extract_colors_from_query, detect_query_type,
    get_color_matches, is_greeting, is_name_question
)
from conversation import (
    get_conversation_context, enhance_message_with_context, update_context
)
from model_manager import generate_chat_response, generate_chat_response_streaming
from wardrobe import handle_wardrobe_chat
from rag import retrieve_relevant_context, format_rag_context
from config import COLOR_HARMONY
from model_manager import style_model

def setup_routes(app):
    @app.get("/")
    async def root():
        return {
            "message": "Style GPT API - Milestone 1",
            "version": "1.0.0",
            "endpoints": {
                "/text": "POST - Text-only chat",
                "/chat": "POST - Chat with optional images",
                "/chat/upload": "POST - Chat with file upload",
                "/chat/upload/stream": "POST - Streaming chat with file upload",
                "/health": "GET - Health check"
            }
        }

    @app.get("/health")
    async def health_check():
        return {
            "status": "healthy" if style_model is not None else "loading",
            "model_loaded": style_model is not None,
            "model_name": "Qwen/Qwen2.5-VL-7B-Instruct"
        }

    @app.post("/text", response_model=ChatResponse)
    async def text_only(request: TextRequest):
        try:
            message = request.message.strip()
            session_id = request.session_id
            
            if not message:
                raise HTTPException(status_code=400, detail="Message cannot be empty")
            
            conv_context = get_conversation_context(session_id)
            
            if is_name_question(message):
                prompt = "What is your name? Respond naturally and friendly."
                rag_chunks = retrieve_relevant_context(message, top_k=2)
                rag_context = format_rag_context(rag_chunks)
                response_text = generate_chat_response(prompt, max_length=100, temperature=0.8, rag_context=rag_context, images=None)
                update_context(session_id, message, {"response": response_text})
                return ChatResponse(response=response_text, session_id=session_id)
            
            if is_greeting(message):
                prompt = f"{message} Respond warmly and offer to help with fashion advice."
                rag_chunks = retrieve_relevant_context(message, top_k=2)
                rag_context = format_rag_context(rag_chunks)
                response_text = generate_chat_response(prompt, max_length=150, temperature=0.8, rag_context=rag_context, images=None)
                update_context(session_id, message, {"response": response_text})
                return ChatResponse(response=response_text, session_id=session_id)
            
            enhanced_message = enhance_message_with_context(message, conv_context["context"])
            query_type = detect_query_type(enhanced_message)
            rag_chunks = retrieve_relevant_context(enhanced_message, top_k=3)
            rag_context = format_rag_context(rag_chunks)
            
            if query_type == "color_compatibility":
                found_colors = extract_colors_from_query(enhanced_message)
                
                if len(found_colors) >= 2:
                    color1_mapped = found_colors[0][1]
                    color2_mapped = found_colors[1][1]
                    color1_original = found_colors[0][0]
                    color2_original = found_colors[1][0]
                    
                    compatible = False
                    if color1_mapped in COLOR_HARMONY:
                        compatible = color2_mapped in COLOR_HARMONY[color1_mapped]
                    elif color2_mapped in COLOR_HARMONY:
                        compatible = color1_mapped in COLOR_HARMONY[color2_mapped]
                    
                    neutrals = ["white", "black", "grey", "gray", "beige", "navy"]
                    if color1_mapped in neutrals or color2_mapped in neutrals:
                        compatible = True
                    
                    if compatible:
                        response_text = f"Yes, {color1_original.title()} will go well with {color2_original.title()}. They create a balanced and stylish combination that works great together!"
                    else:
                        response_text = f"{color1_original.title()} and {color2_original.title()} can work together, though you might want to add some neutral pieces to balance the look."
                    
                    prompt = f"Does {color1_original} go well with {color2_original}? Answer naturally and conversationally."
                    ai_response = generate_chat_response(prompt, max_length=150, temperature=0.8, rag_context=rag_context, images=None)
                    if len(ai_response) > 15:
                        response_text = ai_response
                    
                    update_context(session_id, message, {
                        "response": response_text,
                        "color": color1_original,
                        "colors": [color1_original, color2_original]
                    })
                    
                    return ChatResponse(
                        response=response_text,
                        session_id=session_id
                    )
            
            elif query_type == "color_suggestion":
                clothing_info = extract_clothing_info(enhanced_message)
                base_color = clothing_info.get("color")
                
                if not base_color:
                    found_colors = extract_colors_from_query(enhanced_message)
                    if found_colors:
                        base_color = found_colors[0][1]
                    elif conv_context["context"].get("last_color"):
                        base_color = conv_context["context"]["last_color"]
                
                if not base_color:
                    return ChatResponse(
                        response="I'd love to help you with colors! Could you tell me which color you're working with? For example, 'what colors go with red?'",
                        session_id=session_id
                    )
                
                matching_colors = get_color_matches(base_color)
                clothing_item = clothing_info.get("existing_item") or clothing_info.get("type") or conv_context["context"].get("last_item", "outfit")
                
                suggested_colors = [c.title() for c in matching_colors[:4]]
                
                message_lower_for_style = message.lower()
                style_keywords = []
                if "stylish" in message_lower_for_style or "standout" in message_lower_for_style or "stand out" in message_lower_for_style:
                    style_keywords.append("stylish and eye-catching")
                if "professional" in message_lower_for_style or "formal" in message_lower_for_style:
                    style_keywords.append("professional")
                if "casual" in message_lower_for_style:
                    style_keywords.append("casual")
                
                style_note = ""
                if style_keywords:
                    style_note = f" The user wants something {', '.join(style_keywords)}."
                
                prompt = f"What colors go well with {base_color} {clothing_item}?{style_note} Give me a natural, conversational answer with specific color suggestions."
                ai_response = generate_chat_response(prompt, max_length=300, temperature=0.8, rag_context=rag_context, images=None)
                if len(ai_response) > 30:
                    response_text = ai_response
                else:
                    response_text = f"For your {base_color} {clothing_item}, I'd suggest pairing it with {', '.join(suggested_colors[:3])}, or {suggested_colors[3] if len(suggested_colors) > 3 else 'other neutrals'}. These colors complement each other beautifully!"
                
                update_context(session_id, message, {
                    "response": response_text,
                    "color": base_color,
                    "item": clothing_item,
                    "colors": suggested_colors
                })
                
                return ChatResponse(
                    response=response_text,
                    session_id=session_id
                )
            
            else:
                clothing_info = extract_clothing_info(enhanced_message)
                
                if not clothing_info.get("color") and conv_context["context"].get("last_color"):
                    enhanced_message = f"{enhanced_message} {conv_context['context']['last_color']}"
                    clothing_info = extract_clothing_info(enhanced_message)
                
                context_info = ""
                if clothing_info.get("color"):
                    context_info += f"Color preference: {clothing_info.get('color')}. "
                if clothing_info.get("type"):
                    context_info += f"Item type: {clothing_info.get('type')}. "
                if clothing_info.get("existing_item"):
                    context_info += f"User has: {clothing_info.get('existing_item')}. "
                
                occasion_keywords = ["defense", "project", "presentation", "meeting", "interview", "formal", "casual", "party", "wedding"]
                occasion = next((word for word in occasion_keywords if word in enhanced_message.lower()), None)
                if occasion:
                    context_info += f"Occasion: {occasion}. "
                
                prompt = f"{enhanced_message}"
                if context_info:
                    prompt += f"\n\nContext: {context_info.strip()}"
                prompt += "\n\nGive helpful, detailed outfit suggestions that are practical and stylish. Be specific about item combinations and explain why they work well."
                
                response_text = generate_chat_response(prompt, max_length=1024, temperature=0.8, rag_context=rag_context, images=None)
                
                update_context(session_id, message, {
                    "response": response_text,
                    "color": clothing_info.get("color"),
                    "item": clothing_info.get("type") or clothing_info.get("requested_item"),
                    "items": clothing_info.get("items", [])
                })
                
                return ChatResponse(
                    response=response_text,
                    session_id=session_id
                )
        
        except Exception as e:
            raise HTTPException(status_code=500, detail=f"Error processing text message: {str(e)}")

    @app.post("/chat", response_model=ChatResponse)
    async def chat(request: ChatRequest):
        try:
            message = request.message.strip()
            session_id = request.session_id
            
            if not message:
                raise HTTPException(status_code=400, detail="Message cannot be empty")
            
            if request.wardrobe and len(request.wardrobe) > 0:
                print(f"[WARDROBE CHAT] ===== WARDROBE REQUEST DETECTED =====")
                if request.wardrobe_description:
                    print(f"[WARDROBE CHAT] Using provided wardrobe description ({len(request.wardrobe_description)} chars)")
                return await handle_wardrobe_chat(message, request.wardrobe, session_id, images=request.images, wardrobe_description=request.wardrobe_description)
            
            conv_context = get_conversation_context(session_id)
            
            if is_name_question(message):
                prompt = "What is your name? Respond naturally and friendly."
                rag_chunks = retrieve_relevant_context(message, top_k=2)
                rag_context = format_rag_context(rag_chunks)
                response_text = generate_chat_response(prompt, max_length=100, temperature=0.8, rag_context=rag_context, images=request.images)
                update_context(session_id, message, {"response": response_text})
                return ChatResponse(response=response_text, session_id=session_id)
            
            if is_greeting(message):
                prompt = f"{message} Respond warmly and offer to help with fashion advice."
                rag_chunks = retrieve_relevant_context(message, top_k=2)
                rag_context = format_rag_context(rag_chunks)
                response_text = generate_chat_response(prompt, max_length=150, temperature=0.8, rag_context=rag_context, images=request.images)
                update_context(session_id, message, {"response": response_text})
                return ChatResponse(response=response_text, session_id=session_id)
            
            enhanced_message = enhance_message_with_context(message, conv_context["context"])
            query_type = detect_query_type(enhanced_message)
            rag_chunks = retrieve_relevant_context(enhanced_message, top_k=3)
            rag_context = format_rag_context(rag_chunks)
            
            if query_type == "color_compatibility":
                found_colors = extract_colors_from_query(enhanced_message)
                
                if len(found_colors) >= 2:
                    color1_mapped = found_colors[0][1]
                    color2_mapped = found_colors[1][1]
                    color1_original = found_colors[0][0]
                    color2_original = found_colors[1][0]
                    
                    compatible = False
                    if color1_mapped in COLOR_HARMONY:
                        compatible = color2_mapped in COLOR_HARMONY[color1_mapped]
                    elif color2_mapped in COLOR_HARMONY:
                        compatible = color1_mapped in COLOR_HARMONY[color2_mapped]
                    
                    neutrals = ["white", "black", "grey", "gray", "beige", "navy"]
                    if color1_mapped in neutrals or color2_mapped in neutrals:
                        compatible = True
                    
                    if compatible:
                        response_text = f"Yes, {color1_original.title()} will go well with {color2_original.title()}. They create a balanced and stylish combination that works great together!"
                    else:
                        response_text = f"{color1_original.title()} and {color2_original.title()} can work together, though you might want to add some neutral pieces to balance the look."
                    
                    prompt = f"Does {color1_original} go well with {color2_original}? Answer naturally and conversationally."
                    ai_response = generate_chat_response(prompt, max_length=150, temperature=0.8, rag_context=rag_context, images=request.images)
                    if len(ai_response) > 15:
                        response_text = ai_response
                    
                    update_context(session_id, message, {
                        "response": response_text,
                        "color": color1_original,
                        "colors": [color1_original, color2_original]
                    })
                    
                    return ChatResponse(
                        response=response_text,
                        session_id=session_id
                    )
            
            elif query_type == "color_suggestion":
                clothing_info = extract_clothing_info(enhanced_message)
                base_color = clothing_info.get("color")
                
                if not base_color:
                    found_colors = extract_colors_from_query(enhanced_message)
                    if found_colors:
                        base_color = found_colors[0][1]
                    elif conv_context["context"].get("last_color"):
                        base_color = conv_context["context"]["last_color"]
                
                if not base_color:
                    return ChatResponse(
                        response="I'd love to help you with colors! Could you tell me which color you're working with? For example, 'what colors go with red?'",
                        session_id=session_id
                    )
                
                matching_colors = get_color_matches(base_color)
                clothing_item = clothing_info.get("existing_item") or clothing_info.get("type") or conv_context["context"].get("last_item", "outfit")
                
                suggested_colors = [c.title() for c in matching_colors[:4]]
                
                message_lower_for_style = message.lower()
                style_keywords = []
                if "stylish" in message_lower_for_style or "standout" in message_lower_for_style or "stand out" in message_lower_for_style:
                    style_keywords.append("stylish and eye-catching")
                if "professional" in message_lower_for_style or "formal" in message_lower_for_style:
                    style_keywords.append("professional")
                if "casual" in message_lower_for_style:
                    style_keywords.append("casual")
                
                style_note = ""
                if style_keywords:
                    style_note = f" The user wants something {', '.join(style_keywords)}."
                
                prompt = f"What colors go well with {base_color} {clothing_item}?{style_note} Give me a natural, conversational answer with specific color suggestions."
                ai_response = generate_chat_response(prompt, max_length=300, temperature=0.8, rag_context=rag_context, images=request.images)
                if len(ai_response) > 30:
                    response_text = ai_response
                else:
                    response_text = f"For your {base_color} {clothing_item}, I'd suggest pairing it with {', '.join(suggested_colors[:3])}, or {suggested_colors[3] if len(suggested_colors) > 3 else 'other neutrals'}. These colors complement each other beautifully!"
                
                update_context(session_id, message, {
                    "response": response_text,
                    "color": base_color,
                    "item": clothing_item,
                    "colors": suggested_colors
                })
                
                return ChatResponse(
                    response=response_text,
                    session_id=session_id
                )
            
            else:
                clothing_info = extract_clothing_info(enhanced_message)
                
                if not clothing_info.get("color") and conv_context["context"].get("last_color"):
                    enhanced_message = f"{enhanced_message} {conv_context['context']['last_color']}"
                    clothing_info = extract_clothing_info(enhanced_message)
                
                context_info = ""
                if clothing_info.get("color"):
                    context_info += f"Color preference: {clothing_info.get('color')}. "
                if clothing_info.get("type"):
                    context_info += f"Item type: {clothing_info.get('type')}. "
                if clothing_info.get("existing_item"):
                    context_info += f"User has: {clothing_info.get('existing_item')}. "
                
                occasion_keywords = ["defense", "project", "presentation", "meeting", "interview", "formal", "casual", "party", "wedding"]
                occasion = next((word for word in occasion_keywords if word in enhanced_message.lower()), None)
                if occasion:
                    context_info += f"Occasion: {occasion}. "
                
                prompt = f"{enhanced_message}"
                if context_info:
                    prompt += f"\n\nContext: {context_info.strip()}"
                prompt += "\n\nGive helpful, detailed outfit suggestions that are practical and stylish. Be specific about item combinations and explain why they work well."
                
                response_text = generate_chat_response(prompt, max_length=1024, temperature=0.8, rag_context=rag_context, images=request.images)
                
                update_context(session_id, message, {
                    "response": response_text,
                    "color": clothing_info.get("color"),
                    "item": clothing_info.get("type") or clothing_info.get("requested_item"),
                    "items": clothing_info.get("items", [])
                })
                
                return ChatResponse(
                    response=response_text,
                    session_id=session_id
                )
        
        except Exception as e:
            raise HTTPException(status_code=500, detail=f"Error processing chat message: {str(e)}")

    @app.post("/chat/upload", response_model=ChatResponse)
    async def chat_with_upload(
        message: str = Form(...),
        session_id: str = Form(default="default"),
        wardrobe: Optional[str] = Form(default=None),
        wardrobe_description: Optional[str] = Form(default=None),
        images: List[UploadFile] = File(default=[])
    ):
        try:
            wardrobe_items = []
            if wardrobe and wardrobe.strip() and wardrobe.strip() not in ["[]", "", "string"]:
                try:
                    wardrobe_data = json.loads(wardrobe)
                    if isinstance(wardrobe_data, list):
                        wardrobe_items = [WardrobeItem(**item) for item in wardrobe_data]
                except json.JSONDecodeError:
                    print(f"[UPLOAD] Ignoring invalid wardrobe value: {wardrobe[:50]}")
            
            image_data_urls = []
            for img_file in images:
                if img_file.filename:
                    content = await img_file.read()
                    content_type = img_file.content_type or "image/jpeg"
                    base64_data = base64.b64encode(content).decode("utf-8")
                    data_url = f"data:{content_type};base64,{base64_data}"
                    image_data_urls.append(data_url)
                    print(f"[UPLOAD] Processed image: {img_file.filename} ({len(content)} bytes)")
            
            request = ChatRequest(
                message=message,
                session_id=session_id,
                wardrobe=wardrobe_items if wardrobe_items else None,
                wardrobe_description=wardrobe_description if wardrobe_description and wardrobe_description.strip() else None,
                images=image_data_urls if image_data_urls else None
            )
            
            print(f"[UPLOAD] Processing chat request: message='{message[:50]}...', images={len(image_data_urls)}, wardrobe={len(wardrobe_items)}")
            result = await chat(request)
            print(f"[UPLOAD] Response generated: {len(result.response)} chars")
            return result
        
        except Exception as e:
            print(f"[UPLOAD] Error: {e}")
            raise HTTPException(status_code=500, detail=f"Error processing upload: {str(e)}")

    @app.post("/chat/upload/stream")
    async def chat_with_upload_stream(
        message: str = Form(...),
        session_id: str = Form(default="default"),
        wardrobe: Optional[str] = Form(default=None),
        wardrobe_description: Optional[str] = Form(default=None),
        images: List[UploadFile] = File(default=[])
    ):
        image_data_urls = []
        for img_file in images:
            if img_file.filename:
                content = await img_file.read()
                content_type = img_file.content_type or "image/jpeg"
                base64_data = base64.b64encode(content).decode("utf-8")
                data_url = f"data:{content_type};base64,{base64_data}"
                image_data_urls.append(data_url)
                print(f"[STREAM UPLOAD] Processed image: {img_file.filename} ({len(content)} bytes)")
        
        wardrobe_items = []
        if wardrobe and wardrobe.strip() and wardrobe.strip() not in ["[]", "", "string"]:
            try:
                wardrobe_data = json.loads(wardrobe)
                if isinstance(wardrobe_data, list):
                    wardrobe_items = [WardrobeItem(**item) for item in wardrobe_data]
            except json.JSONDecodeError:
                print(f"[STREAM UPLOAD] Ignoring invalid wardrobe value: {wardrobe[:50]}")
        
        rag_chunks = retrieve_relevant_context(message, top_k=3)
        rag_context = format_rag_context(rag_chunks)
        
        wardrobe_context = ""
        if wardrobe_description and wardrobe_description.strip():
            wardrobe_context = wardrobe_description
            print(f"[STREAM UPLOAD] Using provided wardrobe description ({len(wardrobe_context)} chars)")
        elif wardrobe_items:
            from wardrobe import format_wardrobe_for_prompt
            wardrobe_context = format_wardrobe_for_prompt(wardrobe_items)
            print(f"[STREAM UPLOAD] Generated wardrobe context ({len(wardrobe_context)} chars)")
        
        if wardrobe_context:
            prompt = f"""{wardrobe_context}

User request: {message}

Suggest a complete outfit using ONLY the items listed above. Reference items by their exact names. Include accessories if available. Be friendly and conversational."""
        else:
            prompt = message
        
        print(f"[STREAM UPLOAD] Starting streaming response for: {message[:50]}...")
        
        async def generate():
            yield f"data: {json.dumps({'type': 'start', 'session_id': session_id})}\n\n"
            
            full_response = ""
            async for chunk in generate_chat_response_streaming(
                prompt=prompt,
                max_length=512,
                temperature=0.7,
                rag_context=rag_context,
                images=image_data_urls if image_data_urls else None
            ):
                full_response += chunk
                yield f"data: {json.dumps({'type': 'chunk', 'content': chunk})}\n\n"
            
            yield f"data: {json.dumps({'type': 'end', 'full_response': full_response, 'session_id': session_id})}\n\n"
            print(f"[STREAM UPLOAD] Streaming complete: {len(full_response)} chars")
        
        return StreamingResponse(
            generate(),
            media_type="text/event-stream",
            headers={
                "Cache-Control": "no-cache",
                "Connection": "keep-alive",
                "X-Accel-Buffering": "no",
            }
        )