Show HN: AI SQL Copilot LogicLoop – AI to Generate, Optimize and Debug SQL (logicloop.com)

Yeah, it's actually pretty good at fixing existing queries if you give it your schema. I'd say on queries from 10-50 lines of SQL it works 9 out of 10 times.

Agreed, that's always a risk with making data too easy to access, its context doesn't always come with it. While we initially thought it'd be most valuable for new SQL users, a lot of existing SQL experts actually find that it gives them a huge productivity boost.

Thanks! We're starting with just the schema, good clarification.

Good question. A few different ways:

* Model Selection: We pre-select the best models for SQL generation, so you don't have to A/B test and figure it out. As the number of models available increases, this can take a lot of time if you want to do it yourself.

* Prompt Engineering: Similarly, we've had to fiddle around with a lot of different prompts to deal with common issues like hallucinations (and are constantly updating them, fixing bugs, telling the model what it can't ignore etc.) This also takes a lot of time, patience and skill.

* Directly on your data schema: You'd have to copy-paste your schema every single time you wanted to generate or modify a SQL query, which can be cumbersome. You connect your data sources once with LogicLoop, and it gets auto-included correctly for you.

Hope that helps.

Right, here's some ways we're different -

* On your data schema: A lot of the ones we've seen recently don't actually generate these queries on your schema, so you still have to do a lot of manipulation or calculation to get to SQL that just produces results.

* A better experience in productionizing SQL: Like code, writing production SQL is rarely one-shot. LogicLoop comes with a suite of AI SQL Copilot tools, like a fixer, optimizer, editor etc. that allow you to continue to iterate on your SQL logic in an integrated environment.

* Go beyond SQL: Generating query results might often be the first step of a business process. With LogicLoop, you can actually visualize results, send Slack alerts, create tickets, run automations etc. on a schedule.

Hope that helps clarify your question.

Haha, almost :P Some problems can't be solved. For everything else, there's LogicLoop.

Thank you, let us know what you think when you try it out.

It varies. Theoretically, someone who doesn't know SQL should be able to generate it from this, but you'd still have to get someone to connect your datasource, and have a general sense for what you want to do.

I'd say it's not optimized for the high-level executive or entry-level operations person. It's more geared for someone that is data-aware but wouldn't consider themselves confident in SQL, like a junior data analyst profile.

I understand your skepticism, here's how we think about things -

* We only send data schema to underlying models: No data is actually sent there, but yes when you connect your database, we run the generated SQL on your database and store results to show those to you. They get regularly deleted though.

* We take security seriously: We're SOC2 Type 2 compliant and publish our security practices online, have a DPA, etc.

* Companies value convenience: We've often seen that companies _prefer_ cloud deploys so they don't have to DevOps themselves. Can be a matter of personal preference.

Hope that helps clarify.

We use a combination of APIs from existing LLM providers, and do some serious prompt engineering to get the best from them. We're starting to train models on more SQL-specific prompts now.

On simpler multi-table joins we've been able to product good results, and we've done a lot of prompt engineering to make sure it takes the schema very seriously so that prevents hallucinations too. We're always finding new edge cases and fixing those as we go.

Yes, we can actually connect to third-party apps via APIs https://docs.logicloop.com/data-sources/supported-data-sourc...

Thanks for sharing your experience. By organizing and storing queries, we hope to be able to improve these suggestions as well behind the scenes.

The submitter is active in the comments and has a long history of submissions related to similar ideas. That seems to indicate to me that this is a real project and an iteration on a premise they've been thinking about for a long time. I do notice even their comments follow a suspiciously GPT-like template with three bullet points. But even if they're using GPT to generate the comments and blog posts, is that really so wrong? Maybe English is their second language and GPT helps them to more effectively communicate their ideas.

I haven't investigated the blog or pricing, but I'm inclined to give them the benefit of the doubt. I don't think this post should be flagged, so I've vouched for it (although I think my vouching powers were nerfed a long time ago, so hopefully someone else will vouch for it too).

EDIT: I actually don't even see the vouch button. Paging @dang

Yup, our bet is that people are going to want better integration into data sources and actions resulting from their queries, and that a lot of business value will come from that.

According to docs it only takes in your data schema and outputs the SQL query to run, so the AI model doesn't actually get information about your data, just the schema. Otherwise data is handled according to SOC II principles similar to any other BI tools.

That's a great idea for future improvements, thank you for trying it out!

Thanks for sharing how you keep the discussion quality high here. We responded to some thought-provoking questions on this thread about differentiation vs ChatGPT, other SQL Copilots, edge cases like more complicated queries etc. that we believe other HN users and makers in this space will benefit from. As I responded in @chatmasta's vouch (thank you!), we're data and software engineering experts, and happy to provide our original perspective on any other questions you have for us. Thanks for your consideration!