Apache Answer Plugin: Milvus Vector Search

This plugin enables semantic/vector search in Apache Answer using Milvus.

Prerequisites

Milvus 2.x instance (self-hosted or Zilliz Cloud)
An OpenAI-compatible embedding API (e.g., OpenAI, Azure OpenAI, or any compatible provider)

Installation

Build Apache Answer with this plugin:

./answer build --with github.com/apache/answer-plugins/vector-search-milvus

Configuration

After enabling the plugin in the Admin UI (Admin > Plugins > Vector Search), configure the following fields:

Field	Description	Example
Milvus Endpoint	Milvus gRPC endpoint	`localhost:19530`
Embedding API Host	OpenAI-compatible API base URL	`https://api.openai.com`
Embedding API Key	API key for the embedding service	`sk-...`
Embedding Model	Model name for generating embeddings	`text-embedding-3-small`
Embedding Level	`question` embeds question + all answers + comments together; `answer` embeds each answer separately	`question`
Similarity Threshold	Minimum cosine similarity score (0-1). Default `0` means no filtering	`0.5`

How It Works

Embedding dimensions are auto-detected from the configured model. No manual dimension configuration is needed.
On first configuration, the plugin creates a collection answer_vector_embeddings with VarChar fields and a FloatVector field.
An IvfFlat index with cosine metric is created on the embedding field.
If the embedding model changes and produces different dimensions, the collection is automatically dropped and recreated.
Upsert is implemented via delete-by-expression + insert (Milvus v2 pattern).
A full sync of all questions/answers is triggered when the plugin starts.

Running Milvus Locally

Using Docker Compose:

# Download the docker-compose file
wget https://github.com/milvus-io/milvus/releases/download/v2.4.0/milvus-standalone-docker-compose.yml -O docker-compose.yml

# Start Milvus
docker compose up -d

The default gRPC endpoint will be localhost:19530.

License

Apache License 2.0