It’s a GitHub App that listens to webhook events and helps automate common repository tasks like PR review, issue triage, and repo health checks. I tried to focus more on architecture reliability — guardrails, async webhook handling, and validation of AI responses. Would love feedback from other developers on the approach. Repo: https://github.com/Shweta-Mishra-ai/github-autopilot

Hey everyone, glad to be here. I'm Winner, founder of Kybernis. I'm currently working on infrastructure that ensures AI agents execute real-world actions safely when interacting with production systems (APIs, databases, payments, infrastructure, etc.). Recently I've been experimenting a lot with Google ADK and agent workflows, especially around what happens once agents move beyond reasoning and start triggering real system mutations. Really excited to learn from others building production-grade multi-agent systems, and happy to exchange ideas or collaborate. Looking forward to the discussions here

Hey, guys! I need your help. 😭 I’m a Machine Learning intern, and I built an application in Vertex AI Studio in one GCP project. The problem is that we couldn’t connect it to BigQuery because of a service account error (everyone on my team tried and we all failed). My manager asked me to recreate the application in Vertex AI Studio, but in a different project. I’m doing that now, but even using the exact same prompts from the original project, the new version isn’t performing nearly as well. I know some variation is expected with generative AI, but the original version was so good that I don’t want to give up on it. I downloaded the ZIP file with the code from the original application — is there a way to upload/import what I already have into the new project so I can keep the same behavior?