dltHub, maker of dlt "data load tool" is releasing a LLM-native workflow for creating pipelines. This is OSS and you can try it yourself (15-30min is a good time)
Is this BS or does it work? A: it works decently well, significantly better than starting from scratch. We held back this feature for a couple months to improve it until we were happy to offer it.
How is this better than from scratch?
- We do not generate code, we generate python config dictionary, making review easy. - We offer a debug app to validate correctness, - finally, the LLM context ensures low hallucination rate for higher success rates.
Feedback welcome! Missing a scaffold? let me know and we will add it
No comments yet.
