Mirascope + Anthropic + Streamlit Website URL Extraction Demo

[willbakst]

• Uses Mirascope AnthropicExtractor to generate a schema based on a user query.
• Uses another AnthropicExtractor to extract the generated schema from scraped data from the given url.
• Streamlit app for easy use and deployment with Ploomber.
• README blog post

---

📚 Documentation preview 📚: https://ploomber-doc--182.org.readthedocs.build/en/182/

[edublancas]

@bryannho please review this

this is a demo + blog post, please review both. once that's done please open a PR to our website repo so we publish the blog post and link to this example

[edublancas]

@neelasha23 please also review

[bryannho]

@willbakst Looks great! Drafted a blog post which we'll publish pending @neelasha23's review.

@edublancas Do we want to deploy a live demo as well?

[neelasha23]

A few observations:

1. Please add the requirements.lock.txt file
2. Since there is a corresponding PR to blog post, we should remove the blog content from README and include more details on setting up the app locally. Need to add the API keys that should be added as env variables. Also, need to add the URL for generating Anthropic key (https://console.anthropic.com/settings/keys ?)
3. I think we need credits for running it so I haven't been able to test the functionality

@willbakst @edublancas

[edublancas]

shared an anthropic key via slack

[edublancas]

Since there is a corresponding PR to blog post, we should remove the blog content from README

I think let's keep the post (let's just renamed it to post.md and have a readme.md with a shorter description and essential details). once all observations are addressed and this is merged, we can move this to our website's repo (link from the blog post to this repo), and delete it from here

[willbakst]

A few observations:

1. Please add the requirements.lock.txt file
2. Since there is a corresponding PR to blog post, we should remove the blog content from README and include more details on setting up the app locally. Need to add the API keys that should be added as env variables. Also, need to add the URL for generating Anthropic key (https://console.anthropic.com/settings/keys ?)
3. I think we need credits for running it so I haven't been able to test the functionality

@willbakst @edublancas

1. Done
2. I've renamed the post to post.md as Eduardo suggested and added a true README.md
3. Yes, you need an API key with credits.

[neelasha23]

Works fine but I think we need to refresh for every query? The first time I tried a new query it displayed the previous query's answers and then an error. but on refreshing the page it shows the correct response.

First attempt (Without refresh). Note that it still displays the painting title and links output of the default query

Second attempt:

@willbakst @edublancas

[willbakst]

@neelasha23 I've just pushed some changes to hopefully fix your concerns:

1. Your first attempt seems to have hit the token/min limit (which is why it failed).
2. I've updated the streamlit app so if you change the url or query it should clear the previous output. It will also clear when you hit Extract

[edublancas]

@willbakst great work! I'll follow up with the team so we publish this in our blog, will send you the link once it's live!