Consuming a custom Azure AI Foundry Endpoint #7166
scara1701
started this conversation in
Help Wanted
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
I managed to get our OpenAI Azure endpoint working with the 4o-mini model. Which works on librechat with following config:
azureOpenAI: groups: - group: "somename-gpt4o" apiKey: "somekey" serverless: true baseURL: "https://somename.openai.azure.com/openai/deployments/somename-gpt4o-mini" version: "2023-12-01-preview" models: gpt-4o-mini: deploymentName: somename-gpt4o-mini version: "2024-07-18"
I also have published an Azure AI Prompt flow (a chat flow) to an endpoint on azure.
Now I would like to utilize this endpoint on a libreChat interface. I am however a bit confused on how to implement this.
As an input the flow expects a 'Chat_history' (a list of previous messages) and 'question'.
custom: - name: "anothername" apiKey: "anotherkey" baseURL: "https://anothername.westeurope.inference.ml.azure.com/score" directEndpoint: true models: default: ["modelname"]
The endpoint responds with an error 424, due toe the missing input fields ['chat_history', 'question'] . How do I add these to my request in the model?
Am I supposed to use the addParams option? If so, how can I refer to the chat history and most recent question?
This is the swagger generated by Azure AI for the endpoint.
{ "components": { "securitySchemes": { "bearerAuth": { "scheme": "bearer", "type": "http" } } }, "info": { "title": "Promptflow[model] API", "version": "1.0.0", "x-chat-history": "chat_history", "x-chat-input": "question", "x-chat-output": "answer", "x-flow-name": "model", "x-flow-type": "chat" }, "openapi": "3.0.0", "paths": { "/feedback": { "parameters": [ { "description": "flatten the feedback data into traced data", "in": "query", "name": "flatten", "required": false, "schema": { "type": "boolean" } } ], "post": { "requestBody": { "content": { "application/json": { "schema": { "additionalProperties": {}, "type": "object" } } }, "description": "promptflow feedback data", "required": true }, "responses": { "200": { "content": { "application/json": { "schema": { "additionalProperties": {}, "type": "object" } } }, "description": "successful operation" }, "400": { "description": "Invalid input" }, "default": { "description": "unexpected error" } }, "summary": "collect promptflow feedback" } }, "/score": { "post": { "requestBody": { "content": { "application/json": { "example": {}, "schema": { "properties": { "chat_history": { "items": { "additionalProperties": {}, "type": "object" }, "type": "array" }, "question": { "type": "string" } }, "required": [ "chat_history", "question" ], "type": "object" } } }, "description": "promptflow input data", "required": true }, "responses": { "200": { "content": { "application/json": { "schema": { "properties": { "answer": { "type": "string" } }, "type": "object" } } }, "description": "successful operation" }, "400": { "description": "Invalid input" }, "default": { "description": "unexpected error" } }, "summary": "run promptflow: model with an given input" } } }, "security": [ { "bearerAuth": [] } ] }
Beta Was this translation helpful? Give feedback.
All reactions