Building with recent libraries

[talkingtoaj]

Any thoughts on a recommended approach to equip this fantastic project to work with recent python libraries or repos?

For example:

• GPT4 does not know about LangChain, it is not included in its training data. Thus, gpt-engineer could never hope to build gpt-engineer!
• There might be an open-source repo the user wants to use within their project, that is well documented, if it isn't in the LLM's training set, we can't.

Possible solutions:

• A method to add a flag when launching gpt-engineer that prompts it to ask the user for documentation pages to read before beginning to digest the prompt file.

$ gpt-engineer my_folder --add_documentation_sources

Please add URL links to documentation folders to be parsed, each separated by a space
>> https://github.com/dmitrizzle/chat-bubble https://github.com/langchain-ai/langchain

Checking...
LangChain found.
Chat-Bubble found.

Please provide any URL link to documentation for LangChain, or just hit enter if none exists. It is best to provide the root directory from which I can crawl to find all subpages.
>> https://python.langchain.com/docs/

Checking... 
25 pages downloaded...

Please provide any URL link to documentation for Chat-Bubble, or just hit enter if none exists. It is best to provide the root directory from which I can crawl to find all subpages.
>> 

No link provided. Trying README.md ... 
1 page downloaded...

Downloading...
Parsing...