💡 idea: Detailed instruction on how to enable and configure Embedding Search

[jprusch]

Description

Latest plugin 1.2.0 / server 10.9.0

I installed the pgvector plugin on our PostgreSQL server and also configured it for use by the Mattermost db/schema.

We also have enabled ElasticSearch as search backend.

When I try to reindex the post, an error message is shown "Failed to start reindexing". This might be a bug or a misconfiguration.

Log shows:

error [2025-06-16 17:30:11.573 +02:00] search functionality is not configured caller="app/plugin_api.go:1112" plugin_id=mattermost-ai

To rule out a misconfiguration, some clear instruction what to do to enable Embedding Search would be a great step forward.

[crspeller]

Here is a sample working configuration with PGVector:

"embeddingSearchConfig": {
    "chunkingOptions": {
        "chunkOverlap": 200,
        "chunkSize": 1000,
        "chunkingStrategy": "sentences",
        "minChunkSize": 0.75
    },
    "dimensions": 768,
    "embeddingProvider": {
        "parameters": {
            "apiKey": "",
            "apiURL": "http://localhost:11434/v1",
            "embeddingModel": "nomic-embed-text"
        },
        "type": "openai-compatible"
    },
    "parameters": {},
    "type": "composite",
    "vectorStore": {
        "parameters": {},
        "type": "pgvector"
    }
}

But the error you are getting sounds like you might not have "type": "composite" set.