ollama-qdrant-vectorstore-example

Qdrant Vector Store Example with Ollama Embeddings

This example demonstrates how to use the Qdrant vector store with Ollama for semantic search. The program initializes the store, adds a sample dataset of cities, performs several types of queries, and then cleans up the created resources.

Features Demonstrated

• Initializing the Qdrant store with an Ollama embedder.
• Creating a collection automatically.
• Adding documents and generating embeddings with nomic-embed-text.
• Performing similarity searches.
• Applying metadata filters to narrow down search results.
• Using a score threshold to filter by relevance.
• Deleting documents from the collection.

Prerequisites

Before running the example, ensure you have the following installed and running:

1. Go: Version 1.21 or later.
2. Ollama: The Ollama server must be running.
3. Qdrant: The easiest way to run Qdrant is with Docker:

   docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant
   

Setup

1. Pull the Embedding Model: The example uses the nomic-embed-text model. Pull it from Ollama:

   ollama pull nomic-embed-text
   

2. Install Go Dependencies: Navigate to the root of the goframe project and run:

   go mod tidy
   

How to Run

Execute the main.go file from the root of the project:

go run ./examples/ollama-qdrant-vectorstore-example/main.go

Expected Output

You will see a series of logs indicating the progress of the demo:

1. Initialization of Ollama and Qdrant components.
2. Creation of a new collection (e.g., cities_demo_...).
3. Successful addition of 7 documents.
4. A series of search scenarios will be executed, each with its own header and results printed to the console (e.g., === Basic Capital Search ===).
5. Logs showing that the metadata filter for "Europe" worked correctly.
6. A demonstration of document deletion.
7. Finally, a cleanup message indicating the test collection has been deleted.
8. The program will exit with the message: Vector store demo completed successfully.