GitHub Incident: Forks not being recognized, PRs automatically closed

[baskaryan]

UPDATE:
Not sure what happened but looks like we're back! We've reopened as many of the closed PRs as we can. Unfortunately we're not able to do this for forks that we don't have write access to, so those contributors will have to reopen the PRs themselves. Apologies for the inconvenience!

---

ORIGINAL:
As of Jan 30, 2024 9:30am PST we're aware that most LangChain forks have stopped being recognized as forks, and the corresponding PRs have automatically been closed. We're in contact with the GitHub team to try and understand why this happened and how to fix it. We'll keep you updated, thanks for your patience!

[jamesbraza]

This is interesting, it didn't seem to affect any other of my GitHub forks. I wonder what about LangChain triggered this, maybe it was some upper limit in number of outstanding forks.

[baskaryan]

Not sure, though repos like https://github.com/pytorch/pytorch have twice as many forks as we did

[baskaryan]

Not sure what happened but looks like we're back! We've reopened as many of the closed PRs as we can. Unfortunately we're not able to do this for forks that we don't have write access to, so those contributors will have to reopen the PRs themselves. Apologies for the inconvenience!

[dluc]

I think I've seen a similar impact a while back when changing a repo visibility from public to private, or when touching permissions/branch protections, can't remember the details...

[codysaint]

Hi folks, Vinay this side from India.

I am actually new to this LLM and LangChain world and I was trying my hands on this langchain part and created a simple chatbot that will provide information based on Indian Budget documents.

Here is the link - https://github.com/codysaint/streamlit-pdf-qa-langchain-app

But as part of the response I also want the citation and url of the source/doc from where it fetched the answer as we see in the Bing Copilot.

As of now I am trying it by modifying the prompt but unable to get proper results.

template = """\
You are an AI assistant having expertise in answering questions about the most recent union budget of India for the year 2024.
You are given the knowledge base of union budget documents and a question. Provide a conversational answer based on appropriate and relevant search from provided knowledge sources.  

Combine search results and URL together into a coherent answer. Do not repeat text. Cite search results using [${{number}}] notation. Only cite the most relevant results that answer the question accurately. Place these citations at the end of the sentence or paragraph that reference them - do not put them all at the end. If different results refer to different entities within the same name, write separate answers for each entity. ALWAYS return a "SOURCES" part in your answer.

Take note of the sources and the URL of relevant documents where that information can be found and include them in the answer in the format: \

"SOURCES: [source1](URL1) [source2](URL2) ...", use "SOURCES" in capital letters regardless of the number of sources.

Remember: If you don't know the answer, just say "Hmm, I'm not sure." Don't try to make up an answer.Use an unbiased tone.

Question: {question}
=========
{context}
=========
Answer in Markdown:"""

I will be grateful as to how I can achieve it.

PFA the screenshot where 1st is my chatbot and 2nd is the output of Copilot with proper citation and urls pertaining to resources.