Merge branch 'main' into mme_optimizations

Author: sun1lach

microservices/audio-analyzer/docker/compose-minio.yaml

@@ -1,6 +1,6 @@
 services:
   minio:
-    image: minio/minio:latest
+    image: chainguard/minio@sha256:cb84dfa704c648c4b14858aa288576bb1cf756a9b326112a0934db00e87d0bb8
     ports:
       - "9000:9000"
       - "9001:9001"
@@ -11,7 +11,7 @@ services:
       - MINIO_ROOT_PASSWORD=${MINIO_SECRET_KEY}
     command: server /data --console-address ":9001"
     healthcheck:
-      test: ["CMD", "curl", "-f", "http://localhost:9000/minio/health/live"]
+      test: [ "CMD-SHELL", "echo > /dev/tcp/127.0.0.1/80 && exit 0 || exit 1" ]
       interval: 30s
       timeout: 10s
       retries: 3

microservices/visual-data-preparation-for-retrieval/vdms/docker/compose-with-embedding.yaml

@@ -3,7 +3,7 @@
 
 services:
     minio-server:
-        image: minio/minio:RELEASE.2025-02-07T23-21-09Z
+        image: chainguard/minio@sha256:cb84dfa704c648c4b14858aa288576bb1cf756a9b326112a0934db00e87d0bb8
         environment:
             - MINIO_ROOT_USER=${MINIO_ROOT_USER}
             - MINIO_ROOT_PASSWORD=${MINIO_ROOT_PASSWORD}

microservices/visual-data-preparation-for-retrieval/vdms/docker/compose.yaml

@@ -2,7 +2,7 @@
 # SPDX-License-Identifier: Apache-2.0
 services:
     minio-server:
-        image: minio/minio:RELEASE.2025-02-07T23-21-09Z
+        image: chainguard/minio@sha256:cb84dfa704c648c4b14858aa288576bb1cf756a9b326112a0934db00e87d0bb8
         environment:
             - MINIO_ROOT_USER=${MINIO_ROOT_USER}
             - MINIO_ROOT_PASSWORD=${MINIO_ROOT_PASSWORD}

microservices/visual-data-preparation-for-retrieval/vdms/src/common/schema.py

@@ -157,6 +157,25 @@ class VideoSummaryRequest(BaseModel):
     bucket_name: Annotated[
         str, Field(description="The Minio bucket name where the referenced video is stored")
     ]
+    video_id: Annotated[
+        str, Field(description="The video ID (directory) containing the referenced video")
+    ]
+    video_summary: Annotated[
+        str, Field(description="The text summary for the video to be embedded")
+    ]
+    video_start_time: Annotated[
+        float, Field(description="Start timestamp in seconds for the video or video chunk")
+    ]
+    video_end_time: Annotated[
+        float, Field(description="End timestamp in seconds for the video or video chunk")
+    ]
+    tags: Annotated[
+        Optional[List[str]],
+        Field(
+            default_factory=list,
+            description="List of tags to be associated with the video. Useful for filtering the search.",
+        ),
+    ]
 
 
 class TelemetryStageTiming(BaseModel):

microservices/visual-data-preparation-for-retrieval/vdms/src/endpoints/document_processing/process_text.py

@@ -1,10 +1,12 @@
 # Copyright (C) 2025 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
+import datetime
 from http import HTTPStatus
 from typing import Annotated, List
 
 from fastapi import APIRouter, Body, HTTPException
+from tzlocal import get_localzone
 
 from src.common import DataPrepException, Strings, logger
 from src.common.schema import DataPrepResponse, VideoSummaryRequest
@@ -114,6 +116,13 @@ async def process_video_summary(
             bucket_name, video_id, video_start_time, video_end_time, video_summary, tags
         )
 
+        local_timezone = get_localzone()
+        created_at = (
+            datetime.datetime.now(datetime.timezone.utc)
+            .astimezone(local_timezone)
+            .isoformat()
+        )
+
         # Create metadata for summary text
         text_metadata = {
             "bucket_name": bucket_name,
@@ -124,6 +133,7 @@ async def process_video_summary(
             "content_type": "text",
             "timestamp": video_start_time,
             "tags": comma_separated_tags,
+            "created_at": created_at,
         }
 
         logger.info(f"Text metadata for summary: {text_metadata}")

sample-applications/document-summarization/app/.env

@@ -5,4 +5,4 @@ CORS_ALLOW_ORIGINS="*"
 CORS_ALLOW_METHODS="*"
 CORS_ALLOW_HEADERS="*"
 LLM_ENDPOINT_URL="http://ovms-service"
-CHUNK_SIZE=1024
+CHUNK_SIZE=4096

sample-applications/document-summarization/app/server.py

@@ -75,13 +75,14 @@
 )
 
 # Update OpenAILike configuration with proper timeout and retry settings
+
 model = OpenAILike(
     api_base="{}/v3".format(LLM_INFERENCE_URL),
     model=model_name,  
     is_chat_model=True,
     is_function_calling_model=False,
-    timeout=120,  # Increase timeout to 120 seconds
-    max_retries=2,  # Limit number of retries
+    timeout=600,  # Increased timeout for long responses
+    max_retries=10,  # Allow more retries for transient failures
     api_key="not-needed"  # Some implementations require a non-empty API key
 )
 
@@ -198,8 +199,8 @@ async def stream_data_endpoint(file: UploadFile = File(...), query: str = "Summa
         try:
             logger.info("Loading documents")
             if file.filename.endswith(".txt"):
-                logger.info("Chunking TXT file before summarization")
-                documents = chunk_text_file(file_location, max_chars=2000)
+                logger.info(f"Chunking TXT file before summarization with chunk size: {config.CHUNK_SIZE}")
+                documents = chunk_text_file(file_location, max_chars=config.CHUNK_SIZE)
             else:
                 documents = SimpleDirectoryReader(input_files=[file_location]).load_data()
 

sample-applications/document-summarization/app/simple_summary_pack/llama_index/packs/simple_summary/base.py

@@ -49,7 +49,7 @@ def __init__(
         Settings.embed_model = None
         Settings.llm = llm
         self.verbose = verbose
-        self.splitter = SentenceSplitter(chunk_size=config.CHUNK_SIZE or 1024)
+        self.splitter = SentenceSplitter(chunk_size=config.CHUNK_SIZE or 4096)
 
         self.response_synthesizer = get_response_synthesizer(
             response_mode="tree_summarize", use_async=True)

sample-applications/document-summarization/docs/user-guide/troubleshooting.md

@@ -22,6 +22,19 @@ A: Summaries are not stored by default; they are returned in the response.
 
 A: Edit the `.env` file in the project root and restart the services.
 
+### Q: How do I configure the chunk size for document processing?
+
+A: The chunk size determines how large documents are split for processing. You can configure it by setting the `CHUNK_SIZE` environment variable in the `.env` file located in `app/.env`.
+
+**Default value:** 4096 characters per chunk
+
+**Recommended values based on file size:**
+- **Small documents (<100 KB):** 1024-2048 characters
+- **Medium documents (100-200 KB):** 4096-6144 characters
+- **Large documents (>200 KB):** 8192-16384 characters
+
+**Note:** These values are suggestive guidelines. Smaller chunk sizes create more chunks, which increases processing time and may cause timeouts for large files. Larger chunk sizes reduce the number of chunks but require more memory per chunk. **Experiment with different values** to find the optimal setting for your specific use case, file types, and system resources.
+
 ### Q: What is the ideal time for services or pods to become ready when deployed via Helm?
 
 A: The typical initialization time for services and pods deployed using Helm is approximately 6 to 8 minutes, depending on the system resources and network conditions.

sample-applications/video-search-and-summarization/chart/subchart/minio-server/values.yaml

@@ -23,8 +23,8 @@ name: minio-server
 claimSize: ""
 
 image:
-  repository: minio/minio
-  tag: RELEASE.2025-02-07T23-21-09Z-cpuv1
+  repository: chainguard/minio@sha256
+  tag: cb84dfa704c648c4b14858aa288576bb1cf756a9b326112a0934db00e87d0bb8
   pullPolicy: IfNotPresent
 
 containerApiPortName: api-port  # Optional name to refer apiPort