llms/mistral: Implementing embeddings.EmbedderClient for Mistral and an example with PGVector #1086
Conversation
|
I did a replace in my mod file for my local dev, and was unsure how to handle that in the PR, which I think is the reason for the CI failures. Happy to take pointers on if I should change something in my repo to fix it. KR |
…ents CreateEmbedding and the embeddings.EmbedderClient interface
|
Ok, I think good to go for reviewing. Also, as said earlier, any pointers or feedback welcome as this is my first PR in this project. |
| "errors" | ||
| ) | ||
|
|
||
| func ConvertFloat64ToFloat32(input []float64) []float32 { |
There was a problem hiding this comment.
I dont think this function needs to be exported.
| allEmbds := make([][]float32, len(embsRes.Data)) | ||
| for i, embs := range embsRes.Data { | ||
| if len(embs.Embedding) == 0 { | ||
| return nil, errors.New("empty embeddings") |
There was a problem hiding this comment.
This could be an exported error variable. var ErrEmptyEmbeddings = errors.new("...")
| t.Run(tt.name, func(t *testing.T) { | ||
| t.Parallel() | ||
| output := sdk.ConvertFloat64ToFloat32(tt.input) | ||
| if len(output) != len(tt.expected) { |
There was a problem hiding this comment.
We should use stretchr/testify in test.
| value := os.Getenv(envVar) | ||
|
|
||
| // Check if it is set (non-empty) | ||
| if value == "" { |
|
|
||
| model, err := sdk.New() | ||
| if err != nil { | ||
| panic(err) |
There was a problem hiding this comment.
Do not panic the errors. Use require.NoError(t, err) from testify instead
| if err != nil { | ||
| panic(err) | ||
| } | ||
| t.Logf("Document embeddings: %v\n", docEmbeddings) |
There was a problem hiding this comment.
I don't think we need to include the log statements.
| } | ||
| fmt.Println("store.SimilaritySearch1:\n", docs) | ||
|
|
||
| time.Sleep(2 * time.Second) // Don't trigger cloudflare |
There was a problem hiding this comment.
Is there no better solution for this?
This PR implements the embeddings.EmbedderClient interface for mistral, using the "mistral-embed" model. There is an example, heavily inspired by the
openai-embeddings-exampleof how it can be used with pgvector.PR Checklist
memory: add interfaces for X, Yorutil: add whizzbang helpers).Fixes #123).golangci-lintchecks.