正在加载视频...
视频加载失败
Introducing SQrL - a Text2SQL Reasoning Agent built with GPT4.1 It “thinks” before running SQL and "analyzes" the results of its queries to deliver the best possible answer to your questions. The video is 1x speed. Give it a try and let me know if you like it. Code... show more
11 条评论

Code:

Testing Salesforce presents unique challenges due to its complexity, scalability and customizability. RTTS can plan, design & automate a successful testing process for you.

I’ve got very good results with quite basic models in the past. What am I missing?

Not missing anything, I generally found that the response quality improves when the model validates the response: as in does this data make sense. Eg: see the 3rd question in the video (I think) - what’s the longest career in F1. The data has quality issues, and returns Piquet (sr and jr). This is technically correct based on the data but the answer is wrong. If the model can analyze: does a 31 year racing career sound plausible - then it figures out that these are 2 drives racing under the same name/tag. And then returns the correct answer. Don’t think you’re missing anything. Just sharing the approach we’re taking.

Q: why would you build a reasoning agent with a non reasoning model by shoving the think tool down its throat vs. using a reasoning model? cost?

Interesting tone. This research should help:

Interesting. Do you have a connector to databricks?

would it take a sqlalchemy url?

I looked into Vanna recently, but problem is: there are a lot of SQL dialects out there, many models may favor SQLite/Postgres, but when you tried go get a schema from SQL Server with it, they just don't get it right and you end up coding it.

sir, can i apply this also to cypher

great one ... added to VibeApp Store. I hope that you vibe coded Ashpreet 😉
