Hello,

After upgrading Meilisearch from version 1.9 to 1.11, I'm encountering errors when configuring the custom embedder. Below is my previous request configuration that worked in version 1.9:

curl --location --request PATCH 'http://10.8.2.70:7006/indexes/spp_doc/settings' \
--header 'Content-Type: application/json' \
--data '{
    "embedders": {
        "default": {
            "source": "rest",
            "url": "http://10.8.2.70:7005/v1/embeddings",
            "dimensions": 1024,
            "documentTemplate": "标题:{{doc.title}};工艺描述:{{doc.step_content}}",
            "inputField": [
                "input"
            ],
            "inputType": "textArray",
            "query": {
                "model": "host-bge-large-zh-v1.5"
            },
            "pathToEmbeddings": [
                "data"
            ],
            "embeddingObject": [
                "embedding"
            ]
        }
    }
}'

In the latest documentation, I noticed that the rest source no longer supports the model configuration, and it seems that both request and response configurations are now mandatory.

I'm using Xinference as my local service for embeddings. Xinference is compatible with OpenAI in some aspects.

I have a few questions:

1. In my case, should I set the source to openAi or rest?
2. If I use rest, how can I specify the model name since model and rest cannot be used together?
3. If I use rest, how should I configure the request and response fields? I couldn't fully understand this part from the documentation.

I couldn't find detailed explanations or configurations for these parameters in the documentation. Could you please provide guidance on how to configure the custom embedder in the current version? Additionally, is there detailed blog available for these configurations?

Thank you!