ReplayAutonomous app testing for teams who ship at AI-speed
Promoted
Bonacci Studio
Agentic, Visual Data Engineering and Data Science Platform
9 followers
Agentic, Visual Data Engineering and Data Science Platform
9 followers
Most AI data tools stop at generating a snippet you still have to run yourself. Bonacci Studio's agent has real tool access: it connects to your database, inspects schemas, writes PySpark or SQL, executes it on Spark, streams logs back, and debugs its own failures. Also included: a visual drag-and-drop DAG builder, Kafka streaming, Apache Camel API integration, RAG knowledge bases, and native MCP support. Bring your own model - Claude, GPT, Gemini, Groq, or fully local via Ollama
The agent debugging its own failures is honestly such a nice touch. One thing though, would be great if you could save and version the generated PySpark or SQL as actual files in a connected Git repo right from the studio, basically turning the agent runs into proper commit history you can review or roll back later.
@adahiber Thanks for the honest feedback! Thrilled to hear you like the self-debugging feature. As per your suggestion - Git repo integration is a such a great idea, we will be adding this feature in couple of days.
honestly the self debugging loop sounds super useful, but it would be great to see some kind of cost or token usage tracker built in so you know what each agent run is actually spending, especially when you switch between claude and the local ollama setup
@selmagentrj4jv Thanks so much for the feedback Selma, cost tracker is up in next release.
honestly the auto-debug loop is wild, like watching it actually catch its own PySpark errors and recover without me babysitting the run. that kind of self-correcting execution is what most "AI data tools" skip entirely.
@barztalnnug Thanks so much for the feedback.
the self-debugging loop where it streams logs back and fixes its own PySpark failures is genuinely impressive, not just a wrapper that pretends to be an agent. most teams would have shipped the snippet generator and called it done.
@kaanzclu Thanks so much for the positive feedback Kaan.