From Ticket to First Commit: Building a Graph-Powered Contribution Strategist with GitLab Orbit

2026-06-22 · Source: Artificial Intelligence on Medium · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, long

Summary

Orbit Compass, a GitLab Duo Custom Agent developed during the GitLab Transcend Hackathon, addresses a common developer onboarding challenge by transforming issue tickets into actionable contribution roadmaps. The author, frustrated by scattered context across GitLab's tools, found traditional search and Retrieval-Augmented Generation (RAG) inadequate for understanding the "relationships" between development entities. This led to leveraging GitLab Orbit, a knowledge graph modeling issues, merge requests, files, and contributors. Initial LLM attempts failed due to a lack of graph schema understanding. The project then shifted to a deterministic investigation workflow, using Orbit's schema to identify historical context, relevant merge requests, modified files, dependencies, and active contributors. Key design principles included providing explicit evidence for recommendations, replacing numerical scores with qualitative categories (High, Medium, Low), generating fully qualified URLs, and prioritizing clear output. Testing showed Orbit Compass generates a comprehensive contribution roadmap, explaining the "why" behind code context.

Key takeaway

For AI Engineers building developer productivity tools, consider integrating knowledge graphs like GitLab Orbit to move beyond document retrieval. Your solutions should focus on reconstructing context by modeling "relationships" between development entities, not just finding information. Prioritize transparent, evidence-backed recommendations over opaque scoring, and ensure output clarity with fully qualified links to build user trust and provide actionable "next-step" guidance for contributors.

Key insights

Knowledge graphs excel at contextualizing software development relationships beyond simple search.

Principles

• Software development context relies on entity relationships.
• AI recommendations need explicit, verifiable evidence.
• Prioritize clarity and actionable "next-step" information.

Method

Orbit Compass uses a deterministic workflow: identify historical MRs, examine modified files, investigate dependencies, identify active contributors, and check for ongoing work.

In practice

• Model software entities as a knowledge graph.
• Expose reasoning paths for AI-generated recommendations.
• Replace numerical confidence scores with qualitative evidence.

Topics

• GitLab Orbit
• Knowledge Graphs
• Developer Productivity
• AI Agents
• Contributor Onboarding
• Explainable AI

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Agentic coding is only as good as its context — Agentic coding's effectiveness hinges on robust lifecycle context, enhancing speed and review processes.
• Connecting the Dots with Context Graphs — Stephen Chin, Neo4j — Context graphs integrate LLMs with knowledge graphs to provide grounded, explainable, and auditable AI-driven decision-making by connecting disparate enterprise data.
• GitLab Flex: Commit once, reshape your seats and AI spend — GitLab Flex offers a single, adaptable commitment for managing seats and AI spend in unpredictable development environments.
• abhigyanpatwari / GitNexus — GitNexus builds a knowledge graph of codebases to provide AI agents with deep, precomputed architectural context.
• AI Dev 26 x SF | Brandon Waselnuk: Building the Context Engine AI Agents Need — AI agents require a dynamic context engine to achieve organizational intent and avoid costly errors.
• Noticed a big spike on OpenRouter’s Cloud Agents (Coding) leaderboard this week — The gitlawb platform's 582 billion token usage on OpenRouter signals significant AI agent activity in decentralized code collaboration.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence on Medium.