zainhoda

Author of the package here. That's pretty much what this package does just with optimization around what context gets sent to the LLM about the database:

https://github.com/vanna-ai/vanna/blob/main/src/vanna/base/b...

And then of course once you have the SQL, you can put it in an interface where you the SQL can be run automatically and then get a chart etc.

Awesome! I've been looking for almost exactly this. The only issue is that I'm not totally comfortable installing a Mac app to try it out. I'd be happy to try it out as a web app first.

When you snooze it for >=1 day, does it reappear at the time that it was snoozed or do the tasks somehow group into buckets where if you snooze all tasks that appear today for 1 day, they'll all appear tomorrow at the same time?

Wow, this is super cool! From the code it seems like the speech to text and text to speech are using the browser’s built-in features. I always forget those capabilities even exist!

That information is just regurgitated from the New Yorker article not from this new interview with Toner.

Oh man, I would love to compare notes.

I pivoted in the reverse direction from:

https://github.com/pycob/pyvibe

To

https://github.com/vanna-ai/vanna

> All in all, it felt more like a loose collection of unfinished PhD theses than an actual programming language.

Yeah, to me Haskell is the experimental forefront of programming language theory and the most developed concepts get incorporated into the actual programming languages over time.

Nice! I’m going to try it out and possibly integrate it into my Python package: https://vanna.ai

We tested Bard (aka Bison in GCP) for generating SQL.

It has worse generalization capabilities than even GPT-3.5 but actually does as well at GPT-4 when given contextually relevant examples selected from a large corpus of examples.

https://vanna.ai/blog/ai-sql-accuracy.html

This suggests to me that it needs longer prompts to avoid the hallucination problem that everyone else seems be experiencing.

I like small form factor so I use iPhone 12 mini and before that was the original iPhone SE.

Sadly the small form factor seems to be going away so I’m not really sure what my next phone will be.

Very interesting! I’ve been thinking along similar lines for data stories using vanna.ai — if you’re reading a research report or blog post that has charts, you could theoretically dig deeper into it in the same way that you continue your blog post.

There isn't technically a limit on the storage side but it's generally better if you keep documentation to a manageable length.

You call vn.train(sql=...) on each individual SQL statement that you have.

What'll happen under the hood is the package will use the 10 most relevant SQL statement examples, 10 most relevant pieces of documentation, and 10 most relevant DDL statements.

If using 10 examples exceeds the (approximate) token limit for the model, it'll pare down to a smaller number that'll fit into the context limit.

It probably could but it would require adjusting the prompt. You’d have to override the generate_prompt function and tell it that you’re using a variant of SQL and describe the differences.

Thanks for the link! I'll check it out.

In the meantime, this post we wrote might be interesting for you:

https://vanna.ai/blog/ai-sql-accuracy.html

API key is optional -- that's only if you want to use the hosted vector database. When running locally, no Vanna API key is necessary. Instead, you can put in an OpenAI API key as shown here: https://vanna.ai/docs/local.html

If you want to use a locally-hosted LLM, that's also possible by implementing the necessary abstract methods: https://vanna.ai/docs/vanna.html#open-source-and-extending

We use "train" in a colloquial sense because you're right -- it's RAG. We've tried to think about a different term but "train" seems to be the one that most data analysts resonate with.

I like this! I think I’ll use it for my AI product. It actually really quickly answers a lot of customer questions that we have to address via FAQs currently.

Nice job! We're building something relatively similar at Vanna AI: https://vanna.ai/

Thanks! You can see the 3 static examples in this notebook here: https://github.com/vanna-ai/research/blob/main/notebooks/tra...

The set of larger SQL queries includes the initial static set but is brought into the prompt only if it's contextually relevant to the question. I know that's a confusing statement so let me know if that makes sense.

I'm in a similar boat. I've got ADEL50LYL1. It would be good to know what the risk level is.

In addition to the minisplit cost, there's also the cost of the electrical work. Anyone can plug in a dehumidifier into an outlet but drilling a hole into the wall, running the linesets and communication wires for a minisplit is beyond most people's comfort level.

Vanna: A Python package to generate SQL using LLMs

Package: https://github.com/vanna-ai/vanna

A paper we're just finished: https://github.com/vanna-ai/vanna/blob/main/papers/ai-sql-ac...

This is a great start! Is there any way I can style the components myself? If so, I'd use it

I found Chroma to be really slow. I recently switched to pgvector and I'm really happy with it.

This is an example of using pgvector that might be relevant for your use case:

SELECT d.id, d.doc, 1 - (embedding <=> (SELECT embedding FROM documentation_embedding WHERE id = 1)) AS similarity FROM documentation_embedding de join documentation d on de.documentation_id = d.id ORDER BY similarity desc;

Reminds me of the time that my mom ordered shoes off of some random FB ad. Her credit card was charged and we waited a few weeks. Tried to go back to the site to see what the order status was and the site didn't exist anymore! We assumed that she just got scammed and did a chargeback to get a refund. A YEAR later the shoes randomly show up in the mail -- original website has long since vanished off the face of the internet.

Looks promising! Any thoughts about making this a VSCode extension?

Same — I just asked ChatGPT to generate Go code to do cosine similarity and I was surprised when the code it generated worked and the worst performance I get is like 10ms

I’m building exactly this! https://vanna.ai

How has your experience been with runtime errors? One of the things I like about Elm is that for pure Elm code, there's very little surface area where a runtime error can occur.

373

Long overdue -- there's a very noticeable difference in runtime errors with Swift libraries vs the rest. This is especially true of AVFoundation, which is an absolute mess of hidden and undocumented state. Swift optionals and enums will go a long way to fixing that.