execution time is now accounted for in recording timing

blueprints/event_summary.yaml

@@ -213,14 +213,14 @@ action:
                 sequence:
                   - service: llmvision.image_analyzer
                     data:
-                      image_entity: "{{trigger.payload_json['after']['camera']}}"
+                      image_entity: "{{ ['camera.' + trigger.payload_json['after']['camera']|lower] }}"
                       provider: !input provider
                       model: !input model
                       message: "{{importance_prompt}}"
                       include_filename: true
                       target_width: 1280
                       detail: low
-                      max_tokens: 5
+                      max_tokens: 1
                       temperature: 0.1
                     response_variable: importance
               - conditions:
@@ -236,7 +236,7 @@ action:
                       include_filename: true
                       target_width: 1280
                       detail: low
-                      max_tokens: 5
+                      max_tokens: 1
                       temperature: 0.1
                     response_variable: importance
 
@@ -264,7 +264,7 @@ action:
             data:
               video: "{{video}}"
               image: "{{image}}"
-              entity_id: "{{trigger.entity_id if mode=='Camera' else ''}}"
+              entity_id: "{{trigger.entity_id | default('') if mode=='Camera' else ''}}"
               url: !input tap_navigate #iOS
               clickAction: !input tap_navigate #Android
               tag: "{{tag}}"
@@ -278,7 +278,7 @@ action:
         sequence:
           - service: llmvision.video_analyzer
             data:
-              event_id: '{{trigger.payload_json["after"]["id"]}}'
+              event_id: "{{ trigger.payload_json['after']['id'] }}"
               provider: !input provider
               model: !input model
               message: !input message
@@ -319,7 +319,7 @@ action:
       group: "{{group}}"
       image: "{{image}}"
       video: "{{video}}"
-      entity_id: "{{trigger.entity_id if mode=='Camera' else ''}}"
+      entity_id: "{{trigger.entity_id | default('') if mode=='Camera' else ''}}"
       url: !input tap_navigate #iOS
       clickAction: !input tap_navigate #Android
       push:

custom_components/llmvision/media_handlers.py

@@ -95,7 +95,6 @@ async def resize_image(self, target_width, image_path=None, image_data=None, img
             img_byte_arr.write(image_data)
             img = await self.hass.loop.run_in_executor(None, Image.open, img_byte_arr)
             with img:
-                _LOGGER.debug(f"Image format: {img.format}")
                 img = self._convert_to_rgb(img)
                 # calculate new height based on aspect ratio
                 width, height = img.size
@@ -133,7 +132,6 @@ async def record(self, image_entities, duration, max_frames, target_width, inclu
         import asyncio
 
         interval = 1 if duration < 3 else 2 if duration < 10 else 4 if duration < 30 else 6 if duration < 60 else 10
-        duration += 1 # add 1 second in case the first fetch fails
         camera_frames = {}
 
         # Record on a separate thread for each camera
@@ -142,18 +140,24 @@ async def record_camera(image_entity, camera_number):
             frame_counter = 0
             frames = {}
             previous_frame = None
-            while time.time() - start < duration:
+            iteration_time = 0
+
+            while time.time() - start < duration + iteration_time:
+                fetch_start_time = time.time()
                 base_url = get_url(self.hass)
-                # fetched every cycle to ensure latest access token
                 frame_url = base_url + \
                     self.hass.states.get(image_entity).attributes.get(
                         'entity_picture')
                 frame_data = await self.client._fetch(frame_url)
-                
-                # Skip frame when fetch failed
+
+                # Skip frame if fetch failed
                 if not frame_data:
                     continue
 
+                fetch_duration = time.time() - fetch_start_time
+                _LOGGER.info(f"Fetched {image_entity} in {fetch_duration:.2f} seconds")
+
+                preprocessing_start_time = time.time()
                 img = Image.open(io.BytesIO(frame_data))
                 current_frame_gray = np.array(img.convert('L'))
 
@@ -178,7 +182,17 @@ async def record_camera(image_entity, camera_number):
                     # Initialize previous_frame with the first frame
                     previous_frame = current_frame_gray
 
-                await asyncio.sleep(interval)
+                preprocessing_duration = time.time() - preprocessing_start_time
+                _LOGGER.info(f"Preprocessing took: {preprocessing_duration:.2f} seconds")
+
+                adjusted_interval = max(0, interval - fetch_duration - preprocessing_duration)
+
+                if iteration_time == 0:
+                    iteration_time = time.time() - start
+                    _LOGGER.info(f"First iteration took: {iteration_time:.2f} seconds, interval adjusted to: {adjusted_interval}")
+
+                await asyncio.sleep(adjusted_interval)
+
             camera_frames.update({image_entity: frames})
 
         _LOGGER.info(f"Recording {', '.join([entity.replace(
@@ -217,10 +231,12 @@ async def add_images(self, image_entities, image_paths, target_width, include_fi
                         self.hass.states.get(image_entity).attributes.get(
                             'entity_picture')
                     image_data = await self.client._fetch(image_url)
-                    
-                    # Skip frame when fetch failed
+
+                    # Skip frame if fetch failed
                     if not image_data:
-                        continue
+                        if len(image_entities) == 1:
+                            raise ServiceValidationError(
+                                f"Failed to fetch image from {image_entity}")
 
                     # If entity snapshot requested, use entity name as 'filename'
                     self.client.add_frame(
@@ -265,10 +281,10 @@ async def add_videos(self, video_paths, event_ids, max_frames, target_width, inc
                     base_url = get_url(self.hass)
                     frigate_url = base_url + "/api/frigate/notifications/" + event_id + "/clip.mp4"
                     clip_data = await self.client._fetch(frigate_url)
-
-                    # Skip frame when fetch failed
+
                     if not clip_data:
-                        continue
+                            raise ServiceValidationError(
+                                f"Failed to fetch frigate clip {event_id}")
 
                     # create tmp dir to store video clips
                     os.makedirs(tmp_clips_dir, exist_ok=True)

custom_components/llmvision/request_handlers.py

@@ -1,6 +1,7 @@
 from homeassistant.exceptions import ServiceValidationError
 from homeassistant.helpers.aiohttp_client import async_get_clientsession
 import logging
+import asyncio
 import inspect
 from .const import (
     DOMAIN,
@@ -415,8 +416,7 @@ async def _post(self, url, headers, data):
             _LOGGER.info(f"Response data: {response_data}")
             return response_data
 
-    async def _fetch(self, url, max_retries=1, retry_delay=1):
-        import asyncio
+    async def _fetch(self, url, max_retries=2, retry_delay=1):
         """Fetch image from url and return image data"""
         retries = 0
         while retries < max_retries:
@@ -435,8 +435,7 @@ async def _fetch(self, url, max_retries=1, retry_delay=1):
                 _LOGGER.error(f"Fetch failed: {e}")
                 retries += 1
                 await asyncio.sleep(retry_delay)
-
-        _LOGGER.error(f"Failed to fetch {url} after {max_retries} retries")
+        _LOGGER.warning(f"Failed to fetch {url} after {max_retries} retries")
         return None
 
     def _validate_call(self, provider, api_key, base64_images, ip_address=None, port=None):