batmanscode

> The llm_predictor is also specified when building the index, and the default is OpenAI. That is why you still need the API Key, I think. I see, thanks!

> Hi You mean the number of retweets? Sorry no I mean there is a boolean `retweet` column that indicated whether or not something tweeted by a user is a...

@ClemLanoe in the [wiki](https://github.com/twintproject/twint/wiki/Configuration) there's `Retweets (bool) - Get retweets done by the user.` Haven't tried it yet but give it a shot if you haven't already.

> @ClemLanoe in the [wiki](https://github.com/twintproject/twint/wiki/Configuration) there's `Retweets (bool) - Get retweets done by the user.` > > Haven't tried it yet but give it a shot if you haven't already....

Maybe an option to customize the prompt would be a general solution

Instruct models would be great! It can be "fine tuned" via prompt to do asymmetric retrieval (not just similarity) See an example here: https://github.com/batmanscode/Talk2Book/blob/main/Book2Vec.ipynb

`include_context` mode sounds good. Might be useful for other loaders too. > I think it's reasonable to include the entire thread in a `Document`. > The text splitters will take...

> Hm interesting. I've heard that using ### to separate things is generally effective. So basically your document would end up being a bunch of tweets in a thread separated...

Having a similar issue. The package is in requirements and installed via pypi but I still get a module not found error. Link to action: https://github.com/batmanscode/LinkNinja/actions/runs/3720662974/jobs/6310286199

https://github.com/muellerzr/nbdev-extensions should be able to help with that 😄