I think, notably, one of the errors has been to name functions calls "tools"...

I guess that it's only a matter of finetuning.

LLM have lots of experience with bash so I get they figure out how to work with it. They don't have experience with custom tools you provide it.

And also, LLM "tools" as we know it need better design (to show states, dynamic actions).

Given both, AI with the right tools will outperform AI with generic and uncontrolled tool.

I've built a SWE agent too (for fun), check it out => https://github.com/myriade-ai/autocode

Most data analysts are convinced that AI for data analysis is a scam because data is messy and complex. So, nobody even tries to use AI for data analysis.

Even though agentic AI is getting better and better at handling messy and complex data and most startups, relying on simple stacks (database replica) actually have pretty good data.

I have made two ways for people to see the potential of AI for business intelligence:

1. As a docker command

docker run -p 8080:8080 -v $(pwd)/data:/app/data myriadeai/myriade:latest

2. As a live demo - no login required

https://demo.myriade.ai

I would love to hear your feedback.

Comments URL: https://news.ycombinator.com/item?id=44745742

Points: 1

# Comments: 0

You are right that there is lots of way to measure this but quality comments is way harder to judge and we don't have quantity traffic info.

Comments URL: https://news.ycombinator.com/item?id=44625897

Points: 12

# Comments: 8

As a data guy, I’ve always been frustrated by the effort involved in answering simple questions. Not that SQL is hard, but a simple question can easily require 10+ queries and 1h+ of our time.

Which means it’s not worth doing - more often than not.

Myriade is a natural language interface to analyze your database, which allows you to get answers in seconds/minutes, not hours.

Live Demo: https://hn.myriade.ai (explore HN data!)

Repository: https://github.com/myriade-ai/myriade

What it does:

- Ask "Why was there a drop in sales on July 14?” instead of writing JOIN statements

- The data analyst agent will explore, try, correct, adapt, inspect, analyze & synthetize a response… beyond the classical NL2SQL.

- See the agent's full interaction and “take over” at any moment.

- Works with Postgres, MySQL, Snowflake, BigQuery

- Self-hostable (so you can control where your data goes)

Under the hood

- Developed my own agent library for this (https://github.com/BenderV/autochat)

- Uses Anthropic Claude (also compatible with OpenAI and others)

- Postgres/Flask/SQLAlchemy/Vue/Tailwind/Echarts

I’ve used it a lot, and it shows great potential that goes beyond just “getting information”.

Interesting use cases I've seen:

- “What KPIs should I focus on ?” → Select your KPIs

- “Detect quality issues in billing data" → Review data quality

- “Flatten this table” → Prepare a view

- “Make this query run faster: X” → Optimize a query

This clearly is not the one-size-fits-all solution to solve BI, but I find it really useful for day to day.

Try it for yourself - let me know what you think! I’d love to get any feedback, feature requests and (especially) criticism you may have!

Comments URL: https://news.ycombinator.com/item?id=44625870

Points: 2

# Comments: 0

> I wish you luck in refining your differentiation. Can't agree more with you. It's about distribution (which Snowflake/Databricks/... have) or differentiation.

Still, chatting with your data is already working and useful for lots.

> Do you need an expert to verify if the answer from AI is correct?

If the underling data has a quality issue that is not obvious to a human, the AI will miss it too. Otherwise, the AI will correct it for you. But I would argue that it's highly probable that your expert would have missed it too... So, no, it's not a silver bullet yet, and the AI model often lacks enough context that humans have, and the capacity to take a step back.

> How is it time saved refining prompts instead of SQL?

I wouldn't call that "prompting". It's just a chat. I'm at least ~10x faster (for reasonable complex & interesting queries).

Over the past few weeks/months, I’ve been working on Autochat, a lightweight Python library designed to make building AI agents simple and intuitive.

The focus of Autochat is simplicity:

- Extend an AI assistant's functionality by adding Python functions directly, or even class.

- Hide all the complexities/particularities of the providers (openai, anthropic, …)

I’d love your feedback on this. If you’ve got ideas for features, use cases, or critiques, let me know!

Comments URL: https://news.ycombinator.com/item?id=42621962

Points: 4

# Comments: 1

Copying the interface is not what I would call "not reinventing the wheel".

Not that you shouldn't have inspiration, but really I couldn't tell it was not linear...

Futhermore, I think it tell that your product lack AI-first design.

I wanted to share with you Ada, an open-source tool to accelerate data analysis using AI.

In short, it's a (very) lightweight mix between ChatGPT and Metabase.

I have been working around this idea for quite a while, in my spare time. The goal is to make data analysis fun and fast... just like when Tony Stark discover a new element with the help of Jarvis (https://youtu.be/Ddk9ci6geSs?si=Y6sBvZZn8779nPV8&t=32)

I'm sharing it because, while imperfect, I now use it quite regularly in my professional life, and I just learned that a Vietnam company is using it internally. So I figured some of you might be interested, and I would love to have your feedback.

I invite you to install it and try it yourself (https://github.com/BenderV/ada) Or you can try a demo at https://ada.universaldata.io

Best, Ben

PS: Since it’s currently based on OpenAI GPT4, you will need to have an OpenAI API Key.