Mirrors
/
hf_text-generation-inference
mirror of https://github.com/huggingface/text-generation-inference.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update response type for `/v1/chat/completions` and `/v1/completions` (#1747) 

`/v1/chat/completions` and `/v1/completions` have different output types
depending on the `stream` parameter. This PR aims at fixing the
inconsistency in the auto-generated
[openapi.json](https://huggingface.github.io/text-generation-inference/openapi.json)
specs.

cc @OlivierDehaene @drbh I reused what had been done for the `/`
endpoint but haven't tested anything myself. Could you confirm this is
the correct way of handling things?

Also, should I update the openapi.json file manually? If yes, how can I
do it?
This commit is contained in:
Lucain 2024-04-16 19:26:32 +02:00 committed by GitHub
parent 7276d43495
commit 00f365353e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 11 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
router/src/server.rs
View File

@ -548,7 +548,11 @@ async fn generate_stream_internal(
path = "/v1/completions",
request_body = CompletionRequest,
responses(
(status = 200, description = "Generated Text", body = ChatCompletionChunk),
(status = 200, description = "Generated Chat Completion",
content(
("application/json" = Completion),
("text/event-stream" = CompletionCompleteChunk),
)),
(status = 424, description = "Generation Error", body = ErrorResponse,
example = json ! ({"error": "Request failed during generation"})),
(status = 429, description = "Model is overloaded", body = ErrorResponse,
@ -652,7 +656,7 @@ async fn completions(
})
.map_or_else(
|e| {
println!("Failed to serialize ChatCompletionChunk: {:?}", e);
println!("Failed to serialize CompletionCompleteChunk: {:?}", e);
Event::default()
},
|data| data,
@ -725,7 +729,11 @@ async fn completions(
path = "/v1/chat/completions",
request_body = ChatRequest,
responses(
(status = 200, description = "Generated Text", body = ChatCompletionChunk),
(status = 200, description = "Generated Chat Completion",
content(
("application/json" = ChatCompletion),
("text/event-stream" = ChatCompletionChunk),
)),
(status = 424, description = "Generation Error", body = ErrorResponse,
example = json ! ({"error": "Request failed during generation"})),
(status = 429, description = "Model is overloaded", body = ErrorResponse,