community: Add the additonnal kward 'context' for openai

[lucasiscovici]

• Description:
  Add the additonnal kward 'context' for openai into convert_dict_to_message and convert_message_to_dict functions.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchain	⬜️ Ignored (Inspect)	Visit Preview		Dec 2, 2024 5:38pm

[lucasiscovici]

hi :) @baskaryan @ccurme

[ccurme]

Hi @lucasiscovici, can you explain the motivation for this?

[lucasiscovici]

Hi @lucasiscovici, can you explain the motivation for this?

@ccurme In recent versions of the Azure OpenAI Chat With Your Data, a significant enhancement has been introduced with the addition of the "context" property into the response chat message.

The "context" property is added to the assistant's messages to provide extra information about the steps taken by Azure OpenAI while processing a request. This includes retrieved documents, citations, and other relevant metadata.

Chat message

The response assistant message schema inherits from the chat completions assistant chat message, and is extended with the property context.

Name	Type	Required	Description
context	Context	False	Represents the incremental steps performed by the Azure OpenAI On Your Data while processing the request, including the retrieved documents.

Context

Name	Type	Required	Description
citations	Citation[]	False	The data source retrieval result, used to generate the assistant message in the response. Clients can render references from the citations.
intent	string	False	The detected intent from the chat history. Passing back the previous intent is no longer needed. Ignore this property.
all_retrieved_documents	Retrieved documents[]	False	All the retrieved documents.

Citation

Name	Type	Required	Description
content	string	True	The content of the citation.
title	string	False	The title of the citation.
url	string	False	The URL of the citation.
filepath	string	False	The file path of the citation.
chunk_id	string	False	The chunk ID of the citation.

Retrieved documents

Name	Type	Required	Description
search_queries	string[]	True	The search queries used to retrieve the document.
data_source_index	integer	True	The index of the data source.
original_search_score	double	True	The original search score of the retrieved document.
rerank_score	double	False	The rerank score of the retrieved document.
filter_reason	string	False	Represents the rationale for filtering the document. If the document does not undergo filtering, this field will remain unset. Will be score if the document is filtered by original search score threshold defined by strictness. Will be rerank if the document is not filtered by original search score threshold, but is filtered by rerank score and top_n_documents.

[lucasiscovici]

Hi @ccurme and @baskaryan,

Just following up to see if the motivation I shared addresses your questions. Let me know if there’s anything else I can clarify or adjust!

Thanks for your time! 😊

[ccurme]

Hi @lucasiscovici,

Thanks for following up. Can you elaborate on how you are accessing the Azure chat-with-your-data service via LangChain?

Asking in case AzureChatOpenAI is involved, since the community version is deprecated in favor of the version in langchain-openai.

[lucasiscovici]

Hi @lucasiscovici,

Thanks for following up. Can you elaborate on how you are accessing the Azure chat-with-your-data service via LangChain?

Asking in case AzureChatOpenAI is involved, since the community version is deprecated in favor of the version in langchain-openai.

Hi @ccurme, i create the PR for langchain-openai (partners)