Update app.py
Browse files
app.py
CHANGED
@@ -10,35 +10,18 @@ from datetime import datetime, timezone, timedelta
|
|
10 |
from dotenv import load_dotenv
|
11 |
import json
|
12 |
|
13 |
-
# Load environment variables
|
14 |
-
# This is useful for local development. In production on platforms like Hugging Face,
|
15 |
-
# you'll set these as environment variables directly in the settings.
|
16 |
load_dotenv()
|
17 |
|
18 |
app = Flask(__name__)
|
19 |
CORS(app)
|
20 |
|
21 |
# Initialize Gemini client
|
22 |
-
|
23 |
-
api_key = os.getenv('GOOGLE_API_KEY')
|
24 |
-
if not api_key:
|
25 |
-
print("Error: GOOGLE_API_KEY environment variable not set.")
|
26 |
-
# In a real app, you might exit or raise an exception here.
|
27 |
-
# For this example, we'll print an error but allow the app to start;
|
28 |
-
# API calls will fail if the key is missing.
|
29 |
-
# If running locally, make sure you have a .env file with GOOGLE_API_KEY=YOUR_API_KEY
|
30 |
-
pass # Allows the app to run without a key for debugging non-API parts
|
31 |
|
32 |
-
|
33 |
-
client = genai.Client(api_key=api_key)
|
34 |
-
except Exception as e:
|
35 |
-
print(f"Failed to initialize Gemini client: {e}")
|
36 |
-
client = None # Set client to None if initialization fails
|
37 |
-
|
38 |
-
# In-memory storage for demo (in production, use a database like Redis or PostgreSQL)
|
39 |
-
# Maps our internal cache_id (UUID) to Gemini's cache_name and other info
|
40 |
document_caches = {}
|
41 |
-
user_sessions = {}
|
42 |
|
43 |
# HTML template for the web interface
|
44 |
HTML_TEMPLATE = """
|
@@ -48,58 +31,62 @@ HTML_TEMPLATE = """
|
|
48 |
<meta charset="UTF-8">
|
49 |
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
50 |
<title>Smart Document Analysis Platform</title>
|
51 |
-
<link href="https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&display=swap" rel="stylesheet">
|
52 |
<style>
|
53 |
* {
|
54 |
margin: 0;
|
55 |
padding: 0;
|
56 |
box-sizing: border-box;
|
57 |
}
|
58 |
-
|
59 |
body {
|
60 |
font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
|
61 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
62 |
min-height: 100vh;
|
63 |
color: #333;
|
64 |
-
line-height: 1.6;
|
65 |
}
|
66 |
-
|
67 |
.container {
|
68 |
max-width: 1400px;
|
69 |
margin: 0 auto;
|
70 |
padding: 20px;
|
71 |
min-height: 100vh;
|
72 |
-
display: flex;
|
73 |
-
flex-direction: column;
|
74 |
}
|
75 |
-
|
76 |
.header {
|
77 |
text-align: center;
|
78 |
margin-bottom: 30px;
|
79 |
color: white;
|
80 |
}
|
81 |
-
|
82 |
.header h1 {
|
83 |
font-size: 2.8em;
|
84 |
font-weight: 700;
|
85 |
margin-bottom: 10px;
|
86 |
text-shadow: 0 2px 4px rgba(0,0,0,0.3);
|
87 |
}
|
88 |
-
|
89 |
.header p {
|
90 |
font-size: 1.2em;
|
91 |
opacity: 0.9;
|
92 |
font-weight: 300;
|
93 |
}
|
94 |
-
|
95 |
.main-content {
|
96 |
display: grid;
|
97 |
grid-template-columns: 1fr 1fr;
|
98 |
gap: 30px;
|
99 |
-
|
100 |
}
|
101 |
-
|
102 |
-
.left-panel
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
background: white;
|
104 |
border-radius: 20px;
|
105 |
padding: 30px;
|
@@ -107,11 +94,7 @@ HTML_TEMPLATE = """
|
|
107 |
display: flex;
|
108 |
flex-direction: column;
|
109 |
}
|
110 |
-
|
111 |
-
.left-panel {
|
112 |
-
overflow-y: auto; /* Allow scrolling if content is tall */
|
113 |
-
}
|
114 |
-
|
115 |
.panel-title {
|
116 |
font-size: 1.5em;
|
117 |
font-weight: 600;
|
@@ -121,11 +104,11 @@ HTML_TEMPLATE = """
|
|
121 |
align-items: center;
|
122 |
gap: 10px;
|
123 |
}
|
124 |
-
|
125 |
.upload-section {
|
126 |
margin-bottom: 30px;
|
127 |
}
|
128 |
-
|
129 |
.upload-area {
|
130 |
border: 2px dashed #667eea;
|
131 |
border-radius: 15px;
|
@@ -134,31 +117,30 @@ HTML_TEMPLATE = """
|
|
134 |
background: #f8fafc;
|
135 |
transition: all 0.3s ease;
|
136 |
margin-bottom: 20px;
|
137 |
-
cursor: pointer; /* Indicate clickable area */
|
138 |
}
|
139 |
-
|
140 |
.upload-area:hover {
|
141 |
border-color: #764ba2;
|
142 |
background: #f0f2ff;
|
143 |
transform: translateY(-2px);
|
144 |
}
|
145 |
-
|
146 |
.upload-area.dragover {
|
147 |
border-color: #764ba2;
|
148 |
background: #e8f0ff;
|
149 |
transform: scale(1.02);
|
150 |
}
|
151 |
-
|
152 |
.upload-icon {
|
153 |
font-size: 3em;
|
154 |
color: #667eea;
|
155 |
margin-bottom: 15px;
|
156 |
}
|
157 |
-
|
158 |
.file-input {
|
159 |
display: none;
|
160 |
}
|
161 |
-
|
162 |
.upload-btn {
|
163 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
164 |
color: white;
|
@@ -171,12 +153,12 @@ HTML_TEMPLATE = """
|
|
171 |
transition: all 0.3s ease;
|
172 |
margin: 10px;
|
173 |
}
|
174 |
-
|
175 |
.upload-btn:hover {
|
176 |
transform: translateY(-2px);
|
177 |
box-shadow: 0 10px 20px rgba(102, 126, 234, 0.3);
|
178 |
}
|
179 |
-
|
180 |
.url-input {
|
181 |
width: 100%;
|
182 |
padding: 15px;
|
@@ -186,12 +168,12 @@ HTML_TEMPLATE = """
|
|
186 |
margin-bottom: 15px;
|
187 |
transition: border-color 0.3s ease;
|
188 |
}
|
189 |
-
|
190 |
.url-input:focus {
|
191 |
outline: none;
|
192 |
border-color: #667eea;
|
193 |
}
|
194 |
-
|
195 |
.btn {
|
196 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
197 |
color: white;
|
@@ -203,18 +185,18 @@ HTML_TEMPLATE = """
|
|
203 |
font-weight: 500;
|
204 |
transition: all 0.3s ease;
|
205 |
}
|
206 |
-
|
207 |
.btn:hover {
|
208 |
transform: translateY(-1px);
|
209 |
box-shadow: 0 5px 15px rgba(102, 126, 234, 0.3);
|
210 |
}
|
211 |
-
|
212 |
.btn:disabled {
|
213 |
opacity: 0.6;
|
214 |
cursor: not-allowed;
|
215 |
transform: none;
|
216 |
}
|
217 |
-
|
218 |
.chat-container {
|
219 |
flex: 1;
|
220 |
border: 1px solid #e2e8f0;
|
@@ -223,44 +205,40 @@ HTML_TEMPLATE = """
|
|
223 |
padding: 20px;
|
224 |
background: #f8fafc;
|
225 |
margin-bottom: 20px;
|
226 |
-
display: flex;
|
227 |
-
flex-direction: column;
|
228 |
}
|
229 |
-
|
230 |
.message {
|
231 |
margin-bottom: 15px;
|
232 |
padding: 15px;
|
233 |
border-radius: 12px;
|
234 |
max-width: 85%;
|
235 |
animation: fadeIn 0.3s ease;
|
236 |
-
word-wrap: break-word; /* Ensure long words wrap */
|
237 |
}
|
238 |
-
|
239 |
@keyframes fadeIn {
|
240 |
from { opacity: 0; transform: translateY(10px); }
|
241 |
to { opacity: 1; transform: translateY(0); }
|
242 |
}
|
243 |
-
|
244 |
.user-message {
|
245 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
246 |
color: white;
|
247 |
margin-left: auto;
|
248 |
box-shadow: 0 4px 12px rgba(102, 126, 234, 0.3);
|
249 |
}
|
250 |
-
|
251 |
.ai-message {
|
252 |
background: white;
|
253 |
color: #333;
|
254 |
border: 1px solid #e2e8f0;
|
255 |
box-shadow: 0 2px 8px rgba(0,0,0,0.1);
|
256 |
-
margin-right: auto; /* Align AI messages to the left */
|
257 |
}
|
258 |
-
|
259 |
.input-group {
|
260 |
display: flex;
|
261 |
gap: 10px;
|
262 |
}
|
263 |
-
|
264 |
.question-input {
|
265 |
flex: 1;
|
266 |
padding: 15px;
|
@@ -269,12 +247,12 @@ HTML_TEMPLATE = """
|
|
269 |
font-size: 1em;
|
270 |
transition: border-color 0.3s ease;
|
271 |
}
|
272 |
-
|
273 |
.question-input:focus {
|
274 |
outline: none;
|
275 |
border-color: #667eea;
|
276 |
}
|
277 |
-
|
278 |
.cache-info {
|
279 |
background: linear-gradient(135deg, #48bb78 0%, #38a169 100%);
|
280 |
border-radius: 12px;
|
@@ -283,28 +261,18 @@ HTML_TEMPLATE = """
|
|
283 |
color: white;
|
284 |
box-shadow: 0 4px 12px rgba(72, 187, 120, 0.3);
|
285 |
}
|
286 |
-
|
287 |
.cache-info h3 {
|
288 |
margin-bottom: 10px;
|
289 |
font-weight: 600;
|
290 |
}
|
291 |
-
|
292 |
-
.cache-info p {
|
293 |
-
font-size: 0.9em;
|
294 |
-
margin-bottom: 5px;
|
295 |
-
}
|
296 |
-
|
297 |
-
.cache-info p:last-child {
|
298 |
-
margin-bottom: 0;
|
299 |
-
}
|
300 |
-
|
301 |
-
|
302 |
.loading {
|
303 |
text-align: center;
|
304 |
padding: 40px;
|
305 |
color: #666;
|
306 |
}
|
307 |
-
|
308 |
.loading-spinner {
|
309 |
border: 3px solid #f3f3f3;
|
310 |
border-top: 3px solid #667eea;
|
@@ -314,12 +282,12 @@ HTML_TEMPLATE = """
|
|
314 |
animation: spin 1s linear infinite;
|
315 |
margin: 0 auto 20px;
|
316 |
}
|
317 |
-
|
318 |
@keyframes spin {
|
319 |
0% { transform: rotate(0deg); }
|
320 |
100% { transform: rotate(360deg); }
|
321 |
}
|
322 |
-
|
323 |
.error {
|
324 |
background: linear-gradient(135deg, #f56565 0%, #e53e3e 100%);
|
325 |
border-radius: 12px;
|
@@ -328,7 +296,7 @@ HTML_TEMPLATE = """
|
|
328 |
margin-bottom: 20px;
|
329 |
box-shadow: 0 4px 12px rgba(245, 101, 101, 0.3);
|
330 |
}
|
331 |
-
|
332 |
.success {
|
333 |
background: linear-gradient(135deg, #48bb78 0%, #38a169 100%);
|
334 |
border-radius: 12px;
|
@@ -337,13 +305,13 @@ HTML_TEMPLATE = """
|
|
337 |
margin-bottom: 20px;
|
338 |
box-shadow: 0 4px 12px rgba(72, 187, 120, 0.3);
|
339 |
}
|
340 |
-
|
341 |
@media (max-width: 768px) {
|
342 |
.main-content {
|
343 |
grid-template-columns: 1fr;
|
344 |
gap: 20px;
|
345 |
}
|
346 |
-
|
347 |
.header h1 {
|
348 |
font-size: 2em;
|
349 |
}
|
@@ -355,190 +323,146 @@ HTML_TEMPLATE = """
|
|
355 |
<div class="header">
|
356 |
<h1>📚 Smart Document Analysis Platform</h1>
|
357 |
<p>Upload PDF documents once, ask questions forever with Gemini API caching</p>
|
358 |
-
<p style="font-size:0.9em; margin-top: 5px; opacity: 0.8;">Powered by Google Gemini API - Explicit Caching</p>
|
359 |
</div>
|
360 |
-
|
361 |
<div class="main-content">
|
362 |
<!-- Left Panel - Upload Section -->
|
363 |
<div class="left-panel">
|
364 |
<div class="panel-title">
|
365 |
📤 Upload PDF Document
|
366 |
</div>
|
367 |
-
|
368 |
<div class="upload-section">
|
369 |
<div class="upload-area" id="uploadArea">
|
370 |
<div class="upload-icon">📄</div>
|
371 |
<p>Drag and drop your PDF file here, or click to select</p>
|
372 |
<input type="file" id="fileInput" class="file-input" accept=".pdf">
|
373 |
-
|
374 |
-
<button type="button" class="upload-btn" onclick="document.getElementById('fileInput').click()">
|
375 |
Choose PDF File
|
376 |
</button>
|
377 |
</div>
|
378 |
-
|
379 |
<div style="margin-top: 20px;">
|
380 |
<h3>Or provide a URL:</h3>
|
381 |
<input type="url" id="urlInput" class="url-input" placeholder="https://example.com/document.pdf">
|
382 |
-
<button
|
383 |
</div>
|
384 |
</div>
|
385 |
-
|
386 |
<div id="loading" class="loading" style="display: none;">
|
387 |
<div class="loading-spinner"></div>
|
388 |
<p id="loadingText">Processing your PDF... This may take a moment.</p>
|
389 |
</div>
|
390 |
-
|
391 |
<div id="error" class="error" style="display: none;"></div>
|
392 |
<div id="success" class="success" style="display: none;"></div>
|
393 |
</div>
|
394 |
-
|
395 |
<!-- Right Panel - Chat Section -->
|
396 |
<div class="right-panel">
|
397 |
<div class="panel-title">
|
398 |
💬 Ask Questions
|
399 |
</div>
|
400 |
-
|
401 |
<div id="cacheInfo" class="cache-info" style="display: none;">
|
402 |
<h3>✅ Document Cached Successfully!</h3>
|
403 |
<p>Your PDF has been cached using Gemini API. You can now ask multiple questions without re-uploading.</p>
|
404 |
<p><strong>Cache ID:</strong> <span id="cacheId"></span></p>
|
405 |
<p><strong>Tokens Cached:</strong> <span id="tokenCount"></span></p>
|
406 |
-
<p>Note: Caching is ideal for larger documents (typically 1024+ tokens required).</p>
|
407 |
</div>
|
408 |
-
|
409 |
<div class="chat-container" id="chatContainer">
|
410 |
<div class="message ai-message">
|
411 |
-
👋 Hello!
|
412 |
</div>
|
413 |
</div>
|
414 |
-
|
415 |
<div class="input-group">
|
416 |
-
<input type="text" id="questionInput" class="question-input" placeholder="Ask a question about your document..."
|
417 |
-
<button
|
418 |
</div>
|
419 |
</div>
|
420 |
</div>
|
421 |
</div>
|
|
|
422 |
<script>
|
423 |
let currentCacheId = null;
|
424 |
-
|
425 |
-
// Disable input/button initially
|
426 |
-
document.getElementById('questionInput').disabled = true;
|
427 |
-
document.getElementById('askBtn').disabled = true;
|
428 |
-
|
429 |
// File upload handling
|
430 |
const uploadArea = document.getElementById('uploadArea');
|
431 |
const fileInput = document.getElementById('fileInput');
|
432 |
-
|
433 |
-
|
434 |
-
['dragenter', 'dragover', 'dragleave', 'drop'].forEach(eventName => {
|
435 |
-
uploadArea.addEventListener(eventName, preventDefaults, false);
|
436 |
-
});
|
437 |
-
|
438 |
-
function preventDefaults (e) {
|
439 |
e.preventDefault();
|
440 |
-
|
441 |
-
}
|
442 |
-
|
443 |
-
// Highlight drop area when item is dragged over
|
444 |
-
['dragenter', 'dragover'].forEach(eventName => {
|
445 |
-
uploadArea.addEventListener(eventName, () => uploadArea.classList.add('dragover'), false);
|
446 |
});
|
447 |
-
|
448 |
-
|
449 |
-
uploadArea.
|
450 |
});
|
451 |
-
|
452 |
-
|
453 |
-
|
454 |
-
|
455 |
-
|
456 |
-
const dt = e.dataTransfer;
|
457 |
-
const files = dt.files;
|
458 |
if (files.length > 0) {
|
459 |
uploadFile(files[0]);
|
460 |
}
|
461 |
-
}
|
462 |
-
|
463 |
fileInput.addEventListener('change', (e) => {
|
464 |
if (e.target.files.length > 0) {
|
465 |
uploadFile(e.target.files[0]);
|
466 |
-
// Clear the input so the same file can be selected again if needed
|
467 |
-
e.target.value = '';
|
468 |
}
|
469 |
});
|
470 |
-
|
471 |
async function uploadFile(file) {
|
472 |
if (!file.type.includes('pdf')) {
|
473 |
showError('Please select a PDF file.');
|
474 |
return;
|
475 |
}
|
476 |
-
|
477 |
-
// Clear previous status messages
|
478 |
-
hideError();
|
479 |
-
hideSuccess();
|
480 |
-
document.getElementById('cacheInfo').style.display = 'none'; // Hide old cache info
|
481 |
-
currentCacheId = null; // Clear old cache ID
|
482 |
-
|
483 |
showLoading('Uploading PDF...');
|
484 |
-
|
485 |
const formData = new FormData();
|
486 |
formData.append('file', file);
|
487 |
-
|
488 |
try {
|
489 |
const response = await fetch('/upload', {
|
490 |
method: 'POST',
|
491 |
body: formData
|
492 |
});
|
493 |
-
|
494 |
const result = await response.json();
|
495 |
-
|
496 |
if (result.success) {
|
497 |
currentCacheId = result.cache_id;
|
498 |
document.getElementById('cacheId').textContent = result.cache_id;
|
499 |
document.getElementById('tokenCount').textContent = result.token_count;
|
500 |
document.getElementById('cacheInfo').style.display = 'block';
|
501 |
-
showSuccess('PDF uploaded and cached successfully!
|
502 |
-
|
503 |
-
// Enable chat input and button
|
504 |
-
document.getElementById('questionInput').disabled = false;
|
505 |
-
document.getElementById('askBtn').disabled = false;
|
506 |
-
document.getElementById('questionInput').focus(); // Focus input
|
507 |
-
|
508 |
// Add initial message
|
509 |
addMessage("I've analyzed your PDF document. What would you like to know about it?", 'ai');
|
510 |
-
|
511 |
} else {
|
512 |
showError(result.error);
|
513 |
-
// Disable chat input/button if upload/cache failed
|
514 |
-
document.getElementById('questionInput').disabled = true;
|
515 |
-
document.getElementById('askBtn').disabled = true;
|
516 |
}
|
517 |
} catch (error) {
|
518 |
showError('Error uploading file: ' + error.message);
|
519 |
-
// Disable chat input/button on network/server error
|
520 |
-
document.getElementById('questionInput').disabled = true;
|
521 |
-
document.getElementById('askBtn').disabled = true;
|
522 |
} finally {
|
523 |
hideLoading();
|
524 |
}
|
525 |
}
|
526 |
-
|
527 |
async function uploadFromUrl() {
|
528 |
const url = document.getElementById('urlInput').value;
|
529 |
-
if (!url
|
530 |
showError('Please enter a valid URL.');
|
531 |
return;
|
532 |
}
|
533 |
-
|
534 |
-
// Clear previous status messages
|
535 |
-
hideError();
|
536 |
-
hideSuccess();
|
537 |
-
document.getElementById('cacheInfo').style.display = 'none'; // Hide old cache info
|
538 |
-
currentCacheId = null; // Clear old cache ID
|
539 |
-
|
540 |
showLoading('Uploading PDF from URL...');
|
541 |
-
|
542 |
try {
|
543 |
const response = await fetch('/upload-url', {
|
544 |
method: 'POST',
|
@@ -547,61 +471,47 @@ HTML_TEMPLATE = """
|
|
547 |
},
|
548 |
body: JSON.stringify({ url: url })
|
549 |
});
|
550 |
-
|
551 |
const result = await response.json();
|
552 |
-
|
553 |
if (result.success) {
|
554 |
currentCacheId = result.cache_id;
|
555 |
document.getElementById('cacheId').textContent = result.cache_id;
|
556 |
document.getElementById('tokenCount').textContent = result.token_count;
|
557 |
document.getElementById('cacheInfo').style.display = 'block';
|
558 |
-
showSuccess('PDF uploaded and cached successfully!
|
559 |
-
|
560 |
-
// Enable chat input and button
|
561 |
-
document.getElementById('questionInput').disabled = false;
|
562 |
-
document.getElementById('askBtn').disabled = false;
|
563 |
-
document.getElementById('questionInput').focus(); // Focus input
|
564 |
-
|
565 |
// Add initial message
|
566 |
addMessage("I've analyzed your PDF document. What would you like to know about it?", 'ai');
|
567 |
-
|
568 |
} else {
|
569 |
showError(result.error);
|
570 |
-
// Disable chat input/button if upload/cache failed
|
571 |
-
document.getElementById('questionInput').disabled = true;
|
572 |
-
document.getElementById('askBtn').disabled = true;
|
573 |
}
|
574 |
} catch (error) {
|
575 |
showError('Error uploading from URL: ' + error.message);
|
576 |
-
// Disable chat input/button on network/server error
|
577 |
-
document.getElementById('questionInput').disabled = true;
|
578 |
-
document.getElementById('askBtn').disabled = false; // Should be false? Fix: should be true
|
579 |
} finally {
|
580 |
hideLoading();
|
581 |
}
|
582 |
}
|
583 |
-
|
584 |
async function askQuestion() {
|
585 |
-
const
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
if (!currentCacheId) {
|
590 |
showError('Please upload a PDF document first.');
|
591 |
return;
|
592 |
}
|
593 |
-
|
594 |
// Add user message to chat
|
595 |
addMessage(question, 'user');
|
596 |
-
questionInput.value = '';
|
597 |
-
|
598 |
// Show loading state
|
599 |
const askBtn = document.getElementById('askBtn');
|
600 |
const originalText = askBtn.textContent;
|
601 |
askBtn.textContent = 'Generating...';
|
602 |
askBtn.disabled = true;
|
603 |
-
|
604 |
-
|
605 |
try {
|
606 |
const response = await fetch('/ask', {
|
607 |
method: 'POST',
|
@@ -610,12 +520,12 @@ HTML_TEMPLATE = """
|
|
610 |
},
|
611 |
body: JSON.stringify({
|
612 |
question: question,
|
613 |
-
cache_id: currentCacheId
|
614 |
})
|
615 |
});
|
616 |
-
|
617 |
const result = await response.json();
|
618 |
-
|
619 |
if (result.success) {
|
620 |
addMessage(result.answer, 'ai');
|
621 |
} else {
|
@@ -626,565 +536,260 @@ HTML_TEMPLATE = """
|
|
626 |
} finally {
|
627 |
askBtn.textContent = originalText;
|
628 |
askBtn.disabled = false;
|
629 |
-
questionInput.disabled = false; // Re-enable input
|
630 |
-
questionInput.focus(); // Put focus back on input
|
631 |
-
// Ensure button is disabled only if no cache is active
|
632 |
-
if (!currentCacheId) {
|
633 |
-
askBtn.disabled = true;
|
634 |
-
questionInput.disabled = true;
|
635 |
-
}
|
636 |
}
|
637 |
}
|
638 |
-
|
639 |
function addMessage(text, sender) {
|
640 |
const chatContainer = document.getElementById('chatContainer');
|
641 |
const messageDiv = document.createElement('div');
|
642 |
messageDiv.className = `message ${sender}-message`;
|
643 |
-
|
644 |
-
// Use innerHTML to handle potential formatting like newlines or markdown
|
645 |
-
// (Basic textContent might be sufficient depending on expected AI output)
|
646 |
-
// For simplicity here, sticking to textContent as AI might output plain text
|
647 |
messageDiv.textContent = text;
|
648 |
-
|
649 |
-
// Basic handling for newlines
|
650 |
-
messageDiv.style.whiteSpace = 'pre-wrap';
|
651 |
-
|
652 |
chatContainer.appendChild(messageDiv);
|
653 |
-
chatContainer.scrollTop = chatContainer.scrollHeight;
|
654 |
}
|
655 |
-
|
656 |
function showLoading(text = 'Processing...') {
|
657 |
document.getElementById('loadingText').textContent = text;
|
658 |
document.getElementById('loading').style.display = 'block';
|
659 |
}
|
660 |
-
|
661 |
function hideLoading() {
|
662 |
document.getElementById('loading').style.display = 'none';
|
663 |
}
|
664 |
-
|
665 |
function showError(message) {
|
666 |
const errorDiv = document.getElementById('error');
|
667 |
errorDiv.textContent = message;
|
668 |
errorDiv.style.display = 'block';
|
669 |
-
// Auto-hide after 5 seconds
|
670 |
setTimeout(() => {
|
671 |
errorDiv.style.display = 'none';
|
672 |
}, 5000);
|
673 |
}
|
674 |
-
|
675 |
function showSuccess(message) {
|
676 |
const successDiv = document.getElementById('success');
|
677 |
successDiv.textContent = message;
|
678 |
successDiv.style.display = 'block';
|
679 |
-
// Auto-hide after 5 seconds
|
680 |
setTimeout(() => {
|
681 |
successDiv.style.display = 'none';
|
682 |
}, 5000);
|
683 |
}
|
684 |
-
|
685 |
-
function hideError() {
|
686 |
-
document.getElementById('error').style.display = 'none';
|
687 |
-
}
|
688 |
-
|
689 |
-
function hideSuccess() {
|
690 |
-
document.getElementById('success').style.display = 'none';
|
691 |
-
}
|
692 |
-
|
693 |
// Enter key to ask question
|
694 |
document.getElementById('questionInput').addEventListener('keypress', (e) => {
|
695 |
-
|
696 |
-
if (!document.getElementById('questionInput').disabled && e.key === 'Enter') {
|
697 |
-
e.preventDefault(); // Prevent default form submission if input is part of a form
|
698 |
askQuestion();
|
699 |
}
|
700 |
});
|
701 |
-
|
702 |
-
// Initial message visibility
|
703 |
-
// addMessage("👋 Hello! Upload a PDF document using the panel on the left, and I'll help you analyze it using Gemini API caching!", 'ai'); // Added this directly in HTML
|
704 |
</script>
|
705 |
</body>
|
706 |
</html>
|
707 |
"""
|
708 |
|
709 |
-
# --- Flask Routes ---
|
710 |
-
|
711 |
@app.route('/')
|
712 |
def index():
|
713 |
-
# Ensure API key is set before rendering, or add a warning to the template
|
714 |
-
if not api_key:
|
715 |
-
# You could modify the template or pass a variable to indicate error state
|
716 |
-
print("Warning: API key not set. API calls will fail.")
|
717 |
return render_template_string(HTML_TEMPLATE)
|
718 |
|
719 |
-
@app.route('/health', methods=['GET'])
|
720 |
-
def health_check():
|
721 |
-
# A simple endpoint to check if the application is running
|
722 |
-
# Can optionally check API client status if needed, but basic 200 is common.
|
723 |
-
if client is None and api_key is not None: # Client failed to initialize despite key being present
|
724 |
-
return jsonify({"status": "unhealthy", "reason": "Gemini client failed to initialize"}), 500
|
725 |
-
# Note: This doesn't check if the API key is *valid* or if the API is reachable,
|
726 |
-
# just if the Flask app is running and the client object was created.
|
727 |
-
return jsonify({"status": "healthy"}), 200
|
728 |
-
|
729 |
-
|
730 |
@app.route('/upload', methods=['POST'])
|
731 |
def upload_file():
|
732 |
-
if client is None or api_key is None:
|
733 |
-
return jsonify({'success': False, 'error': 'API key not configured or Gemini client failed to initialize.'}), 500
|
734 |
-
|
735 |
try:
|
736 |
if 'file' not in request.files:
|
737 |
return jsonify({'success': False, 'error': 'No file provided'})
|
738 |
-
|
739 |
file = request.files['file']
|
740 |
-
|
741 |
if file.filename == '':
|
742 |
return jsonify({'success': False, 'error': 'No file selected'})
|
743 |
-
|
744 |
# Read file content
|
745 |
file_content = file.read()
|
746 |
file_io = io.BytesIO(file_content)
|
747 |
-
|
748 |
-
#
|
749 |
-
|
750 |
-
|
751 |
-
|
752 |
-
|
753 |
-
|
754 |
-
# The mime_type is crucial for the API to correctly process the file.
|
755 |
-
# The filename is used as the display_name by default if not provided.
|
756 |
-
document = client.upload_file(
|
757 |
-
file=(file.filename, file_io, 'application/pdf'), # Use the 'file' argument with tuple format
|
758 |
-
# display_name=file.filename # Optional: explicitly provide a display name
|
759 |
-
)
|
760 |
-
print(f"File uploaded successfully to Gemini File API: {document.name}") # Log for debugging
|
761 |
-
# Note: client.upload_file returns a google.generativeai.types.File object
|
762 |
-
# which contains the resource name (e.g., 'files/xyz123').
|
763 |
-
except Exception as upload_error:
|
764 |
-
# Attempt to provide more specific feedback if possible
|
765 |
-
error_msg = str(upload_error)
|
766 |
-
print(f"Error uploading file to Gemini API: {error_msg}")
|
767 |
-
# Check for common upload errors like exceeding file size limits
|
768 |
-
if "file content size exceeds limit" in error_msg.lower():
|
769 |
-
return jsonify({'success': False, 'error': f'Error uploading file: File size exceeds API limit. {error_msg}'}), 413 # 413 Payload Too Large
|
770 |
-
return jsonify({'success': False, 'error': f'Error uploading file to Gemini API: {error_msg}'}), 500
|
771 |
-
# --- END CORRECTED FILE UPLOAD CALL ---
|
772 |
-
|
773 |
# Create cache with system instruction
|
774 |
-
cache = None # Initialize cache variable
|
775 |
try:
|
776 |
system_instruction = "You are an expert document analyzer. Provide detailed, accurate answers based on the uploaded document content. Always be helpful and thorough in your responses."
|
777 |
-
|
778 |
# Use the correct model format as per documentation
|
779 |
-
# Using a specific stable version is recommended for production
|
780 |
model = 'models/gemini-2.0-flash-001'
|
781 |
-
|
782 |
-
print(f"Attempting to create cache for file: {document.name}") # Log
|
783 |
cache = client.caches.create(
|
784 |
model=model,
|
785 |
config=types.CreateCachedContentConfig(
|
786 |
-
display_name=
|
787 |
system_instruction=system_instruction,
|
788 |
-
contents=[document],
|
789 |
-
ttl="3600s", # 1 hour TTL
|
790 |
)
|
791 |
)
|
792 |
-
|
793 |
-
|
794 |
-
# Store cache info in our in-memory dictionary
|
795 |
-
# We map our internal UUID cache_id to the Gemini API's cache.name (resource name)
|
796 |
cache_id = str(uuid.uuid4())
|
797 |
document_caches[cache_id] = {
|
798 |
-
'
|
799 |
'document_name': file.filename,
|
800 |
-
'
|
801 |
-
'created_at': datetime.now().isoformat(),
|
802 |
-
'expires_at': (datetime.now(timezone.utc) + timedelta(seconds=3600)).isoformat(), # Store expiry time for reference
|
803 |
}
|
804 |
-
|
805 |
-
# Get token count from cache metadata if available
|
806 |
-
# Note: cached_token_count might be available on the cache object after creation
|
807 |
-
token_count = 'Unknown'
|
808 |
-
if hasattr(cache, 'usage_metadata') and cache.usage_metadata:
|
809 |
-
token_count = getattr(cache.usage_metadata, 'cached_token_count', 'Unknown')
|
810 |
-
print(f"Cached token count: {token_count}")
|
811 |
-
|
812 |
-
|
813 |
return jsonify({
|
814 |
'success': True,
|
815 |
-
'cache_id': cache_id,
|
816 |
-
'token_count':
|
817 |
})
|
818 |
-
|
819 |
except Exception as cache_error:
|
820 |
-
|
821 |
-
|
822 |
-
|
823 |
-
|
824 |
-
|
825 |
-
|
826 |
-
|
827 |
-
except Exception as cleanup_error:
|
828 |
-
print(f"Failed to clean up file {document.name}: {cleanup_error}")
|
829 |
-
|
830 |
-
# Handle specific cache creation errors
|
831 |
-
# Note: The exact error message for content size can vary or might not be specific
|
832 |
-
# The documentation mentions minimum tokens for caching.
|
833 |
-
if "Cached content is too small" in error_msg or "minimum size" in error_msg.lower() or "tokens required" in error_msg.lower():
|
834 |
-
return jsonify({
|
835 |
-
'success': False,
|
836 |
-
'error': f'PDF content is too small for caching. Minimum token count varies by model, but is typically 1024+ for Flash. {error_msg}',
|
837 |
-
'suggestion': 'Try uploading a longer document or combine multiple documents.'
|
838 |
-
}), 400 # 400 Bad Request - client error
|
839 |
else:
|
840 |
-
|
841 |
-
|
842 |
-
|
843 |
-
|
844 |
except Exception as e:
|
845 |
-
|
846 |
-
return jsonify({'success': False, 'error': str(e)}), 500
|
847 |
|
848 |
@app.route('/upload-url', methods=['POST'])
|
849 |
def upload_from_url():
|
850 |
-
if client is None or api_key is None:
|
851 |
-
return jsonify({'success': False, 'error': 'API key not configured or Gemini client failed to initialize.'}), 500
|
852 |
-
|
853 |
try:
|
854 |
data = request.get_json()
|
855 |
url = data.get('url')
|
856 |
-
|
857 |
if not url:
|
858 |
-
return jsonify({'success': False, 'error': 'No URL provided'})
|
859 |
-
|
860 |
# Download file from URL
|
861 |
-
response =
|
862 |
-
|
863 |
-
|
864 |
-
# Add a timeout to prevent hanging on unresponsive URLs.
|
865 |
-
response = httpx.get(url, follow_redirects=True, timeout=30.0)
|
866 |
-
response.raise_for_status() # Raise HTTPError for bad responses (4xx or 5xx)
|
867 |
-
|
868 |
-
# Basic check for PDF mime type (optional but good practice)
|
869 |
-
content_type = response.headers.get('Content-Type', '').lower()
|
870 |
-
if 'application/pdf' not in content_type:
|
871 |
-
print(f"Warning: URL content type is not application/pdf: {content_type}")
|
872 |
-
# Decide if you want to block non-PDFs or try to upload anyway
|
873 |
-
# For now, we'll proceed but log a warning. API might reject it.
|
874 |
-
# If strictly PDF required, return an error here:
|
875 |
-
# return jsonify({'success': False, 'error': f'URL does not point to a PDF document (Content-Type: {content_type})'}), 415 # 415 Unsupported Media Type
|
876 |
-
|
877 |
-
|
878 |
-
except httpx.HTTPStatusError as e:
|
879 |
-
print(f"HTTP error downloading file from URL {url}: {e.response.status_code} - {e.response.text}")
|
880 |
-
return jsonify({'success': False, 'error': f'HTTP error downloading file from URL: {e.response.status_code} - {e.response.text}'}), e.response.status_code
|
881 |
-
except httpx.RequestError as e:
|
882 |
-
print(f"Error downloading file from URL {url}: {e}")
|
883 |
-
return jsonify({'success': False, 'error': f'Error downloading file from URL: {e}'}), 500
|
884 |
-
|
885 |
-
|
886 |
file_io = io.BytesIO(response.content)
|
887 |
-
|
888 |
-
#
|
889 |
-
|
890 |
-
|
891 |
-
|
892 |
-
|
893 |
-
|
894 |
-
# Attempt to get filename from URL or headers, otherwise use generic
|
895 |
-
filename = os.path.basename(url)
|
896 |
-
if not filename or '.' not in filename:
|
897 |
-
filename = 'downloaded_document.pdf' # Default generic name
|
898 |
-
|
899 |
-
# Use the mime type from the response headers if available and looks right
|
900 |
-
mime_type = content_type if 'application/pdf' in content_type else 'application/pdf'
|
901 |
-
|
902 |
-
|
903 |
-
document = client.upload_file(
|
904 |
-
file=(filename, file_io, mime_type), # Use parsed filename and mime_type
|
905 |
-
display_name=url # Use the URL as display name in Gemini API
|
906 |
-
)
|
907 |
-
print(f"File from URL uploaded successfully to Gemini File API: {document.name}") # Log
|
908 |
-
# Note: client.upload_file returns a google.generativeai.types.File object
|
909 |
-
# which contains the resource name (e.g., 'files/xyz123').
|
910 |
-
|
911 |
-
except Exception as upload_error:
|
912 |
-
# Attempt to provide more specific feedback if possible
|
913 |
-
error_msg = str(upload_error)
|
914 |
-
print(f"Error uploading file from URL to Gemini API: {error_msg}")
|
915 |
-
# Check for common upload errors like exceeding file size limits
|
916 |
-
if "file content size exceeds limit" in error_msg.lower():
|
917 |
-
return jsonify({'success': False, 'error': f'Error uploading file: File size exceeds API limit. {error_msg}'}), 413 # 413 Payload Too Large
|
918 |
-
return jsonify({'success': False, 'error': f'Error uploading file from URL to Gemini API: {error_msg}'}), 500
|
919 |
-
# --- END CORRECTED FILE UPLOAD CALL ---
|
920 |
-
|
921 |
-
|
922 |
# Create cache with system instruction
|
923 |
-
cache = None # Initialize cache variable
|
924 |
try:
|
925 |
system_instruction = "You are an expert document analyzer. Provide detailed, accurate answers based on the uploaded document content. Always be helpful and thorough in your responses."
|
926 |
-
|
927 |
# Use the correct model format as per documentation
|
928 |
model = 'models/gemini-2.0-flash-001'
|
929 |
-
|
930 |
-
print(f"Attempting to create cache for file: {document.name}") # Log
|
931 |
cache = client.caches.create(
|
932 |
model=model,
|
933 |
config=types.CreateCachedContentConfig(
|
934 |
-
display_name=
|
935 |
system_instruction=system_instruction,
|
936 |
-
contents=[document],
|
937 |
-
ttl="3600s", # 1 hour TTL
|
938 |
)
|
939 |
)
|
940 |
-
|
941 |
-
|
942 |
-
|
943 |
-
# Store cache info in our in-memory dictionary
|
944 |
-
# We map our internal UUID cache_id to the Gemini API's cache.name (resource name)
|
945 |
cache_id = str(uuid.uuid4())
|
946 |
document_caches[cache_id] = {
|
947 |
-
'
|
948 |
-
'document_name': url,
|
949 |
-
'
|
950 |
-
'created_at': datetime.now().isoformat(),
|
951 |
-
'expires_at': (datetime.now(timezone.utc) + timedelta(seconds=3600)).isoformat(), # Store expiry time for reference
|
952 |
}
|
953 |
-
|
954 |
-
# Get token count from cache metadata if available
|
955 |
-
token_count = 'Unknown'
|
956 |
-
if hasattr(cache, 'usage_metadata') and cache.usage_metadata:
|
957 |
-
token_count = getattr(cache.usage_metadata, 'cached_token_count', 'Unknown')
|
958 |
-
print(f"Cached token count: {token_count}")
|
959 |
-
|
960 |
-
|
961 |
return jsonify({
|
962 |
'success': True,
|
963 |
-
'cache_id': cache_id,
|
964 |
-
'token_count':
|
965 |
})
|
966 |
-
|
967 |
except Exception as cache_error:
|
968 |
-
|
969 |
-
|
970 |
-
|
971 |
-
|
972 |
-
|
973 |
-
|
974 |
-
|
975 |
-
except Exception as cleanup_error:
|
976 |
-
print(f"Failed to clean up file {document.name}: {cleanup_error}")
|
977 |
-
|
978 |
-
# Handle specific cache creation errors
|
979 |
-
if "Cached content is too small" in error_msg or "minimum size" in error_msg.lower() or "tokens required" in error_msg.lower():
|
980 |
-
return jsonify({
|
981 |
-
'success': False,
|
982 |
-
'error': f'PDF content is too small for caching. Minimum token count varies by model, but is typically 1024+ for Flash. {error_msg}',
|
983 |
-
'suggestion': 'Try uploading a longer document or combine multiple documents.'
|
984 |
-
}), 400 # 400 Bad Request - client error
|
985 |
else:
|
986 |
-
|
987 |
-
|
988 |
-
|
989 |
-
|
990 |
except Exception as e:
|
991 |
-
|
992 |
-
return jsonify({'success': False, 'error': str(e)}), 500
|
993 |
-
|
994 |
|
995 |
@app.route('/ask', methods=['POST'])
|
996 |
def ask_question():
|
997 |
-
if client is None or api_key is None:
|
998 |
-
return jsonify({'success': False, 'error': 'API key not configured or Gemini client failed to initialize.'}), 500
|
999 |
-
|
1000 |
try:
|
1001 |
data = request.get_json()
|
1002 |
question = data.get('question')
|
1003 |
cache_id = data.get('cache_id')
|
1004 |
-
|
1005 |
if not question or not cache_id:
|
1006 |
-
return jsonify({'success': False, 'error': 'Missing question or cache_id'})
|
1007 |
-
|
1008 |
-
# --- CORRECTED CACHE LOOKUP ---
|
1009 |
-
# Check if our internal cache_id exists in the in-memory dictionary
|
1010 |
if cache_id not in document_caches:
|
1011 |
-
|
1012 |
-
|
1013 |
-
print(f"Cache ID {cache_id} not found in local storage.")
|
1014 |
-
return jsonify({'success': False, 'error': 'Cache not found or expired. Please upload the document again.'}), 404 # 404 Not Found
|
1015 |
-
|
1016 |
-
# If found, retrieve the Gemini API cache name
|
1017 |
cache_info = document_caches[cache_id]
|
1018 |
-
|
1019 |
-
print(f"Using Gemini cache name: {gemini_cache_name} for question.")
|
1020 |
-
# --- END CORRECTED CACHE LOOKUP ---
|
1021 |
-
|
1022 |
# Generate response using cached content with correct model format
|
1023 |
response = client.models.generate_content(
|
1024 |
-
model='models/gemini-2.0-flash-001',
|
1025 |
-
contents=
|
1026 |
-
|
1027 |
-
cached_content=
|
1028 |
)
|
1029 |
)
|
1030 |
-
|
1031 |
-
# Check if response has parts before accessing .text
|
1032 |
-
answer = "Could not generate response from the model."
|
1033 |
-
if response and response.candidates:
|
1034 |
-
# Handle potential tool_code or other non-text parts if necessary
|
1035 |
-
answer_parts = []
|
1036 |
-
for candidate in response.candidates:
|
1037 |
-
if candidate.content and candidate.content.parts:
|
1038 |
-
for part in candidate.content.parts:
|
1039 |
-
if hasattr(part, 'text'):
|
1040 |
-
answer_parts.append(part.text)
|
1041 |
-
# Add handling for other part types if needed (e.g., tool_code, function_response)
|
1042 |
-
# elif hasattr(part, 'tool_code'):
|
1043 |
-
# answer_parts.append(f"\n```tool_code\n{part.tool_code.code}\n```\n")
|
1044 |
-
# elif hasattr(part, 'function_response'):
|
1045 |
-
# answer_parts.append(f"\n```function_response\n{json.dumps(part.function_response, indent=2)}\n```\n")
|
1046 |
-
if answer_parts:
|
1047 |
-
answer = "".join(answer_parts)
|
1048 |
-
else:
|
1049 |
-
# Handle cases where candidates exist but have no text parts (e.g., tool calls)
|
1050 |
-
answer = "Model returned content without text parts (e.g., tool calls)."
|
1051 |
-
print(f"Model returned non-text parts: {response.candidates}") # Log for debugging
|
1052 |
-
|
1053 |
-
elif response and response.prompt_feedback and response.prompt_feedback.block_reason:
|
1054 |
-
# Handle cases where the prompt was blocked
|
1055 |
-
block_reason = response.prompt_feedback.block_reason.name
|
1056 |
-
block_message = getattr(response.prompt_feedback, 'block_reason_message', 'No message provided')
|
1057 |
-
answer = f"Request blocked by safety filters. Reason: {block_reason}. Message: {block_message}"
|
1058 |
-
print(f"Request blocked: {block_reason} - {block_message}")
|
1059 |
-
|
1060 |
-
else:
|
1061 |
-
# Handle other unexpected response structures
|
1062 |
-
print(f"Unexpected response structure from API: {response}")
|
1063 |
-
# answer stays as the initial "Could not generate response..." message
|
1064 |
-
|
1065 |
return jsonify({
|
1066 |
'success': True,
|
1067 |
-
'answer':
|
1068 |
})
|
1069 |
-
|
1070 |
except Exception as e:
|
1071 |
-
|
1072 |
-
# Attempt to provide more specific API error messages
|
1073 |
-
error_msg = str(e)
|
1074 |
-
if "Resource has been exhausted" in error_msg:
|
1075 |
-
error_msg = "API rate limit or quota exceeded. Please try again later."
|
1076 |
-
elif "cached_content refers to a resource that has been deleted" in error_msg:
|
1077 |
-
error_msg = "The cached document has expired or was deleted from Gemini API. Please upload the document again."
|
1078 |
-
# Clean up local entry if API confirms deletion/expiry
|
1079 |
-
if cache_id in document_caches:
|
1080 |
-
print(f"Removing local entry for cache_id {cache_id} as API confirmed deletion.")
|
1081 |
-
del document_caches[cache_id]
|
1082 |
-
elif "invalid cached_content value" in error_msg:
|
1083 |
-
error_msg = "Invalid cache reference. The cached document might have expired or been deleted. Please upload the document again."
|
1084 |
-
# Clean up local entry if API confirms deletion/expiry
|
1085 |
-
if cache_id in document_caches:
|
1086 |
-
print(f"Removing local entry for cache_id {cache_id} as API confirmed deletion (invalid reference).")
|
1087 |
-
del document_caches[cache_id]
|
1088 |
-
elif "model does not exist" in error_msg:
|
1089 |
-
error_msg = "The specified model is not available."
|
1090 |
-
|
1091 |
-
|
1092 |
-
return jsonify({'success': False, 'error': f'Error from Gemini API: {error_msg}'}), 500 # 500 Internal Server Error
|
1093 |
-
|
1094 |
|
1095 |
@app.route('/caches', methods=['GET'])
|
1096 |
def list_caches():
|
1097 |
-
# Lists caches stored *in this application's memory*.
|
1098 |
-
# It does NOT list caches directly from the Gemini API unless you add that logic.
|
1099 |
try:
|
1100 |
caches = []
|
1101 |
-
for cache_id, cache_info in
|
1102 |
-
|
1103 |
-
|
1104 |
-
|
1105 |
-
|
1106 |
-
|
1107 |
-
|
1108 |
-
caches.append({
|
1109 |
-
'cache_id': cache_id, # Our internal ID
|
1110 |
-
'document_name': cache_info['document_name'],
|
1111 |
-
'gemini_cache_name': cache_info['gemini_cache_name'], # Include Gemini name
|
1112 |
-
'created_at': cache_info['created_at'],
|
1113 |
-
'expires_at': getattr(api_cache_info, 'expire_time', 'Unknown'), # Get actual expiry from API
|
1114 |
-
'cached_token_count': getattr(api_cache_info.usage_metadata, 'cached_token_count', 'Unknown') if hasattr(api_cache_info, 'usage_metadata') else 'Unknown'
|
1115 |
-
})
|
1116 |
-
except Exception as e:
|
1117 |
-
# If API lookup fails (e.g., cache expired/deleted), remove from our local map
|
1118 |
-
print(f"Gemini cache {cache_info['gemini_cache_name']} for local ID {cache_id} not found via API. Removing from local storage. Error: {e}")
|
1119 |
-
del document_caches[cache_id]
|
1120 |
-
# Don't add it to the list of active caches
|
1121 |
-
|
1122 |
return jsonify({'success': True, 'caches': caches})
|
1123 |
-
|
1124 |
except Exception as e:
|
1125 |
-
print(f"An error occurred listing caches: {str(e)}")
|
1126 |
return jsonify({'success': False, 'error': str(e)})
|
1127 |
|
1128 |
-
|
1129 |
@app.route('/cache/<cache_id>', methods=['DELETE'])
|
1130 |
def delete_cache(cache_id):
|
1131 |
-
if client is None or api_key is None:
|
1132 |
-
return jsonify({'success': False, 'error': 'API key not configured or Gemini client failed to initialize.'}), 500
|
1133 |
-
|
1134 |
try:
|
1135 |
if cache_id not in document_caches:
|
1136 |
-
return jsonify({'success': False, 'error': 'Cache not found'})
|
1137 |
-
|
1138 |
cache_info = document_caches[cache_id]
|
1139 |
-
|
1140 |
-
|
1141 |
-
|
1142 |
-
|
1143 |
-
#
|
1144 |
-
try:
|
1145 |
-
client.caches.delete(gemini_cache_name_to_delete)
|
1146 |
-
print(f"Gemini cache deleted: {gemini_cache_name_to_delete}") # Log
|
1147 |
-
except Exception as delete_error:
|
1148 |
-
error_msg = str(delete_error)
|
1149 |
-
print(f"Error deleting Gemini cache {gemini_cache_name_to_delete}: {error_msg}") # Log
|
1150 |
-
# Handle case where the cache was already gone (e.g. expired)
|
1151 |
-
if "Resource not found" in error_msg:
|
1152 |
-
print(f"Gemini cache {gemini_cache_name_to_delete} already gone from API.")
|
1153 |
-
else:
|
1154 |
-
# For other errors, you might want to stop and return the error
|
1155 |
-
return jsonify({'success': False, 'error': f'Failed to delete cache from API: {error_msg}'}), 500
|
1156 |
-
|
1157 |
-
|
1158 |
-
# Also delete the associated file from Gemini File API to free up storage
|
1159 |
-
if gemini_file_name_to_delete:
|
1160 |
-
try:
|
1161 |
-
client.files.delete(gemini_file_name_to_delete)
|
1162 |
-
print(f"Associated Gemini file deleted: {gemini_file_name_to_delete}") # Log
|
1163 |
-
except Exception as file_delete_error:
|
1164 |
-
error_msg = str(file_delete_error)
|
1165 |
-
print(f"Error deleting Gemini file {gemini_file_name_to_delete}: {error_msg}") # Log
|
1166 |
-
if "Resource not found" in error_msg:
|
1167 |
-
print(f"Gemini file {gemini_file_name_to_delete} already gone from API.")
|
1168 |
-
else:
|
1169 |
-
# Log but continue, deleting the cache is the primary goal
|
1170 |
-
pass
|
1171 |
-
|
1172 |
-
|
1173 |
-
# Remove from local storage *after* attempting API deletion
|
1174 |
del document_caches[cache_id]
|
1175 |
-
|
1176 |
-
|
1177 |
-
|
1178 |
-
|
1179 |
except Exception as e:
|
1180 |
-
|
1181 |
-
return jsonify({'success': False, 'error': str(e)}), 500
|
1182 |
-
|
1183 |
|
1184 |
if __name__ == '__main__':
|
1185 |
import os
|
1186 |
port = int(os.environ.get("PORT", 7860))
|
1187 |
-
|
1188 |
-
# In production, set debug=False
|
1189 |
-
# Use threaded=True or a production WSGI server (like Gunicorn) for concurrent requests
|
1190 |
-
app.run(debug=True, host='0.0.0.0', port=port, threaded=True)
|
|
|
10 |
from dotenv import load_dotenv
|
11 |
import json
|
12 |
|
13 |
+
# Load environment variables
|
|
|
|
|
14 |
load_dotenv()
|
15 |
|
16 |
app = Flask(__name__)
|
17 |
CORS(app)
|
18 |
|
19 |
# Initialize Gemini client
|
20 |
+
client = genai.Client(api_key=os.getenv('GOOGLE_API_KEY'))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
# In-memory storage for demo (in production, use a database)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
document_caches = {}
|
24 |
+
user_sessions = {}
|
25 |
|
26 |
# HTML template for the web interface
|
27 |
HTML_TEMPLATE = """
|
|
|
31 |
<meta charset="UTF-8">
|
32 |
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
33 |
<title>Smart Document Analysis Platform</title>
|
|
|
34 |
<style>
|
35 |
* {
|
36 |
margin: 0;
|
37 |
padding: 0;
|
38 |
box-sizing: border-box;
|
39 |
}
|
40 |
+
|
41 |
body {
|
42 |
font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
|
43 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
44 |
min-height: 100vh;
|
45 |
color: #333;
|
|
|
46 |
}
|
47 |
+
|
48 |
.container {
|
49 |
max-width: 1400px;
|
50 |
margin: 0 auto;
|
51 |
padding: 20px;
|
52 |
min-height: 100vh;
|
|
|
|
|
53 |
}
|
54 |
+
|
55 |
.header {
|
56 |
text-align: center;
|
57 |
margin-bottom: 30px;
|
58 |
color: white;
|
59 |
}
|
60 |
+
|
61 |
.header h1 {
|
62 |
font-size: 2.8em;
|
63 |
font-weight: 700;
|
64 |
margin-bottom: 10px;
|
65 |
text-shadow: 0 2px 4px rgba(0,0,0,0.3);
|
66 |
}
|
67 |
+
|
68 |
.header p {
|
69 |
font-size: 1.2em;
|
70 |
opacity: 0.9;
|
71 |
font-weight: 300;
|
72 |
}
|
73 |
+
|
74 |
.main-content {
|
75 |
display: grid;
|
76 |
grid-template-columns: 1fr 1fr;
|
77 |
gap: 30px;
|
78 |
+
height: calc(100vh - 200px);
|
79 |
}
|
80 |
+
|
81 |
+
.left-panel {
|
82 |
+
background: white;
|
83 |
+
border-radius: 20px;
|
84 |
+
padding: 30px;
|
85 |
+
box-shadow: 0 20px 40px rgba(0,0,0,0.1);
|
86 |
+
overflow-y: auto;
|
87 |
+
}
|
88 |
+
|
89 |
+
.right-panel {
|
90 |
background: white;
|
91 |
border-radius: 20px;
|
92 |
padding: 30px;
|
|
|
94 |
display: flex;
|
95 |
flex-direction: column;
|
96 |
}
|
97 |
+
|
|
|
|
|
|
|
|
|
98 |
.panel-title {
|
99 |
font-size: 1.5em;
|
100 |
font-weight: 600;
|
|
|
104 |
align-items: center;
|
105 |
gap: 10px;
|
106 |
}
|
107 |
+
|
108 |
.upload-section {
|
109 |
margin-bottom: 30px;
|
110 |
}
|
111 |
+
|
112 |
.upload-area {
|
113 |
border: 2px dashed #667eea;
|
114 |
border-radius: 15px;
|
|
|
117 |
background: #f8fafc;
|
118 |
transition: all 0.3s ease;
|
119 |
margin-bottom: 20px;
|
|
|
120 |
}
|
121 |
+
|
122 |
.upload-area:hover {
|
123 |
border-color: #764ba2;
|
124 |
background: #f0f2ff;
|
125 |
transform: translateY(-2px);
|
126 |
}
|
127 |
+
|
128 |
.upload-area.dragover {
|
129 |
border-color: #764ba2;
|
130 |
background: #e8f0ff;
|
131 |
transform: scale(1.02);
|
132 |
}
|
133 |
+
|
134 |
.upload-icon {
|
135 |
font-size: 3em;
|
136 |
color: #667eea;
|
137 |
margin-bottom: 15px;
|
138 |
}
|
139 |
+
|
140 |
.file-input {
|
141 |
display: none;
|
142 |
}
|
143 |
+
|
144 |
.upload-btn {
|
145 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
146 |
color: white;
|
|
|
153 |
transition: all 0.3s ease;
|
154 |
margin: 10px;
|
155 |
}
|
156 |
+
|
157 |
.upload-btn:hover {
|
158 |
transform: translateY(-2px);
|
159 |
box-shadow: 0 10px 20px rgba(102, 126, 234, 0.3);
|
160 |
}
|
161 |
+
|
162 |
.url-input {
|
163 |
width: 100%;
|
164 |
padding: 15px;
|
|
|
168 |
margin-bottom: 15px;
|
169 |
transition: border-color 0.3s ease;
|
170 |
}
|
171 |
+
|
172 |
.url-input:focus {
|
173 |
outline: none;
|
174 |
border-color: #667eea;
|
175 |
}
|
176 |
+
|
177 |
.btn {
|
178 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
179 |
color: white;
|
|
|
185 |
font-weight: 500;
|
186 |
transition: all 0.3s ease;
|
187 |
}
|
188 |
+
|
189 |
.btn:hover {
|
190 |
transform: translateY(-1px);
|
191 |
box-shadow: 0 5px 15px rgba(102, 126, 234, 0.3);
|
192 |
}
|
193 |
+
|
194 |
.btn:disabled {
|
195 |
opacity: 0.6;
|
196 |
cursor: not-allowed;
|
197 |
transform: none;
|
198 |
}
|
199 |
+
|
200 |
.chat-container {
|
201 |
flex: 1;
|
202 |
border: 1px solid #e2e8f0;
|
|
|
205 |
padding: 20px;
|
206 |
background: #f8fafc;
|
207 |
margin-bottom: 20px;
|
|
|
|
|
208 |
}
|
209 |
+
|
210 |
.message {
|
211 |
margin-bottom: 15px;
|
212 |
padding: 15px;
|
213 |
border-radius: 12px;
|
214 |
max-width: 85%;
|
215 |
animation: fadeIn 0.3s ease;
|
|
|
216 |
}
|
217 |
+
|
218 |
@keyframes fadeIn {
|
219 |
from { opacity: 0; transform: translateY(10px); }
|
220 |
to { opacity: 1; transform: translateY(0); }
|
221 |
}
|
222 |
+
|
223 |
.user-message {
|
224 |
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
|
225 |
color: white;
|
226 |
margin-left: auto;
|
227 |
box-shadow: 0 4px 12px rgba(102, 126, 234, 0.3);
|
228 |
}
|
229 |
+
|
230 |
.ai-message {
|
231 |
background: white;
|
232 |
color: #333;
|
233 |
border: 1px solid #e2e8f0;
|
234 |
box-shadow: 0 2px 8px rgba(0,0,0,0.1);
|
|
|
235 |
}
|
236 |
+
|
237 |
.input-group {
|
238 |
display: flex;
|
239 |
gap: 10px;
|
240 |
}
|
241 |
+
|
242 |
.question-input {
|
243 |
flex: 1;
|
244 |
padding: 15px;
|
|
|
247 |
font-size: 1em;
|
248 |
transition: border-color 0.3s ease;
|
249 |
}
|
250 |
+
|
251 |
.question-input:focus {
|
252 |
outline: none;
|
253 |
border-color: #667eea;
|
254 |
}
|
255 |
+
|
256 |
.cache-info {
|
257 |
background: linear-gradient(135deg, #48bb78 0%, #38a169 100%);
|
258 |
border-radius: 12px;
|
|
|
261 |
color: white;
|
262 |
box-shadow: 0 4px 12px rgba(72, 187, 120, 0.3);
|
263 |
}
|
264 |
+
|
265 |
.cache-info h3 {
|
266 |
margin-bottom: 10px;
|
267 |
font-weight: 600;
|
268 |
}
|
269 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
270 |
.loading {
|
271 |
text-align: center;
|
272 |
padding: 40px;
|
273 |
color: #666;
|
274 |
}
|
275 |
+
|
276 |
.loading-spinner {
|
277 |
border: 3px solid #f3f3f3;
|
278 |
border-top: 3px solid #667eea;
|
|
|
282 |
animation: spin 1s linear infinite;
|
283 |
margin: 0 auto 20px;
|
284 |
}
|
285 |
+
|
286 |
@keyframes spin {
|
287 |
0% { transform: rotate(0deg); }
|
288 |
100% { transform: rotate(360deg); }
|
289 |
}
|
290 |
+
|
291 |
.error {
|
292 |
background: linear-gradient(135deg, #f56565 0%, #e53e3e 100%);
|
293 |
border-radius: 12px;
|
|
|
296 |
margin-bottom: 20px;
|
297 |
box-shadow: 0 4px 12px rgba(245, 101, 101, 0.3);
|
298 |
}
|
299 |
+
|
300 |
.success {
|
301 |
background: linear-gradient(135deg, #48bb78 0%, #38a169 100%);
|
302 |
border-radius: 12px;
|
|
|
305 |
margin-bottom: 20px;
|
306 |
box-shadow: 0 4px 12px rgba(72, 187, 120, 0.3);
|
307 |
}
|
308 |
+
|
309 |
@media (max-width: 768px) {
|
310 |
.main-content {
|
311 |
grid-template-columns: 1fr;
|
312 |
gap: 20px;
|
313 |
}
|
314 |
+
|
315 |
.header h1 {
|
316 |
font-size: 2em;
|
317 |
}
|
|
|
323 |
<div class="header">
|
324 |
<h1>📚 Smart Document Analysis Platform</h1>
|
325 |
<p>Upload PDF documents once, ask questions forever with Gemini API caching</p>
|
|
|
326 |
</div>
|
327 |
+
|
328 |
<div class="main-content">
|
329 |
<!-- Left Panel - Upload Section -->
|
330 |
<div class="left-panel">
|
331 |
<div class="panel-title">
|
332 |
📤 Upload PDF Document
|
333 |
</div>
|
334 |
+
|
335 |
<div class="upload-section">
|
336 |
<div class="upload-area" id="uploadArea">
|
337 |
<div class="upload-icon">📄</div>
|
338 |
<p>Drag and drop your PDF file here, or click to select</p>
|
339 |
<input type="file" id="fileInput" class="file-input" accept=".pdf">
|
340 |
+
<button class="upload-btn" onclick="document.getElementById('fileInput').click()">
|
|
|
341 |
Choose PDF File
|
342 |
</button>
|
343 |
</div>
|
344 |
+
|
345 |
<div style="margin-top: 20px;">
|
346 |
<h3>Or provide a URL:</h3>
|
347 |
<input type="url" id="urlInput" class="url-input" placeholder="https://example.com/document.pdf">
|
348 |
+
<button class="btn" onclick="uploadFromUrl()">Upload from URL</button>
|
349 |
</div>
|
350 |
</div>
|
351 |
+
|
352 |
<div id="loading" class="loading" style="display: none;">
|
353 |
<div class="loading-spinner"></div>
|
354 |
<p id="loadingText">Processing your PDF... This may take a moment.</p>
|
355 |
</div>
|
356 |
+
|
357 |
<div id="error" class="error" style="display: none;"></div>
|
358 |
<div id="success" class="success" style="display: none;"></div>
|
359 |
</div>
|
360 |
+
|
361 |
<!-- Right Panel - Chat Section -->
|
362 |
<div class="right-panel">
|
363 |
<div class="panel-title">
|
364 |
💬 Ask Questions
|
365 |
</div>
|
366 |
+
|
367 |
<div id="cacheInfo" class="cache-info" style="display: none;">
|
368 |
<h3>✅ Document Cached Successfully!</h3>
|
369 |
<p>Your PDF has been cached using Gemini API. You can now ask multiple questions without re-uploading.</p>
|
370 |
<p><strong>Cache ID:</strong> <span id="cacheId"></span></p>
|
371 |
<p><strong>Tokens Cached:</strong> <span id="tokenCount"></span></p>
|
|
|
372 |
</div>
|
373 |
+
|
374 |
<div class="chat-container" id="chatContainer">
|
375 |
<div class="message ai-message">
|
376 |
+
👋 Hello! I'm ready to analyze your PDF documents. Upload a document to get started!
|
377 |
</div>
|
378 |
</div>
|
379 |
+
|
380 |
<div class="input-group">
|
381 |
+
<input type="text" id="questionInput" class="question-input" placeholder="Ask a question about your document...">
|
382 |
+
<button class="btn" onclick="askQuestion()" id="askBtn">Ask</button>
|
383 |
</div>
|
384 |
</div>
|
385 |
</div>
|
386 |
</div>
|
387 |
+
|
388 |
<script>
|
389 |
let currentCacheId = null;
|
390 |
+
|
|
|
|
|
|
|
|
|
391 |
// File upload handling
|
392 |
const uploadArea = document.getElementById('uploadArea');
|
393 |
const fileInput = document.getElementById('fileInput');
|
394 |
+
|
395 |
+
uploadArea.addEventListener('dragover', (e) => {
|
|
|
|
|
|
|
|
|
|
|
396 |
e.preventDefault();
|
397 |
+
uploadArea.classList.add('dragover');
|
|
|
|
|
|
|
|
|
|
|
398 |
});
|
399 |
+
|
400 |
+
uploadArea.addEventListener('dragleave', () => {
|
401 |
+
uploadArea.classList.remove('dragover');
|
402 |
});
|
403 |
+
|
404 |
+
uploadArea.addEventListener('drop', (e) => {
|
405 |
+
e.preventDefault();
|
406 |
+
uploadArea.classList.remove('dragover');
|
407 |
+
const files = e.dataTransfer.files;
|
|
|
|
|
408 |
if (files.length > 0) {
|
409 |
uploadFile(files[0]);
|
410 |
}
|
411 |
+
});
|
412 |
+
|
413 |
fileInput.addEventListener('change', (e) => {
|
414 |
if (e.target.files.length > 0) {
|
415 |
uploadFile(e.target.files[0]);
|
|
|
|
|
416 |
}
|
417 |
});
|
418 |
+
|
419 |
async function uploadFile(file) {
|
420 |
if (!file.type.includes('pdf')) {
|
421 |
showError('Please select a PDF file.');
|
422 |
return;
|
423 |
}
|
424 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
425 |
showLoading('Uploading PDF...');
|
426 |
+
|
427 |
const formData = new FormData();
|
428 |
formData.append('file', file);
|
429 |
+
|
430 |
try {
|
431 |
const response = await fetch('/upload', {
|
432 |
method: 'POST',
|
433 |
body: formData
|
434 |
});
|
435 |
+
|
436 |
const result = await response.json();
|
437 |
+
|
438 |
if (result.success) {
|
439 |
currentCacheId = result.cache_id;
|
440 |
document.getElementById('cacheId').textContent = result.cache_id;
|
441 |
document.getElementById('tokenCount').textContent = result.token_count;
|
442 |
document.getElementById('cacheInfo').style.display = 'block';
|
443 |
+
showSuccess('PDF uploaded and cached successfully!');
|
444 |
+
|
|
|
|
|
|
|
|
|
|
|
445 |
// Add initial message
|
446 |
addMessage("I've analyzed your PDF document. What would you like to know about it?", 'ai');
|
|
|
447 |
} else {
|
448 |
showError(result.error);
|
|
|
|
|
|
|
449 |
}
|
450 |
} catch (error) {
|
451 |
showError('Error uploading file: ' + error.message);
|
|
|
|
|
|
|
452 |
} finally {
|
453 |
hideLoading();
|
454 |
}
|
455 |
}
|
456 |
+
|
457 |
async function uploadFromUrl() {
|
458 |
const url = document.getElementById('urlInput').value;
|
459 |
+
if (!url) {
|
460 |
showError('Please enter a valid URL.');
|
461 |
return;
|
462 |
}
|
463 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
464 |
showLoading('Uploading PDF from URL...');
|
465 |
+
|
466 |
try {
|
467 |
const response = await fetch('/upload-url', {
|
468 |
method: 'POST',
|
|
|
471 |
},
|
472 |
body: JSON.stringify({ url: url })
|
473 |
});
|
474 |
+
|
475 |
const result = await response.json();
|
476 |
+
|
477 |
if (result.success) {
|
478 |
currentCacheId = result.cache_id;
|
479 |
document.getElementById('cacheId').textContent = result.cache_id;
|
480 |
document.getElementById('tokenCount').textContent = result.token_count;
|
481 |
document.getElementById('cacheInfo').style.display = 'block';
|
482 |
+
showSuccess('PDF uploaded and cached successfully!');
|
483 |
+
|
|
|
|
|
|
|
|
|
|
|
484 |
// Add initial message
|
485 |
addMessage("I've analyzed your PDF document. What would you like to know about it?", 'ai');
|
|
|
486 |
} else {
|
487 |
showError(result.error);
|
|
|
|
|
|
|
488 |
}
|
489 |
} catch (error) {
|
490 |
showError('Error uploading from URL: ' + error.message);
|
|
|
|
|
|
|
491 |
} finally {
|
492 |
hideLoading();
|
493 |
}
|
494 |
}
|
495 |
+
|
496 |
async function askQuestion() {
|
497 |
+
const question = document.getElementById('questionInput').value;
|
498 |
+
if (!question.trim()) return;
|
499 |
+
|
|
|
500 |
if (!currentCacheId) {
|
501 |
showError('Please upload a PDF document first.');
|
502 |
return;
|
503 |
}
|
504 |
+
|
505 |
// Add user message to chat
|
506 |
addMessage(question, 'user');
|
507 |
+
document.getElementById('questionInput').value = '';
|
508 |
+
|
509 |
// Show loading state
|
510 |
const askBtn = document.getElementById('askBtn');
|
511 |
const originalText = askBtn.textContent;
|
512 |
askBtn.textContent = 'Generating...';
|
513 |
askBtn.disabled = true;
|
514 |
+
|
|
|
515 |
try {
|
516 |
const response = await fetch('/ask', {
|
517 |
method: 'POST',
|
|
|
520 |
},
|
521 |
body: JSON.stringify({
|
522 |
question: question,
|
523 |
+
cache_id: currentCacheId
|
524 |
})
|
525 |
});
|
526 |
+
|
527 |
const result = await response.json();
|
528 |
+
|
529 |
if (result.success) {
|
530 |
addMessage(result.answer, 'ai');
|
531 |
} else {
|
|
|
536 |
} finally {
|
537 |
askBtn.textContent = originalText;
|
538 |
askBtn.disabled = false;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
539 |
}
|
540 |
}
|
541 |
+
|
542 |
function addMessage(text, sender) {
|
543 |
const chatContainer = document.getElementById('chatContainer');
|
544 |
const messageDiv = document.createElement('div');
|
545 |
messageDiv.className = `message ${sender}-message`;
|
|
|
|
|
|
|
|
|
546 |
messageDiv.textContent = text;
|
|
|
|
|
|
|
|
|
547 |
chatContainer.appendChild(messageDiv);
|
548 |
+
chatContainer.scrollTop = chatContainer.scrollHeight;
|
549 |
}
|
550 |
+
|
551 |
function showLoading(text = 'Processing...') {
|
552 |
document.getElementById('loadingText').textContent = text;
|
553 |
document.getElementById('loading').style.display = 'block';
|
554 |
}
|
555 |
+
|
556 |
function hideLoading() {
|
557 |
document.getElementById('loading').style.display = 'none';
|
558 |
}
|
559 |
+
|
560 |
function showError(message) {
|
561 |
const errorDiv = document.getElementById('error');
|
562 |
errorDiv.textContent = message;
|
563 |
errorDiv.style.display = 'block';
|
|
|
564 |
setTimeout(() => {
|
565 |
errorDiv.style.display = 'none';
|
566 |
}, 5000);
|
567 |
}
|
568 |
+
|
569 |
function showSuccess(message) {
|
570 |
const successDiv = document.getElementById('success');
|
571 |
successDiv.textContent = message;
|
572 |
successDiv.style.display = 'block';
|
|
|
573 |
setTimeout(() => {
|
574 |
successDiv.style.display = 'none';
|
575 |
}, 5000);
|
576 |
}
|
577 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
578 |
// Enter key to ask question
|
579 |
document.getElementById('questionInput').addEventListener('keypress', (e) => {
|
580 |
+
if (e.key === 'Enter') {
|
|
|
|
|
581 |
askQuestion();
|
582 |
}
|
583 |
});
|
|
|
|
|
|
|
584 |
</script>
|
585 |
</body>
|
586 |
</html>
|
587 |
"""
|
588 |
|
|
|
|
|
589 |
@app.route('/')
|
590 |
def index():
|
|
|
|
|
|
|
|
|
591 |
return render_template_string(HTML_TEMPLATE)
|
592 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
593 |
@app.route('/upload', methods=['POST'])
|
594 |
def upload_file():
|
|
|
|
|
|
|
595 |
try:
|
596 |
if 'file' not in request.files:
|
597 |
return jsonify({'success': False, 'error': 'No file provided'})
|
598 |
+
|
599 |
file = request.files['file']
|
600 |
+
|
601 |
if file.filename == '':
|
602 |
return jsonify({'success': False, 'error': 'No file selected'})
|
603 |
+
|
604 |
# Read file content
|
605 |
file_content = file.read()
|
606 |
file_io = io.BytesIO(file_content)
|
607 |
+
|
608 |
+
# Upload to Gemini File API
|
609 |
+
document = client.files.upload(
|
610 |
+
file=file_io,
|
611 |
+
config=dict(mime_type='application/pdf')
|
612 |
+
)
|
613 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
614 |
# Create cache with system instruction
|
|
|
615 |
try:
|
616 |
system_instruction = "You are an expert document analyzer. Provide detailed, accurate answers based on the uploaded document content. Always be helpful and thorough in your responses."
|
617 |
+
|
618 |
# Use the correct model format as per documentation
|
|
|
619 |
model = 'models/gemini-2.0-flash-001'
|
620 |
+
|
|
|
621 |
cache = client.caches.create(
|
622 |
model=model,
|
623 |
config=types.CreateCachedContentConfig(
|
624 |
+
display_name='pdf document cache',
|
625 |
system_instruction=system_instruction,
|
626 |
+
contents=[document],
|
627 |
+
ttl="3600s", # 1 hour TTL
|
628 |
)
|
629 |
)
|
630 |
+
|
631 |
+
# Store cache info
|
|
|
|
|
632 |
cache_id = str(uuid.uuid4())
|
633 |
document_caches[cache_id] = {
|
634 |
+
'cache_name': cache.name,
|
635 |
'document_name': file.filename,
|
636 |
+
'created_at': datetime.now().isoformat()
|
|
|
|
|
637 |
}
|
638 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
639 |
return jsonify({
|
640 |
'success': True,
|
641 |
+
'cache_id': cache_id,
|
642 |
+
'token_count': getattr(cache.usage_metadata, 'cached_token_count', 'Unknown')
|
643 |
})
|
644 |
+
|
645 |
except Exception as cache_error:
|
646 |
+
# If caching fails due to small content, provide alternative approach
|
647 |
+
if "Cached content is too small" in str(cache_error):
|
648 |
+
return jsonify({
|
649 |
+
'success': False,
|
650 |
+
'error': 'PDF is too small for caching. Please upload a larger document (minimum 4,096 tokens required).',
|
651 |
+
'suggestion': 'Try uploading a longer document or combine multiple documents.'
|
652 |
+
})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
653 |
else:
|
654 |
+
raise cache_error
|
655 |
+
|
|
|
|
|
656 |
except Exception as e:
|
657 |
+
return jsonify({'success': False, 'error': str(e)})
|
|
|
658 |
|
659 |
@app.route('/upload-url', methods=['POST'])
|
660 |
def upload_from_url():
|
|
|
|
|
|
|
661 |
try:
|
662 |
data = request.get_json()
|
663 |
url = data.get('url')
|
664 |
+
|
665 |
if not url:
|
666 |
+
return jsonify({'success': False, 'error': 'No URL provided'})
|
667 |
+
|
668 |
# Download file from URL
|
669 |
+
response = httpx.get(url)
|
670 |
+
response.raise_for_status()
|
671 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
672 |
file_io = io.BytesIO(response.content)
|
673 |
+
|
674 |
+
# Upload to Gemini File API
|
675 |
+
document = client.files.upload(
|
676 |
+
file=file_io,
|
677 |
+
config=dict(mime_type='application/pdf')
|
678 |
+
)
|
679 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
680 |
# Create cache with system instruction
|
|
|
681 |
try:
|
682 |
system_instruction = "You are an expert document analyzer. Provide detailed, accurate answers based on the uploaded document content. Always be helpful and thorough in your responses."
|
683 |
+
|
684 |
# Use the correct model format as per documentation
|
685 |
model = 'models/gemini-2.0-flash-001'
|
686 |
+
|
|
|
687 |
cache = client.caches.create(
|
688 |
model=model,
|
689 |
config=types.CreateCachedContentConfig(
|
690 |
+
display_name='pdf document cache',
|
691 |
system_instruction=system_instruction,
|
692 |
+
contents=[document],
|
693 |
+
ttl="3600s", # 1 hour TTL
|
694 |
)
|
695 |
)
|
696 |
+
|
697 |
+
# Store cache info
|
|
|
|
|
|
|
698 |
cache_id = str(uuid.uuid4())
|
699 |
document_caches[cache_id] = {
|
700 |
+
'cache_name': cache.name,
|
701 |
+
'document_name': url,
|
702 |
+
'created_at': datetime.now().isoformat()
|
|
|
|
|
703 |
}
|
704 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
705 |
return jsonify({
|
706 |
'success': True,
|
707 |
+
'cache_id': cache_id,
|
708 |
+
'token_count': getattr(cache.usage_metadata, 'cached_token_count', 'Unknown')
|
709 |
})
|
710 |
+
|
711 |
except Exception as cache_error:
|
712 |
+
# If caching fails due to small content, provide alternative approach
|
713 |
+
if "Cached content is too small" in str(cache_error):
|
714 |
+
return jsonify({
|
715 |
+
'success': False,
|
716 |
+
'error': 'PDF is too small for caching. Please upload a larger document (minimum 4,096 tokens required).',
|
717 |
+
'suggestion': 'Try uploading a longer document or combine multiple documents.'
|
718 |
+
})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
719 |
else:
|
720 |
+
raise cache_error
|
721 |
+
|
|
|
|
|
722 |
except Exception as e:
|
723 |
+
return jsonify({'success': False, 'error': str(e)})
|
|
|
|
|
724 |
|
725 |
@app.route('/ask', methods=['POST'])
|
726 |
def ask_question():
|
|
|
|
|
|
|
727 |
try:
|
728 |
data = request.get_json()
|
729 |
question = data.get('question')
|
730 |
cache_id = data.get('cache_id')
|
731 |
+
|
732 |
if not question or not cache_id:
|
733 |
+
return jsonify({'success': False, 'error': 'Missing question or cache_id'})
|
734 |
+
|
|
|
|
|
735 |
if cache_id not in document_caches:
|
736 |
+
return jsonify({'success': False, 'error': 'Cache not found'})
|
737 |
+
|
|
|
|
|
|
|
|
|
738 |
cache_info = document_caches[cache_id]
|
739 |
+
|
|
|
|
|
|
|
740 |
# Generate response using cached content with correct model format
|
741 |
response = client.models.generate_content(
|
742 |
+
model='models/gemini-2.0-flash-001',
|
743 |
+
contents=question,
|
744 |
+
config=types.GenerateContentConfig(
|
745 |
+
cached_content=cache_info['cache_name']
|
746 |
)
|
747 |
)
|
748 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
749 |
return jsonify({
|
750 |
'success': True,
|
751 |
+
'answer': response.text
|
752 |
})
|
753 |
+
|
754 |
except Exception as e:
|
755 |
+
return jsonify({'success': False, 'error': str(e)})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
756 |
|
757 |
@app.route('/caches', methods=['GET'])
|
758 |
def list_caches():
|
|
|
|
|
759 |
try:
|
760 |
caches = []
|
761 |
+
for cache_id, cache_info in document_caches.items():
|
762 |
+
caches.append({
|
763 |
+
'cache_id': cache_id,
|
764 |
+
'document_name': cache_info['document_name'],
|
765 |
+
'created_at': cache_info['created_at']
|
766 |
+
})
|
767 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
768 |
return jsonify({'success': True, 'caches': caches})
|
769 |
+
|
770 |
except Exception as e:
|
|
|
771 |
return jsonify({'success': False, 'error': str(e)})
|
772 |
|
|
|
773 |
@app.route('/cache/<cache_id>', methods=['DELETE'])
|
774 |
def delete_cache(cache_id):
|
|
|
|
|
|
|
775 |
try:
|
776 |
if cache_id not in document_caches:
|
777 |
+
return jsonify({'success': False, 'error': 'Cache not found'})
|
778 |
+
|
779 |
cache_info = document_caches[cache_id]
|
780 |
+
|
781 |
+
# Delete from Gemini API
|
782 |
+
client.caches.delete(cache_info['cache_name'])
|
783 |
+
|
784 |
+
# Remove from local storage
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
785 |
del document_caches[cache_id]
|
786 |
+
|
787 |
+
return jsonify({'success': True, 'message': 'Cache deleted successfully'})
|
788 |
+
|
|
|
789 |
except Exception as e:
|
790 |
+
return jsonify({'success': False, 'error': str(e)})
|
|
|
|
|
791 |
|
792 |
if __name__ == '__main__':
|
793 |
import os
|
794 |
port = int(os.environ.get("PORT", 7860))
|
795 |
+
app.run(debug=True, host='0.0.0.0', port=port)
|
|
|
|
|
|