dearwolf/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2025-05-28 14:35:00 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 12

feat(api): allow to pass audios to backends (#3603)

Browse source 
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2024-09-19 12:26:53 +02:00 • committed by GitHub
parent fbb9facda4
commit 191bc2e50a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 27 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

6
core/http/endpoints/openai/chat.go
View file

@ -644,8 +644,12 @@ func handleQuestion(config *config.BackendConfig, input *schema.OpenAIRequest, m
for _, m := range input.Messages {
videos = append(videos, m.StringVideos...)
}
audios := []string{}
for _, m := range input.Messages {
audios = append(audios, m.StringAudios...)
}
predFunc, err := backend.ModelInference(input.Context, prompt, input.Messages, images, videos, ml, *config, o, nil)
predFunc, err := backend.ModelInference(input.Context, prompt, input.Messages, images, videos, audios, ml, *config, o, nil)
if err != nil {
log.Error().Err(err).Msg("model inference failed")
return "", err

6
core/http/endpoints/openai/inference.go
View file

@ -31,9 +31,13 @@ func ComputeChoices(
for _, m := range req.Messages {
videos = append(videos, m.StringVideos...)
}
audios := []string{}
for _, m := range req.Messages {
audios = append(audios, m.StringAudios...)
}
// get the model function to call for the result
predFunc, err := backend.ModelInference(req.Context, predInput, req.Messages, images, videos, loader, *config, o, tokenCallback)
predFunc, err := backend.ModelInference(req.Context, predInput, req.Messages, images, videos, audios, loader, *config, o, tokenCallback)
if err != nil {
return result, backend.TokenUsage{}, err
}

14
core/http/endpoints/openai/request.go
View file

@ -135,7 +135,7 @@ func updateRequestConfig(config *config.BackendConfig, input *schema.OpenAIReque
}
// Decode each request's message content
imgIndex, vidIndex := 0, 0
imgIndex, vidIndex, audioIndex := 0, 0, 0
for i, m := range input.Messages {
switch content := m.Content.(type) {
case string:
@ -160,9 +160,19 @@ func updateRequestConfig(config *config.BackendConfig, input *schema.OpenAIReque
// set a placeholder for each image
input.Messages[i].StringContent = fmt.Sprintf("[vid-%d]", vidIndex) + input.Messages[i].StringContent
vidIndex++
case "audio_url", "audio":
// Decode content as base64 either if it's an URL or base64 text
base64, err := utils.GetContentURIAsBase64(pp.AudioURL.URL)
if err != nil {
log.Error().Msgf("Failed encoding image: %s", err)
continue CONTENT
}
input.Messages[i].StringAudios = append(input.Messages[i].StringAudios, base64) // TODO: make sure that we only return base64 stuff
// set a placeholder for each image
input.Messages[i].StringContent = fmt.Sprintf("[audio-%d]", audioIndex) + input.Messages[i].StringContent
audioIndex++
case "image_url", "image":
// Decode content as base64 either if it's an URL or base64 text
base64, err := utils.GetContentURIAsBase64(pp.ImageURL.URL)
if err != nil {
log.Error().Msgf("Failed encoding image: %s", err)