amadmateen commited on
Commit
7db73fd
Β·
1 Parent(s): ccc7737

Refactor: Production-grade SatVision backend modularization, structured logging, boot safety validations, and workspace cleanup

Browse files
main.py CHANGED
@@ -15,7 +15,7 @@ from src.app import create_app
15
 
16
  def main():
17
  """Main entry point for the backend server."""
18
- print("πŸš€ SatVision Production Backend Running...", flush=True)
19
 
20
  # Create application instance
21
  app = create_app()
 
15
 
16
  def main():
17
  """Main entry point for the backend server."""
18
+ print("[SatVision] Production Backend Running...", flush=True)
19
 
20
  # Create application instance
21
  app = create_app()
server.py DELETED
@@ -1,790 +0,0 @@
1
- import os
2
- import json
3
- import numpy as np
4
- import tifffile as tiff
5
- import cv2
6
- import torch
7
- import pytorch_lightning as pl
8
- import segmentation_models_pytorch as smp
9
- import ee
10
- import requests
11
- from datetime import datetime, timedelta
12
- from flask import Flask, request, Response, stream_with_context, send_from_directory
13
- from flask_cors import CORS
14
- import math
15
- from google import genai
16
- from google.genai import types
17
- from PIL import Image
18
- from fpdf import FPDF
19
- from fpdf.enums import XPos, YPos
20
- import uuid
21
- from pymongo import MongoClient
22
-
23
- app = Flask(__name__)
24
- CORS(app)
25
-
26
- @app.route("/")
27
- def health_check():
28
- return "βœ… SatVision Backend is alive!", 200
29
-
30
- # --- 1. CONFIGURATION & DATABASE ---
31
- CHECKPOINT_PATH = "./models/SatVision_Model.ckpt"
32
- DOWNLOADS_DIR = "server_downloads"
33
- os.makedirs(DOWNLOADS_DIR, exist_ok=True)
34
-
35
- DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
36
-
37
- MONGO_URI = os.environ.get("MONGO_URI")
38
- try:
39
- mongo_client = MongoClient(MONGO_URI, serverSelectionTimeoutMS=5000)
40
- db = mongo_client.satvision_db
41
- generations_col = db.generations
42
- print("βœ… MongoDB Connected Successfully!", flush=True)
43
- except Exception as e:
44
- print(f"❌ MongoDB Connection Failed: {e}", flush=True)
45
- db, generations_col = None, None
46
-
47
- # --- 2. GOOGLE EARTH ENGINE SETUP ---
48
- print("🌍 Initializing Earth Engine...", flush=True)
49
- try:
50
- credentials_json = os.environ.get("EE_CREDENTIALS")
51
- if credentials_json:
52
- key_dict = json.loads(credentials_json)
53
- from google.oauth2 import service_account
54
- credentials = service_account.Credentials.from_service_account_info(
55
- key_dict,
56
- scopes=['https://www.googleapis.com/auth/earthengine']
57
- )
58
- ee.Initialize(credentials, project='gen-lang-client-0114614261')
59
- print("βœ… Earth Engine Linked via Service Account.", flush=True)
60
- else:
61
- ee.Initialize(project='gen-lang-client-0114614261')
62
- print("βœ… Earth Engine Linked via Local Auth.", flush=True)
63
- except Exception as e:
64
- print(f"❌ Earth Engine Init Failed: {e}", flush=True)
65
-
66
- # --- 3. MODEL DEFINITION ---
67
- class DynamicEvalModel(pl.LightningModule):
68
- def __init__(self, **kwargs):
69
- super().__init__()
70
- self.save_hyperparameters()
71
- self.model_params = kwargs.get('model_params', {'model_type': 'unetplusplus', 'encoder_name': 'resnet34', 'in_channels': 6, 'out_classes': 2})
72
- self._build_model()
73
-
74
- def _build_model(self):
75
- self.model = smp.UnetPlusPlus(
76
- encoder_name=self.model_params.get('encoder_name', 'resnet34'),
77
- encoder_weights=None,
78
- in_channels=self.model_params.get('in_channels', 6),
79
- classes=self.model_params.get('out_classes', 2),
80
- activation=None
81
- )
82
-
83
- def forward(self, x):
84
- return self.model(x)
85
-
86
- print("⏳ Loading model...", flush=True)
87
- try:
88
- model = DynamicEvalModel.load_from_checkpoint(CHECKPOINT_PATH, strict=False)
89
- model.to(DEVICE)
90
- model.eval()
91
- print("βœ… Model Loaded Successfully!", flush=True)
92
- except Exception as e:
93
- print(f"❌ Model Load Failed: {e}", flush=True)
94
- model = None
95
-
96
- # --- 4. GEE FETCHING & UTILS ---
97
-
98
- def download_gee_tile(image, n, s, e, w, prefix, r, c, scale=10):
99
- region = ee.Geometry.BBox(w, s, e, n)
100
- url = image.getDownloadURL({'crs': 'EPSG:4326', 'region': region, 'scale': scale, 'format': 'GEO_TIFF'})
101
- resp = requests.get(url, timeout=120)
102
- if resp.status_code != 200:
103
- print(f"❌ Tile {r}-{c} failed.", flush=True)
104
- return None
105
-
106
- path = os.path.join(DOWNLOADS_DIR, f"{prefix}_tile_{r}_{c}.tiff")
107
- with open(path, 'wb') as f:
108
- f.write(resp.content)
109
- return path
110
-
111
- def fetch_optical_composite(coords, time_interval, filename_prefix):
112
- try:
113
- region = ee.Geometry.BBox(coords['west'], coords['south'], coords['east'], coords['north'])
114
- start_date, end_date = time_interval
115
-
116
- col = (ee.ImageCollection('COPERNICUS/S2_SR_HARMONIZED')
117
- .filterBounds(region)
118
- .filterDate(start_date, end_date)
119
- .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 35)))
120
-
121
- if col.size().getInfo() == 0:
122
- col = (ee.ImageCollection('COPERNICUS/S2_HARMONIZED')
123
- .filterBounds(region)
124
- .filterDate(start_date, end_date)
125
- .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 35)))
126
-
127
- if col.size().getInfo() == 0:
128
- return None, None, None, 100.0
129
-
130
- def maskS2clouds(image):
131
- qa = image.select('QA60')
132
- cloudBitMask = 1 << 10
133
- cirrusBitMask = 1 << 11
134
- mask = qa.bitwiseAnd(cloudBitMask).eq(0).And(qa.bitwiseAnd(cirrusBitMask).eq(0))
135
- return image.updateMask(mask)
136
-
137
- mosaic = col.map(maskS2clouds).median()
138
- s2_bands = ['B2', 'B3', 'B4', 'B8', 'B11', 'B12']
139
- final_image = mosaic.select(s2_bands).toInt16().unmask(0)
140
-
141
- valid_mask = final_image.select('B4').gt(0)
142
- stats = valid_mask.reduceRegion(
143
- reducer=ee.Reducer.mean(),
144
- geometry=region,
145
- scale=100,
146
- maxPixels=1e9
147
- )
148
- clear_fraction = stats.get('B4').getInfo()
149
-
150
- if clear_fraction is None: clear_fraction = 0
151
- cloud_percent = (1.0 - clear_fraction) * 100.0
152
-
153
- if cloud_percent > 85.0:
154
- return None, None, None, cloud_percent
155
-
156
- actual_date = end_date
157
- source_type = "Sentinel-2 Optical (Cloud-Masked Composite)"
158
-
159
- n, s, e, w = coords['north'], coords['south'], coords['east'], coords['west']
160
- is_large = max(n - s, e - w) > 0.5
161
- MAX_DEG_STEP = 0.50 if is_large else 0.08
162
- current_scale = 50 if is_large else 10
163
- rows = math.ceil((n - s) / MAX_DEG_STEP)
164
- cols = math.ceil((e - w) / MAX_DEG_STEP)
165
- lat_steps = np.linspace(n, s, rows + 1)
166
- lon_steps = np.linspace(w, e, cols + 1)
167
-
168
- tile_arrays = []
169
- for r in range(rows):
170
- row_list = []
171
- for c in range(cols):
172
- tile_n, tile_s = lat_steps[r], lat_steps[r+1]
173
- tile_w, tile_e = lon_steps[c], lon_steps[c+1]
174
- tile_path = download_gee_tile(final_image, tile_n, tile_s, tile_e, tile_w, filename_prefix, r, c, scale=current_scale)
175
- if not tile_path: return None, None, None, 100.0
176
- img = tiff.imread(tile_path)
177
- row_list.append(img)
178
- os.remove(tile_path)
179
-
180
- min_h = min(tile.shape[0] for tile in row_list)
181
- row_list = [tile[:min_h, :, :] for tile in row_list]
182
- row_stitched = np.concatenate(row_list, axis=1)
183
- tile_arrays.append(row_stitched)
184
-
185
- min_w = min(row.shape[1] for row in tile_arrays)
186
- tile_arrays = [row[:, :min_w, :] for row in tile_arrays]
187
- final_stitched = np.concatenate(tile_arrays, axis=0)
188
-
189
- save_path = os.path.join(DOWNLOADS_DIR, f"{filename_prefix}.tiff")
190
- tiff.imwrite(save_path, final_stitched)
191
-
192
- return save_path, actual_date, source_type, cloud_percent
193
-
194
- except Exception as e:
195
- print(f"❌ GEE Optical Error {filename_prefix}: {e}", flush=True)
196
- return None, None, None, 100.0
197
-
198
- def fetch_sar_data(coords, time_interval, filename_prefix, is_baseline=False):
199
- try:
200
- region = ee.Geometry.BBox(coords['west'], coords['south'], coords['east'], coords['north'])
201
- start_date, end_date = time_interval
202
-
203
- s1_col = (ee.ImageCollection('COPERNICUS/S1_GRD')
204
- .filterBounds(region)
205
- .filterDate(start_date, end_date)
206
- .filter(ee.Filter.listContains('transmitterReceiverPolarisation', 'VV'))
207
- .filter(ee.Filter.eq('instrumentMode', 'IW')))
208
-
209
- if s1_col.size().getInfo() == 0: return None, None, None
210
-
211
- if is_baseline:
212
- s1_image = s1_col.median().select('VV').unmask(-99)
213
- else:
214
- s1_image = s1_col.mosaic().select('VV').unmask(-99)
215
-
216
- final_image = s1_image.toFloat()
217
-
218
- MAX_DEG_STEP = 0.08
219
- n, s, e, w = coords['north'], coords['south'], coords['east'], coords['west']
220
- rows = math.ceil((n - s) / MAX_DEG_STEP)
221
- cols = math.ceil((e - w) / MAX_DEG_STEP)
222
- lat_steps = np.linspace(n, s, rows + 1)
223
- lon_steps = np.linspace(w, e, cols + 1)
224
-
225
- tile_arrays = []
226
- for r in range(rows):
227
- row_list = []
228
- for c in range(cols):
229
- tile_n, tile_s = lat_steps[r], lat_steps[r+1]
230
- tile_w, tile_e = lon_steps[c], lon_steps[c+1]
231
- tile_path = download_gee_tile(final_image, tile_n, tile_s, tile_e, tile_w, filename_prefix, r, c, scale=current_scale)
232
- if not tile_path: return None, None, None
233
- img = tiff.imread(tile_path)
234
- row_list.append(img)
235
- os.remove(tile_path)
236
-
237
- min_h = min(tile.shape[0] for tile in row_list)
238
- row_list = [tile[:min_h, ...] for tile in row_list]
239
- row_stitched = np.concatenate(row_list, axis=1)
240
- tile_arrays.append(row_stitched)
241
-
242
- min_w = min(row.shape[1] for row in tile_arrays)
243
- tile_arrays = [row[:, :min_w, ...] for row in tile_arrays]
244
- final_stitched = np.concatenate(tile_arrays, axis=0)
245
-
246
- actual_date = end_date
247
- source_type = "Sentinel-1 SAR (VV Backscatter)" if not is_baseline else "Sentinel-1 SAR (Dry-Season Baseline)"
248
-
249
- save_path = os.path.join(DOWNLOADS_DIR, f"{filename_prefix}_sar.tiff")
250
- tiff.imwrite(save_path, final_stitched)
251
-
252
- return save_path, actual_date, source_type
253
-
254
- except Exception as e:
255
- print(f"❌ SAR Error {filename_prefix}: {e}", flush=True)
256
- return None, None, None
257
-
258
- def fetch_dynamic_world_baseline(coords, time_interval, filename_prefix):
259
- try:
260
- region = ee.Geometry.BBox(coords['west'], coords['south'], coords['east'], coords['north'])
261
- start_date, end_date = time_interval
262
-
263
- dw_col = (ee.ImageCollection('GOOGLE/DYNAMICWORLD/V1')
264
- .filterBounds(region)
265
- .filterDate(start_date, end_date))
266
-
267
- if dw_col.size().getInfo() == 0: return None, None
268
-
269
- final_image = dw_col.select('label').mode().eq(0).unmask(0).multiply(255).toByte()
270
-
271
- MAX_DEG_STEP = 0.08
272
- n, s, e, w = coords['north'], coords['south'], coords['east'], coords['west']
273
- rows = math.ceil((n - s) / MAX_DEG_STEP)
274
- cols = math.ceil((e - w) / MAX_DEG_STEP)
275
- lat_steps = np.linspace(n, s, rows + 1)
276
- lon_steps = np.linspace(w, e, cols + 1)
277
-
278
- tile_arrays = []
279
- for r in range(rows):
280
- row_list = []
281
- for c in range(cols):
282
- tile_n, tile_s = lat_steps[r], lat_steps[r+1]
283
- tile_w, tile_e = lon_steps[c], lon_steps[c+1]
284
- tile_path = download_gee_tile(final_image, tile_n, tile_s, tile_e, tile_w, filename_prefix, r, c)
285
- if not tile_path: return None, None
286
- img = tiff.imread(tile_path)
287
- row_list.append(img)
288
- os.remove(tile_path)
289
-
290
- min_h = min(tile.shape[0] for tile in row_list)
291
- row_list = [tile[:min_h, ...] for tile in row_list]
292
- row_stitched = np.concatenate(row_list, axis=1)
293
- tile_arrays.append(row_stitched)
294
-
295
- min_w = min(row.shape[1] for row in tile_arrays)
296
- tile_arrays = [row[:, :min_w, ...] for row in tile_arrays]
297
- final_stitched = np.concatenate(tile_arrays, axis=0)
298
-
299
- save_path = os.path.join(DOWNLOADS_DIR, f"{filename_prefix}.tiff")
300
- tiff.imwrite(save_path, final_stitched)
301
-
302
- return save_path, "Dynamic Dry-Season Anchor (April-May)"
303
-
304
- except Exception as e:
305
- print(f"❌ Dynamic World Error {filename_prefix}: {e}", flush=True)
306
- return None, None
307
-
308
- # --- 5. IMAGE PROCESSING UTILS ---
309
-
310
- def predict_water_mask(tiff_path):
311
- if not model or not os.path.exists(tiff_path): return None
312
-
313
- image = tiff.imread(tiff_path).astype(np.float32)
314
-
315
- if image.ndim == 3 and image.shape[2] == 6:
316
- image_tensor = np.transpose(image, (2, 0, 1))
317
- else:
318
- image_tensor = image.copy()
319
-
320
- scaled_tensor = np.clip(image_tensor / 10000.0, 0, 1)
321
-
322
- tile_size, overlap = 256, 0.5
323
- stride = int(tile_size * (1 - overlap))
324
- c, h, w = scaled_tensor.shape
325
-
326
- pad_h = (tile_size - (h % stride)) % stride + (tile_size - stride) if h % stride != 0 else 0
327
- pad_w = (tile_size - (w % stride)) % stride + (tile_size - stride) if w % stride != 0 else 0
328
- image_padded = np.pad(scaled_tensor, ((0,0), (0, pad_h), (0, pad_w)), mode='reflect')
329
-
330
- padded_h, padded_w = image_padded.shape[1], image_padded.shape[2]
331
-
332
- prob_map_cloud = np.zeros((padded_h, padded_w), dtype=np.float32)
333
- prob_map_water = np.zeros((padded_h, padded_w), dtype=np.float32)
334
- count_map = np.zeros((padded_h, padded_w), dtype=np.float32)
335
-
336
- y_steps = list(range(0, padded_h - tile_size + 1, stride))
337
- x_steps = list(range(0, padded_w - tile_size + 1, stride))
338
-
339
- window_1d = np.hanning(tile_size)
340
- window_2d = np.outer(window_1d, window_1d).astype(np.float32)
341
-
342
- batch_size = 16 # Process 16 tiles simultaneously (adjust to 8 if you hit GPU memory limits)
343
- tiles_batch = []
344
- coords_batch = []
345
-
346
- with torch.no_grad():
347
- for y in y_steps:
348
- for x in x_steps:
349
- # 1. Extract the numpy tile and store it
350
- tile = image_padded[:, y:y+tile_size, x:x+tile_size]
351
- tiles_batch.append(tile)
352
- coords_batch.append((x, y))
353
-
354
- # 2. If we hit the batch size, OR we are on the very last patch, run inference
355
- if len(tiles_batch) == batch_size or (y == y_steps[-1] and x == x_steps[-1]):
356
- # Stack list of arrays into a single (B, C, H, W) tensor
357
- batch_tensor = torch.from_numpy(np.stack(tiles_batch)).to(DEVICE)
358
-
359
- # Forward pass
360
- logits = model(batch_tensor)
361
- probs = torch.sigmoid(logits).cpu().numpy()
362
-
363
- # 3. Reassemble the probabilities back into the full-size maps
364
- for i, (bx, by) in enumerate(coords_batch):
365
- prob_cloud = probs[i, 0, :, :] * window_2d
366
- prob_water = probs[i, 1, :, :] * window_2d
367
-
368
- prob_map_cloud[by:by+tile_size, bx:bx+tile_size] += prob_cloud
369
- prob_map_water[by:by+tile_size, bx:bx+tile_size] += prob_water
370
- count_map[by:by+tile_size, bx:bx+tile_size] += window_2d
371
-
372
- # 4. Clear lists for the next batch
373
- tiles_batch = []
374
- coords_batch = []
375
-
376
- final_cloud_prob = (prob_map_cloud / (count_map + 1e-7))[:h, :w]
377
- final_water_prob = (prob_map_water / (count_map + 1e-7))[:h, :w]
378
-
379
- b2 = scaled_tensor[0, :h, :w]
380
- b3 = scaled_tensor[1, :h, :w]
381
- b4 = scaled_tensor[2, :h, :w]
382
- b8 = scaled_tensor[3, :h, :w]
383
-
384
- brightness = np.sqrt(b2**2 + b3**2 + b4**2)
385
- ndwi = (b3 - b8) / (b3 + b8 + 1e-5)
386
-
387
- base_cloud_mask = ((final_cloud_prob > 0.5) & (brightness > 0.35)).astype(np.uint8)
388
-
389
- cloud_buffer_kernel = np.ones((10, 10), np.uint8)
390
- cloud_and_shadow_mask = cv2.dilate(base_cloud_mask, cloud_buffer_kernel, iterations=1)
391
-
392
- raw_water_mask = (final_water_prob > 0.35) | (ndwi > 0.15)
393
-
394
- water_mask = raw_water_mask & (cloud_and_shadow_mask == 0) & (ndwi > -0.1)
395
- clean_mask = water_mask.astype(np.uint8)
396
-
397
- close_kernel = np.ones((3, 3), np.uint8)
398
- clean_mask = cv2.morphologyEx(clean_mask, cv2.MORPH_CLOSE, close_kernel)
399
-
400
- open_kernel = np.ones((2, 2), np.uint8)
401
- clean_mask = cv2.morphologyEx(clean_mask, cv2.MORPH_OPEN, open_kernel)
402
-
403
- contours, _ = cv2.findContours(clean_mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
404
-
405
- min_area = 50
406
- for cnt in contours:
407
- area = cv2.contourArea(cnt)
408
- if area < min_area:
409
- cv2.drawContours(clean_mask, [cnt], -1, 0, -1)
410
-
411
- return clean_mask * 255
412
-
413
- def process_sar_image(tiff_path, preview_filename):
414
- if not os.path.exists(tiff_path): return None, None
415
-
416
- sar_db = tiff.imread(tiff_path).astype(np.float32)
417
-
418
- water_mask = ((sar_db < -16) & (sar_db > -90)).astype(np.uint8) * 255
419
-
420
- close_kernel = np.ones((3, 3), np.uint8)
421
- water_mask = cv2.morphologyEx(water_mask, cv2.MORPH_CLOSE, close_kernel)
422
-
423
- visual = np.clip((sar_db + 25) / 25.0, 0, 1) * 255
424
- visual = visual.astype(np.uint8)
425
-
426
- h, w = visual.shape
427
- rgba = np.zeros((h, w, 4), dtype=np.uint8)
428
- rgba[:, :, 0] = visual
429
- rgba[:, :, 1] = visual
430
- rgba[:, :, 2] = visual
431
-
432
- rgba[:, :, 3] = np.where(sar_db < -90, 0, 255)
433
-
434
- save_path = os.path.join(DOWNLOADS_DIR, preview_filename)
435
- cv2.imwrite(save_path, rgba)
436
-
437
- return water_mask, preview_filename
438
-
439
- def create_rgb_preview(tiff_path, filename):
440
- if not os.path.exists(tiff_path): return None
441
- image = tiff.imread(tiff_path).astype(np.float32)
442
-
443
- if image.ndim == 2:
444
- rgb = np.zeros((image.shape[0], image.shape[1], 3), dtype=np.uint8)
445
- rgb[image == 255] = [255, 0, 0]
446
- save_path = os.path.join(DOWNLOADS_DIR, filename)
447
- cv2.imwrite(save_path, rgb)
448
- return filename
449
-
450
- if image.shape[0] == 6: image = np.transpose(image, (1, 2, 0))
451
- if np.max(image) > 1.0: image = image / 10000.0
452
-
453
- rgb = image[:, :, [0, 1, 2]]
454
- rgb = np.clip(rgb * 3.0, 0, 1) * 255
455
- rgb = rgb.astype(np.uint8)
456
-
457
- gaussian = cv2.GaussianBlur(rgb, (0, 0), 2.0)
458
- sharpened = cv2.addWeighted(rgb, 1.5, gaussian, -0.5, 0)
459
-
460
- save_path = os.path.join(DOWNLOADS_DIR, filename)
461
- cv2.imwrite(save_path, sharpened)
462
- return filename
463
-
464
- def create_overlay_png(mask, color, filename):
465
- if mask is None: return None
466
- h, w = mask.shape
467
- rgba = np.zeros((h, w, 4), dtype=np.uint8)
468
- b, g, r = color
469
- rgba[mask == 255, 0] = b
470
- rgba[mask == 255, 1] = g
471
- rgba[mask == 255, 2] = r
472
- rgba[mask == 255, 3] = 180
473
- save_path = os.path.join(DOWNLOADS_DIR, filename)
474
- cv2.imwrite(save_path, rgba)
475
- return filename
476
-
477
- # --- 6. AI REPORTING ---
478
-
479
- def generate_flood_report(stats, coords, dates, image_path):
480
- try:
481
- gemini_api_key = os.environ.get("GEMINI_API_KEY")
482
- if not gemini_api_key:
483
- return "Error: AI Report unavailable. GEMINI_API_KEY missing in environment."
484
-
485
- if not os.path.exists(image_path):
486
- return "Error: AI Report unavailable. Flood image overlay not found."
487
-
488
- client = genai.Client(api_key=gemini_api_key)
489
-
490
- baseline_date = dates.get('baseline', 'Unknown Baseline')
491
- latest_date = dates.get('latest', 'Unknown Date')
492
- base_sq = stats.get('baseline_sq_km', 0.0)
493
- curr_sq = stats.get('current_sq_km', 0.0)
494
- flood_sq = stats.get('flood_sq_km', 0.0)
495
-
496
- with Image.open(image_path) as img:
497
- prompt = f"""
498
- You are a Lead Geospatial Intelligence Analyst generating an executive-level Flood Damage and Loss Assessment (DaLA) report.
499
- I have provided a satellite image overlay. The red pixels definitively indicate algorithmic detections of newly accumulated floodwater.
500
-
501
- CONTEXTUAL TELEMETRY:
502
- - Bounding Box Coordinates: {coords}
503
- - Baseline Imagery Date: {baseline_date}
504
- - Post-Flood Capture Date: {latest_date}
505
- - Pre-flood Water Extent: {base_sq:.2f} sq km
506
- - Current Total Water Extent: {curr_sq:.2f} sq km
507
- - Net New Inundation Area: {flood_sq:.2f} sq km
508
-
509
- YOUR DIRECTIVES:
510
- 1. Be Authoritative: Eliminate all speculative language (e.g., do not use "likely," "suggests," or "possibly"). Make definitive statements based on the visual evidence and telemetry provided. If a factor is indistinguishable, state that it cannot be determined at this resolution rather than guessing.
511
- 2. Be Analytical: Do not just read the numbers back. Calculate the percentage increase in water extent and explain what that scale of increase means for the terrain visible in the image.
512
- 3. Be Plain Text: You must output strictly in plain text. DO NOT use markdown formatting, asterisks, hashes, bolding, or special characters. Use standard ALL CAPS for your headers.
513
-
514
- REQUIRED REPORT STRUCTURE:
515
- EXECUTIVE SUMMARY
516
- Provide a bottom-line-up-front (BLUF) statement summarizing the severity and scale of the inundation event.
517
-
518
- GEOSPATIAL & CLIMATOLOGICAL CONTEXT
519
- Identify the specific region in Pakistan based on the coordinates. Correlate the capture dates with known seasonal weather patterns (e.g., monsoon cycles or glacial melt) to establish the primary driver of the event.
520
-
521
- INUNDATION METRICS & MORPHOLOGY
522
- Analyze the quantitative data. Describe the morphological spread of the floodwaters (e.g., riverine overspill, pooling in low-lying plains, or flash flood patterns) based strictly on the shape and distribution of the red overlay.
523
-
524
- INFRASTRUCTURE & SOCIO-ECONOMIC IMPACT
525
- Examine the base imagery beneath the flood mask. Definitively state the types of terrain affected (e.g., cultivated agricultural plots, dense urban grids, rural transit corridors). Outline the immediate operational impacts on these specific sectors.
526
-
527
- MODEL CONFIDENCE & FALSE POSITIVE ANALYSIS
528
- Critically evaluate the segmentation mask. Identify specific areas in the image where the model may have misclassified topographical shadows, dense cloud shadows, or highly reflective urban surfaces as water.
529
-
530
- STRATEGIC DIRECTIVES
531
- Provide 3 highly specific, actionable recommendations for on-the-ground disaster response agencies based on the exact terrain and impact identified above.
532
- """
533
-
534
- config = types.GenerateContentConfig(
535
- temperature=0.1,
536
- top_p=0.9,
537
- )
538
-
539
- response = client.models.generate_content(
540
- model='gemini-2.5-flash',
541
- contents=[img, prompt],
542
- config=config
543
- )
544
-
545
- if response.text:
546
- return response.text.strip()
547
- else:
548
- return "Error: Model returned an empty response."
549
-
550
- except Exception as e:
551
- import traceback
552
- print(f"❌ AI Report Error: {e}\n{traceback.format_exc()}", flush=True)
553
- return "Error: AI Report generation failed due to an internal system error."
554
-
555
-
556
- # --- 7. MAIN ENDPOINTS ---
557
-
558
- @app.route('/api/detect_stream', methods=['POST'])
559
- def detect_stream():
560
- data = request.json
561
- coords = data.get('bbox')
562
- apply_buffer = data.get('apply_buffer', True)
563
-
564
- target_date_str = data.get('target_date')
565
- if target_date_str:
566
- target_date = datetime.strptime(target_date_str, '%Y-%m-%d')
567
- else:
568
- target_date = datetime.now()
569
-
570
- session_id = str(uuid.uuid4())
571
- session_logs = []
572
-
573
- def generate_process():
574
- def emit_log(progress, message):
575
- session_logs.append({"progress": progress, "log": message, "timestamp": datetime.now().isoformat()})
576
- return json.dumps({"progress": progress, "log": message}) + "\n"
577
-
578
- yield emit_log(5, "🌍 Locking Coordinates & Validating Dates...")
579
-
580
- start_latest = target_date - timedelta(days=15)
581
- end_latest = target_date + timedelta(days=5)
582
- interval_latest = (start_latest.strftime('%Y-%m-%d'), end_latest.strftime('%Y-%m-%d'))
583
-
584
- target_year = target_date.year
585
- baseline_year = target_year - 1 if target_date.month < 6 else target_year
586
- interval_baseline = (f"{baseline_year}-04-01", f"{baseline_year}-05-31")
587
-
588
- yield emit_log(15, f"πŸ“‘ Building Optical Composite ({interval_latest[0]} to {interval_latest[1]})...")
589
-
590
- path_latest, date_latest, src_latest, cloud_percent = fetch_optical_composite(coords, interval_latest, f"{session_id}_latest")
591
- if cloud_percent is None: cloud_percent = 100
592
-
593
- if path_latest and cloud_percent <= 85.0:
594
- yield emit_log(30, f"🧠 Optical Rescued ({cloud_percent:.1f}% Dead Pixels). Running U-Net++...")
595
- mask_latest = predict_water_mask(path_latest)
596
- rgb_latest = create_rgb_preview(path_latest, f"{session_id}_rgb_latest.jpg")
597
-
598
- yield emit_log(50, "πŸ“‘ Fetching Dry-Season Baseline (Optical)...")
599
- path_pre, src_pre = fetch_dynamic_world_baseline(coords, interval_baseline, f"{session_id}_previous")
600
- date_pre = f"{interval_baseline[0]} to {interval_baseline[1]}"
601
-
602
- if path_pre:
603
- mask_pre = tiff.imread(path_pre).astype(np.uint8)
604
- rgb_pre = create_rgb_preview(path_pre, f"{session_id}_rgb_previous.jpg")
605
- else:
606
- mask_pre = np.zeros_like(mask_latest)
607
- rgb_pre = None
608
-
609
- else:
610
- yield emit_log(30, f"β›ˆοΈ Optical Failed ({cloud_percent:.1f}% Dead Pixels). Switching to SAR Payload...")
611
- path_latest, date_latest, src_latest = fetch_sar_data(coords, interval_latest, f"{session_id}_latest", is_baseline=False)
612
-
613
- if not path_latest:
614
- yield json.dumps({"error": "Critical Error: Failed to acquire both Optical and SAR fallback data."}) + "\n"
615
- return
616
-
617
- mask_latest, rgb_latest = process_sar_image(path_latest, f"{session_id}_rgb_latest_sar.png")
618
-
619
- yield emit_log(50, "πŸ“‘ Fetching Homogeneous SAR Baseline...")
620
- path_pre, date_pre_raw, src_pre = fetch_sar_data(coords, interval_baseline, f"{session_id}_previous", is_baseline=True)
621
- date_pre = f"{interval_baseline[0]} to {interval_baseline[1]}"
622
-
623
- if path_pre:
624
- mask_pre, rgb_pre = process_sar_image(path_pre, f"{session_id}_rgb_previous_sar.png")
625
- else:
626
- mask_pre = np.zeros_like(mask_latest)
627
- rgb_pre = None
628
-
629
- yield emit_log(70, "🌊 Aligning Baseline Maps...")
630
- if mask_latest.shape != mask_pre.shape:
631
- mask_pre = cv2.resize(mask_pre, (mask_latest.shape[1], mask_latest.shape[0]), interpolation=cv2.INTER_NEAREST)
632
-
633
- yield emit_log(80, "🌊 Generating Flood Difference...")
634
- if apply_buffer:
635
- buffer_kernel = np.ones((5, 5), np.uint8)
636
- mask_pre_buffered = cv2.dilate(mask_pre, buffer_kernel, iterations=1)
637
- flood_mask = cv2.bitwise_and(mask_latest, cv2.bitwise_not(mask_pre_buffered))
638
- else:
639
- flood_mask = cv2.bitwise_and(mask_latest, cv2.bitwise_not(mask_pre))
640
-
641
- base_url = request.host_url.rstrip('/')
642
-
643
- yield emit_log(85, "🎨 Generating Visual Overlays...")
644
- url_latest = create_overlay_png(mask_latest, [255, 0, 0], f"{session_id}_overlay_latest.png")
645
- url_pre = create_overlay_png(mask_pre, [255, 255, 0], f"{session_id}_overlay_previous.png")
646
- url_flood = create_overlay_png(flood_mask, [0, 0, 255], f"{session_id}_overlay_flood.png")
647
-
648
- yield emit_log(90, "πŸ“Š Calculating Spatial Metrics...")
649
- n, s, e, w = coords['north'], coords['south'], coords['east'], coords['west']
650
- current_scale = 50 if max(n - s, e - w) > 0.5 else 10
651
- pixel_to_sq_km = (current_scale ** 2) / 1_000_000
652
- stats = {
653
- "baseline_sq_km": np.count_nonzero(mask_pre == 255) * pixel_to_sq_km,
654
- "current_sq_km": np.count_nonzero(mask_latest == 255) * pixel_to_sq_km,
655
- "flood_sq_km": np.count_nonzero(flood_mask == 255) * pixel_to_sq_km,
656
- }
657
- dates = {"baseline": date_pre, "latest": date_latest}
658
-
659
- yield emit_log(95, "πŸ€– Generating AI Assessment Report...")
660
- flood_image_path = os.path.join(DOWNLOADS_DIR, f"{session_id}_overlay_flood.png")
661
- ai_report_text = generate_flood_report(stats, coords, dates, flood_image_path)
662
-
663
- yield emit_log(98, "πŸ“„ Compiling PDF Report...")
664
- report_filename = f"flood_assessment_report_{session_id}.pdf"
665
- report_path = os.path.join(DOWNLOADS_DIR, report_filename)
666
-
667
- pdf_img_path = os.path.join(DOWNLOADS_DIR, f"pdf_img_{session_id}.jpg")
668
-
669
- if rgb_latest:
670
- base_img = cv2.imread(os.path.join(DOWNLOADS_DIR, rgb_latest))
671
- if base_img is not None and base_img.shape[2] == 4:
672
- base_img = cv2.cvtColor(base_img, cv2.COLOR_BGRA2BGR)
673
- else:
674
- base_img = None
675
-
676
- if base_img is not None:
677
- pdf_img = base_img.copy()
678
-
679
- if flood_mask.shape != pdf_img.shape[:2]:
680
- safe_flood_mask = cv2.resize(flood_mask, (pdf_img.shape[1], pdf_img.shape[0]), interpolation=cv2.INTER_NEAREST)
681
- else:
682
- safe_flood_mask = flood_mask
683
-
684
- pdf_img[safe_flood_mask == 255] = (0, 0, 255)
685
- cv2.imwrite(pdf_img_path, pdf_img)
686
- else:
687
- pdf_img_path = flood_image_path
688
-
689
- pdf = FPDF()
690
- pdf.add_page()
691
- pdf.set_font("helvetica", "B", 18)
692
- pdf.cell(0, 15, "SatVision Flood Assessment Report", align="C", new_x=XPos.LMARGIN, new_y=YPos.NEXT)
693
- pdf.ln(5)
694
- pdf.image(pdf_img_path, x=15, w=180)
695
- pdf.ln(10)
696
- pdf.set_font("helvetica", size=11)
697
- pdf.multi_cell(0, 6, text=ai_report_text, markdown=True)
698
- pdf.output(report_path)
699
-
700
- yield emit_log(100, "βœ… Detection Complete!")
701
-
702
- if generations_col is not None:
703
- generation_record = {
704
- "session_id": session_id,
705
- "timestamp": datetime.now(),
706
- "coordinates": coords,
707
- "target_date": target_date_str,
708
- "metadata": {
709
- "latest_source": src_latest,
710
- "baseline_source": src_pre,
711
- "latest_date": date_latest,
712
- "baseline_date": date_pre
713
- },
714
- "metrics": stats,
715
- "ai_report": ai_report_text,
716
- "logs": session_logs,
717
- "feedback": None
718
- }
719
- try:
720
- generations_col.insert_one(generation_record)
721
- except Exception as db_err:
722
- print(f"❌ DB Insert Failed: {db_err}")
723
-
724
- yield json.dumps({
725
- "progress": 100,
726
- "session_id": session_id,
727
- "result": {
728
- "latest": f"{base_url}/mask/{url_latest}",
729
- "previous": f"{base_url}/mask/{url_pre}",
730
- "flood": f"{base_url}/mask/{url_flood}"
731
- },
732
- "meta": {
733
- "latest_date": date_latest,
734
- "previous_date": date_pre,
735
- "latest_source": src_latest,
736
- "previous_source": src_pre,
737
- "latest_rgb": f"{base_url}/mask/{rgb_latest}" if rgb_latest else None,
738
- "previous_rgb": f"{base_url}/mask/{rgb_pre}" if rgb_pre else None
739
- },
740
- "report": {
741
- "text": ai_report_text,
742
- "download_url": f"{base_url}/mask/{report_filename}",
743
- "metrics": stats
744
- }
745
- }) + "\n"
746
-
747
- return Response(stream_with_context(generate_process()), mimetype='application/json')
748
-
749
- @app.route('/api/feedback', methods=['POST'])
750
- def submit_feedback():
751
- if generations_col is None:
752
- return {"error": "Database unavailable"}, 503
753
-
754
- data = request.json
755
- session_id = data.get('session_id')
756
- feedback_text = data.get('feedback_text')
757
- rating = data.get('rating')
758
-
759
- if not session_id or not feedback_text:
760
- return {"error": "Missing session_id or feedback_text"}, 400
761
-
762
- try:
763
- result = generations_col.update_one(
764
- {"session_id": session_id},
765
- {"$set": {
766
- "feedback": {
767
- "text": feedback_text,
768
- "rating": rating,
769
- "submitted_at": datetime.now()
770
- }
771
- }}
772
- )
773
-
774
- if result.matched_count == 0:
775
- return {"error": "Session ID not found"}, 404
776
-
777
- return {"message": "Feedback recorded successfully"}, 200
778
-
779
- except Exception as e:
780
- print(f"❌ Feedback Update Error: {e}", flush=True)
781
- return {"error": "Internal database error"}, 500
782
-
783
- @app.route('/mask/<filename>')
784
- def serve_mask(filename):
785
- is_pdf = filename.endswith('.pdf')
786
- return send_from_directory(DOWNLOADS_DIR, filename, as_attachment=is_pdf)
787
-
788
- if __name__ == '__main__':
789
- print("πŸš€ Flood Backend Running...", flush=True)
790
- app.run(host="0.0.0.0", port=int(os.environ.get("FLASK_PORT", os.environ.get("PORT", 5000))), threaded=True)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
src/app.py CHANGED
@@ -20,6 +20,9 @@ def create_app():
20
  """
21
  app = Flask(__name__)
22
 
 
 
 
23
  # ========================================================================
24
  # CONFIGURATION
25
  # ========================================================================
@@ -36,7 +39,7 @@ def create_app():
36
  try:
37
  gee_service.initialize_earth_engine()
38
  except Exception as e:
39
- print(f"⚠️ Earth Engine initialization warning: {e}", flush=True)
40
 
41
  # ========================================================================
42
  # INITIALIZATION: MONGODB
 
20
  """
21
  app = Flask(__name__)
22
 
23
+ # Run dynamic environment sanity validation
24
+ config.validate_environment()
25
+
26
  # ========================================================================
27
  # CONFIGURATION
28
  # ========================================================================
 
39
  try:
40
  gee_service.initialize_earth_engine()
41
  except Exception as e:
42
+ config.logger.warning(f"⚠️ Earth Engine initialization warning: {e}")
43
 
44
  # ========================================================================
45
  # INITIALIZATION: MONGODB
src/config.py CHANGED
@@ -1,5 +1,14 @@
1
  import os
2
  import torch
 
 
 
 
 
 
 
 
 
3
 
4
  CHECKPOINT_PATH = os.environ.get("CHECKPOINT_PATH", "./models/SatVision_Model.ckpt")
5
  DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
@@ -20,6 +29,23 @@ GEMINI_MODEL_ID = "gemini-2.5-flash"
20
  GEMINI_TEMPERATURE = 0.1
21
  GEMINI_TOP_P = 0.9
22
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
23
  S2_CLOUD_THRESHOLD = 35
24
  OPTICAL_CLEAR_THRESHOLD = 85.0
25
  GEE_SCALE_OPTICAL = 100
 
1
  import os
2
  import torch
3
+ import logging
4
+
5
+ # Configure structured enterprise logging
6
+ logging.basicConfig(
7
+ level=logging.INFO,
8
+ format='%(asctime)s [%(levelname)s] %(name)s: %(message)s',
9
+ handlers=[logging.StreamHandler()]
10
+ )
11
+ logger = logging.getLogger("SatVision")
12
 
13
  CHECKPOINT_PATH = os.environ.get("CHECKPOINT_PATH", "./models/SatVision_Model.ckpt")
14
  DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
29
  GEMINI_TEMPERATURE = 0.1
30
  GEMINI_TOP_P = 0.9
31
 
32
+ def validate_environment():
33
+ """Validates required API keys and credentials at system startup."""
34
+ missing = []
35
+ if not EE_CREDENTIALS_JSON:
36
+ missing.append("EE_CREDENTIALS")
37
+ if not GEMINI_API_KEY:
38
+ missing.append("GEMINI_API_KEY")
39
+
40
+ if missing:
41
+ logger.warning(
42
+ f"⚠️ [Startup Security Audit] Missing required environment variables: {', '.join(missing)}. "
43
+ "Google Earth Engine downloads and Gemini report generation will fail in runtime if not configured."
44
+ )
45
+ else:
46
+ logger.info("βœ… [Startup Security Audit] All core environmental secrets verified successfully.")
47
+
48
+
49
  S2_CLOUD_THRESHOLD = 35
50
  OPTICAL_CLEAR_THRESHOLD = 85.0
51
  GEE_SCALE_OPTICAL = 100
src/database/gee.py CHANGED
@@ -11,7 +11,7 @@ def initialize_earth_engine():
11
  """
12
  Initialize Earth Engine with service account credentials or local auth.
13
  """
14
- print("🌍 Initializing Earth Engine...", flush=True)
15
 
16
  try:
17
  if config.EE_CREDENTIALS_JSON:
@@ -24,11 +24,11 @@ def initialize_earth_engine():
24
  scopes=['https://www.googleapis.com/auth/earthengine']
25
  )
26
  ee.Initialize(credentials, project=config.EE_PROJECT_ID)
27
- print("βœ… Earth Engine Linked via Service Account.", flush=True)
28
  else:
29
  # Use local authentication (cached credentials)
30
  ee.Initialize(project=config.EE_PROJECT_ID)
31
- print("βœ… Earth Engine Linked via Local Auth.", flush=True)
32
  except Exception as e:
33
- print(f"❌ Earth Engine Init Failed: {e}", flush=True)
34
  raise
 
11
  """
12
  Initialize Earth Engine with service account credentials or local auth.
13
  """
14
+ config.logger.info("🌍 Initializing Earth Engine...")
15
 
16
  try:
17
  if config.EE_CREDENTIALS_JSON:
 
24
  scopes=['https://www.googleapis.com/auth/earthengine']
25
  )
26
  ee.Initialize(credentials, project=config.EE_PROJECT_ID)
27
+ config.logger.info("βœ… Earth Engine Linked via Service Account.")
28
  else:
29
  # Use local authentication (cached credentials)
30
  ee.Initialize(project=config.EE_PROJECT_ID)
31
+ config.logger.info("βœ… Earth Engine Linked via Local Auth.")
32
  except Exception as e:
33
+ config.logger.error(f"❌ Earth Engine Init Failed: {e}")
34
  raise
src/database/mongo.py CHANGED
@@ -19,7 +19,7 @@ def initialize_mongodb():
19
  global _mongo_client, _db, _generations_col
20
 
21
  if config.MONGO_URI is None:
22
- print("⚠️ MongoDB URI not configured (MONGO_URI env var missing)", flush=True)
23
  return None, None
24
 
25
  try:
@@ -29,10 +29,10 @@ def initialize_mongodb():
29
 
30
  # Verify connection
31
  _mongo_client.admin.command('ping')
32
- print("βœ… MongoDB Connected Successfully!", flush=True)
33
  return _db, _generations_col
34
  except Exception as e:
35
- print(f"❌ MongoDB Connection Failed: {e}", flush=True)
36
  _mongo_client = None
37
  _db = None
38
  _generations_col = None
@@ -54,4 +54,4 @@ def close_mongodb():
54
  global _mongo_client
55
  if _mongo_client:
56
  _mongo_client.close()
57
- print("βœ… MongoDB connection closed", flush=True)
 
19
  global _mongo_client, _db, _generations_col
20
 
21
  if config.MONGO_URI is None:
22
+ config.logger.warning("⚠️ MongoDB URI not configured (MONGO_URI env var missing)")
23
  return None, None
24
 
25
  try:
 
29
 
30
  # Verify connection
31
  _mongo_client.admin.command('ping')
32
+ config.logger.info("βœ… MongoDB Connected Successfully!")
33
  return _db, _generations_col
34
  except Exception as e:
35
+ config.logger.error(f"❌ MongoDB Connection Failed: {e}")
36
  _mongo_client = None
37
  _db = None
38
  _generations_col = None
 
54
  global _mongo_client
55
  if _mongo_client:
56
  _mongo_client.close()
57
+ config.logger.info("βœ… MongoDB connection closed")
src/services/detection.py CHANGED
@@ -200,7 +200,7 @@ def run_detection_pipeline(coords, target_date, apply_buffer=True, base_url="htt
200
  pdf.multi_cell(0, 6, text=safe_pdf_text)
201
  pdf.output(report_path)
202
  except Exception as pdf_err:
203
- print(f"❌ PDF Engine Failed: {pdf_err}", flush=True)
204
 
205
  yield emit_log(100, "βœ… Detection Pipeline Complete!")
206
 
@@ -219,7 +219,7 @@ def run_detection_pipeline(coords, target_date, apply_buffer=True, base_url="htt
219
  try:
220
  generations_col.insert_one(generation_record)
221
  except Exception as db_err:
222
- print(f"❌ DB Write Blocked: {db_err}", flush=True)
223
 
224
  # Sanitize URLs to point to unique session paths: e.g. base_url/mask/session_id/filename.png
225
  c_lat = f"{session_id}/{os.path.basename(raw_url_latest)}" if raw_url_latest else ""
 
200
  pdf.multi_cell(0, 6, text=safe_pdf_text)
201
  pdf.output(report_path)
202
  except Exception as pdf_err:
203
+ config.logger.error(f"❌ PDF Engine Failed: {pdf_err}")
204
 
205
  yield emit_log(100, "βœ… Detection Pipeline Complete!")
206
 
 
219
  try:
220
  generations_col.insert_one(generation_record)
221
  except Exception as db_err:
222
+ config.logger.error(f"❌ DB Write Blocked: {db_err}")
223
 
224
  # Sanitize URLs to point to unique session paths: e.g. base_url/mask/session_id/filename.png
225
  c_lat = f"{session_id}/{os.path.basename(raw_url_latest)}" if raw_url_latest else ""
src/services/gee_service.py CHANGED
@@ -29,7 +29,7 @@ def download_gee_tile(image, n, s, e, w, prefix, r, c, scale=10):
29
  try:
30
  resp = requests.get(url, timeout=config.GEE_REQUEST_TIMEOUT)
31
  if resp.status_code != 200:
32
- print(f"❌ Tile {r}-{c} failed with status {resp.status_code}.", flush=True)
33
  return None
34
 
35
  path = os.path.join(config.DOWNLOADS_DIR, f"{prefix}_tile_{r}_{c}.tiff")
@@ -37,7 +37,7 @@ def download_gee_tile(image, n, s, e, w, prefix, r, c, scale=10):
37
  f.write(resp.content)
38
  return path
39
  except Exception as e:
40
- print(f"❌ Tile download failed {r}-{c}: {e}", flush=True)
41
  return None
42
 
43
 
@@ -170,7 +170,7 @@ def fetch_optical_composite(coords, time_interval, filename_prefix, scale=10):
170
  return save_path, actual_date, source_type, cloud_percent
171
 
172
  except Exception as e:
173
- print(f"❌ GEE Optical Error {filename_prefix}: {e}", flush=True)
174
  return None, None, None, 100.0
175
 
176
 
@@ -239,7 +239,7 @@ def fetch_sar_data(coords, time_interval, filename_prefix, is_baseline=False, sc
239
  return save_path, actual_date, source_type
240
 
241
  except Exception as e:
242
- print(f"❌ SAR Error {filename_prefix}: {e}", flush=True)
243
  return None, None, None
244
 
245
 
@@ -298,5 +298,5 @@ def fetch_dynamic_world_baseline(coords, time_interval, filename_prefix, scale=1
298
  return save_path, "Dynamic Dry-Season Anchor (April-May)"
299
 
300
  except Exception as e:
301
- print(f"❌ Dynamic World Error {filename_prefix}: {e}", flush=True)
302
  return None, None
 
29
  try:
30
  resp = requests.get(url, timeout=config.GEE_REQUEST_TIMEOUT)
31
  if resp.status_code != 200:
32
+ config.logger.warning(f"❌ Tile {r}-{c} failed with status {resp.status_code}.")
33
  return None
34
 
35
  path = os.path.join(config.DOWNLOADS_DIR, f"{prefix}_tile_{r}_{c}.tiff")
 
37
  f.write(resp.content)
38
  return path
39
  except Exception as e:
40
+ config.logger.error(f"❌ Tile download failed {r}-{c}: {e}")
41
  return None
42
 
43
 
 
170
  return save_path, actual_date, source_type, cloud_percent
171
 
172
  except Exception as e:
173
+ config.logger.error(f"❌ GEE Optical Error {filename_prefix}: {e}")
174
  return None, None, None, 100.0
175
 
176
 
 
239
  return save_path, actual_date, source_type
240
 
241
  except Exception as e:
242
+ config.logger.error(f"❌ SAR Error {filename_prefix}: {e}")
243
  return None, None, None
244
 
245
 
 
298
  return save_path, "Dynamic Dry-Season Anchor (April-May)"
299
 
300
  except Exception as e:
301
+ config.logger.error(f"❌ Dynamic World Error {filename_prefix}: {e}")
302
  return None, None
src/services/gemini.py CHANGED
@@ -87,5 +87,5 @@ Provide 3 highly specific, actionable recommendations for on-the-ground disaster
87
 
88
  except Exception as e:
89
  import traceback
90
- print(f"❌ AI Report Error: {e}\n{traceback.format_exc()}", flush=True)
91
  return "Error: AI Report generation failed due to an internal system error."
 
87
 
88
  except Exception as e:
89
  import traceback
90
+ config.logger.error(f"❌ AI Report Error: {e}\n{traceback.format_exc()}")
91
  return "Error: AI Report generation failed due to an internal system error."
src/services/model.py CHANGED
@@ -52,10 +52,10 @@ def initialize_model():
52
  """
53
  global _model
54
 
55
- print("⏳ Loading model...", flush=True)
56
 
57
  if not os.path.exists(config.CHECKPOINT_PATH):
58
- print(f"❌ Model Checkpoint Not Found: {config.CHECKPOINT_PATH}", flush=True)
59
  _model = None
60
  return None
61
 
@@ -66,10 +66,10 @@ def initialize_model():
66
  )
67
  _model.to(config.DEVICE)
68
  _model.eval()
69
- print("βœ… Model Loaded Successfully!", flush=True)
70
  return _model
71
  except Exception as e:
72
- print(f"❌ Model Load Failed: {e}", flush=True)
73
  _model = None
74
  return None
75
 
 
52
  """
53
  global _model
54
 
55
+ config.logger.info("⏳ Loading model...")
56
 
57
  if not os.path.exists(config.CHECKPOINT_PATH):
58
+ config.logger.error(f"❌ Model Checkpoint Not Found: {config.CHECKPOINT_PATH}")
59
  _model = None
60
  return None
61
 
 
66
  )
67
  _model.to(config.DEVICE)
68
  _model.eval()
69
+ config.logger.info("βœ… Model Loaded Successfully!")
70
  return _model
71
  except Exception as e:
72
+ config.logger.error(f"❌ Model Load Failed: {e}")
73
  _model = None
74
  return None
75