The difference of Assistant api and langchain

I have watched the newest devday video(especially the wonderlust demo), and i am really curious about what exactly is the assistant api. I am now doing a travel advisor project like the demo shown in the video. My project uses the ‘agents’ in Langchain and interacting with different ‘tools’ that developed by me to get information from outer travel apis.

My question is that since the openai assistant api has only few built-in functions (code interpreter, retierivals…), how is it able to interact with travel apis to get the real information? If I have to coding to develop my customize tools (called functions in assistant api), then what is the difference of this way of developing and the langchian way? The assistant api is supposed to be a user-friendly, non-coding way of building your own applications.

Assistant api and langchain are basically doing the same thing. Both require programming.
The only advantage of Assistant API is that memory and context window are automatically managed where in langchain you have explicitly set those things up.

1 Like

well it could also be that the assistants are structured in a way that the GPT models were finetuned with, and so could be better. It could also be the baked in logic in assistants in different/better/worse than the ones implemented in the RAG/Agent implementation in langchain.

1 Like

In terms of RAG, I specifically find the Langchain better than assistants API.

As this has a marked solution closing this topic.