johnpccd/image_management_api
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix vector db dimensionality

Browse Source
This commit is contained in:
johnpccd 2025-05-25 12:48:26 +02:00
parent 273b20a76c
commit d3d6e9050c
2 changed files with 23 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
README.md
View File

@ -8,7 +8,7 @@ SEREACT is a secure API for storing, organizing, and retrieving images with adva
- Team-based organization and access control - Team-based organization and access control
- **Hybrid authentication model**: Public management endpoints + API key protected data endpoints - **Hybrid authentication model**: Public management endpoints + API key protected data endpoints
- **Asynchronous image processing with Pub/Sub and Cloud Functions** - **Asynchronous image processing with Pub/Sub and Cloud Functions**
- **AI-powered image embeddings using Google Cloud Vision API** - **AI-powered image embeddings using Google Vertex AI Multimodal Embedding API**
- **Semantic search using vector similarity with Qdrant Vector Database** - **Semantic search using vector similarity with Qdrant Vector Database**
- **Self-hosted vector database on Google Compute Engine VM** - **Self-hosted vector database on Google Compute Engine VM**
- **Automatic retry mechanism for failed processing (up to 3 attempts)** - **Automatic retry mechanism for failed processing (up to 3 attempts)**
@ -102,7 +102,7 @@ sereact/
- **Cloud Function is triggered by Pub/Sub message** - **Cloud Function is triggered by Pub/Sub message**
- Function updates image status to `"processing"` - Function updates image status to `"processing"`
- **Function downloads image from Cloud Storage** - **Function downloads image from Cloud Storage**
- **Function calls Google Cloud Vision API to generate embeddings** - **Function calls Google Vertex AI Multimodal Embedding API to generate 1408-dimensional embeddings**
- **Embeddings are stored in Qdrant Vector Database on dedicated VM** - **Embeddings are stored in Qdrant Vector Database on dedicated VM**
- **Firestore is updated with embedding info and status: "success"** - **Firestore is updated with embedding info and status: "success"**
@ -123,7 +123,7 @@ sereact/
- **Google Cloud Storage** - Image storage - **Google Cloud Storage** - Image storage
- **Google Pub/Sub** - Message queue for async processing - **Google Pub/Sub** - Message queue for async processing
- **Google Cloud Functions** - Serverless image processing - **Google Cloud Functions** - Serverless image processing
- **Google Cloud Vision API** - AI-powered image analysis and embedding generation - **Google Vertex AI Multimodal Embedding API** - AI-powered image analysis and embedding generation
- **Qdrant** - Self-hosted vector database for semantic search (on Google Compute Engine VM) - **Qdrant** - Self-hosted vector database for semantic search (on Google Compute Engine VM)
- **Google Compute Engine** - VM hosting for vector database - **Google Compute Engine** - VM hosting for vector database
- **Pydantic** - Data validation - **Pydantic** - Data validation
@ -155,6 +155,20 @@ The system includes a dedicated Google Compute Engine VM running Qdrant vector d
- **Cosine Similarity**: Optimized for image embedding comparisons - **Cosine Similarity**: Optimized for image embedding comparisons
- **Metadata Filtering**: Support for complex search filters - **Metadata Filtering**: Support for complex search filters
### **AI Embedding Model**
SEREACT uses Google's Vertex AI multimodal embedding model for generating high-quality image embeddings:
- **Model**: `multimodalembedding@001`
- **Provider**: Google Vertex AI
- **Type**: Multimodal embedding model (supports both images and text)
- **Output Dimensions**: **1408-dimensional vectors**
- **Data Type**: Float32, normalized vectors (norm ≈ 1.0)
- **Similarity Metric**: Cosine similarity
- **Use Case**: Semantic image search and similarity matching
**⚠️ Important Configuration**: Ensure your Qdrant collection is configured with **1408 dimensions** to match the Vertex AI model output. Dimension mismatches will cause embedding storage failures.
## Setup and Installation ## Setup and Installation
### Prerequisites ### Prerequisites
@ -197,8 +211,9 @@ The system includes a dedicated Google Compute Engine VM running Qdrant vector d
PUBSUB_TOPIC=image-processing-topic PUBSUB_TOPIC=image-processing-topic
PUBSUB_SUBSCRIPTION=image-processing-subscription PUBSUB_SUBSCRIPTION=image-processing-subscription
# Google Cloud Vision # Google Vertex AI (for image embeddings)
VISION_API_ENABLED=true VERTEX_AI_LOCATION=us-central1
GOOGLE_CLOUD_PROJECT=your-gcp-project-id
# Security # Security
API_KEY_SECRET=your-secret-key API_KEY_SECRET=your-secret-key
@ -206,6 +221,8 @@ The system includes a dedicated Google Compute Engine VM running Qdrant vector d
# Vector database (Qdrant) # Vector database (Qdrant)
QDRANT_HOST=your-vm-external-ip QDRANT_HOST=your-vm-external-ip
QDRANT_API_KEY=your-qdrant-api-key # Optional QDRANT_API_KEY=your-qdrant-api-key # Optional
QDRANT_COLLECTION=image_vectors
QDRANT_HTTPS=false
``` ```
5. **Deploy Infrastructure** (Required for vector database): 5. **Deploy Infrastructure** (Required for vector database):

2
src/services/vector_db.py
View File

@ -86,7 +86,7 @@ class VectorDatabaseService:
self.client.create_collection( self.client.create_collection(
collection_name=self.collection_name, collection_name=self.collection_name,
vectors_config=VectorParams( vectors_config=VectorParams(
size=512, # Typical size for image embeddings size=1408, # Typical size for image embeddings
distance=Distance.COSINE distance=Distance.COSINE
) )
) )