Description

In the Swagger docs for the /v1/openai/embeddings it gives an example value for the input as:

{
  "input": [
    "This is my first string to embed",
    "This is my second string to embed"
  ],
  "model": null
}

However, the input should be plural inputs otherwise an error gets thrown. I.e.

{
  "inputs": [
    "This is my first string to embed",
    "This is my second string to embed"
  ],
  "model": "text-embedding-3-small"
}

In addition, it should be made clear that in order for this endpoint to work with the correct embedding engine (rather than the default NativeEmbedder), the environment variable EMBEDDING_ENGINE should be set (e.g. to openai) otherwise a 500 error gets thrown.

It would also be useful to give an example of the model, e.g. text-embedding-3-small.

Just for reference if anyone else comes across this issue, the options for the EMBEDDING_ENGINE are

• openai
• azure
• localai
• ollama
• native
• lmstudio
• cohere
• voyageai
• litellm
• mistral
• generic-openai

And they can be seen in function getEmbeddingEngineSelection in server/utils/helpers/index.js