Oracle Intelligence Engine

An AI reasoning engine that traverses the GitLab Orbit Knowledge Graph to generate risk-assessed implementation blueprints.

Comment

🔮 Oracle Intelligence Engine

AI-Powered Software Change Intelligence built on GitLab Orbit Knowledge Graph and GitLab Duo Agent Platform

📖 Overview

Oracle Intelligence Engine is a graph-native software intelligence platform designed to help engineering teams understand the impact of changes before they are implemented.

Modern software systems consist of interconnected repositories, services, merge requests, contributors, work items, and dependencies. Understanding how a proposed change propagates through these relationships is often difficult and time-consuming.

Oracle Intelligence Engine leverages GitLab Orbit Knowledge Graph data and AI-powered reasoning to automatically analyze proposed changes, identify subject matter experts, assess risk, discover dependencies, and generate implementation blueprints.

Built during the GitLab Transcend Hackathon, Oracle demonstrates how graph intelligence and agentic workflows can transform software delivery decision-making.

🚨 Problem Statement

Engineering teams frequently face questions such as:

  • Which repositories will this change affect?
  • Who should review this merge request?
  • What dependencies could break?
  • How risky is this migration?
  • What implementation plan should we follow?

Answering these questions usually requires:

  • Manual repository exploration
  • Searching historical merge requests
  • Consulting multiple teams
  • Reviewing documentation
  • Relying on tribal knowledge

Oracle Intelligence Engine automates this process.

💡 Solution

Oracle combines the following to generate actionable engineering intelligence before code changes are made:

Component Role
GitLab Orbit Knowledge Graph Source of truth for org knowledge
GitLab Duo Agents Agentic orchestration
GitLab Flows Multi-stage coordination
FastAPI Intelligence Services Backend processing
Graph Traversal Logic Relationship mapping
AI Reasoning Evidence-based inference

✨ Key Features

🔍 Project Discovery

Discover repositories, services, and systems related to a technology, domain, or feature.

🧑‍💻 Expert Discovery

Identify engineers with the strongest historical expertise in affected code areas.

Oracle analyzes:

  • Contribution history
  • Repository participation
  • Historical ownership
  • Merge request activity

🗺️ Dependency Discovery

Map project relationships and identify connected systems that may be affected by proposed changes.

📊 Change Impact Analysis

Analyze proposed modifications and estimate downstream effects across repositories and services.

👥 Reviewer Recommendation

Recommend reviewers based on expertise, contribution history, and project involvement.

⚡ Risk Assessment

Generate risk scores that estimate:

  • Change complexity
  • Dependency exposure
  • Potential blast radius
  • Review sensitivity

📋 Blueprint Generation

Generate structured implementation plans containing:

  • Impacted systems
  • Risk assessment
  • Recommended reviewers
  • Dependency analysis
  • Execution strategy
  • Architectural guidance

🏗️ Architecture

Oracle Intelligence Engine follows a layered architecture with four major components:

1. GitLab Duo Platform

Provides the user-facing Agent and Flow orchestration layer.

2. GitLab Orbit Knowledge Graph

Acts as the source of truth for organizational knowledge including Projects, Merge Requests, Issues, and Contributors.

3. Oracle Intelligence Engine

Processes graph evidence using specialized reasoning modules.

4. Actionable Intelligence Layer

Produces implementation-ready outputs for engineering teams.

🔀 Architecture Diagram

🔗 Interactive Mermaid Diagram:

⚙️ Technical Architecture

🔗 Backend Layer

Built using FastAPI and Python.

API Endpoints

Endpoint Purpose
/api/experts Find subject matter experts
/api/project-discovery Discover related repositories
/api/dependencies Map dependency structures
/api/similarity Identify similar historical work
/api/blueprint Generate implementation blueprints
/api/project-risk Calculate project risk scores
/api/reviewers Recommend reviewers
/api/impact Estimate downstream impact
/api/recommend-reviewers Reviewer recommendations
/api/change-impact File-level impact analysis

🔌 Orbit Integration Layer

OrbitClient

Responsible for:

  • Orbit API communication
  • Authentication
  • Query execution
  • Data retrieval

GraphExplorer

Provides graph traversal utilities used throughout the intelligence modules.

🧠 Intelligence Modules

Module Function
ExpertFinder Find subject matter experts
ProjectDiscovery Discover related repositories and projects
DependencyDiscovery Map dependency structures
SimilaritySearcher Identify similar work and historical patterns
ProjectRiskScorer Calculate project and change risk
ImpactAnalyzer Estimate downstream impact
ReviewRecommender Recommend reviewers
ChangeImpactAnalyzer Analyze file-level impact
BlueprintGenerator Aggregate intelligence into actionable plans

🤖 GitLab Duo Integration

🎯 Oracle Agent

A custom GitLab Duo Agent responsible for interacting with users and orchestrating analysis workflows.

🔄 Oracle Intelligence Flow

A GitLab Flow that coordinates intelligence generation across multiple reasoning stages.

📦 GitLab-Native Repository

The complete solution is managed through GitLab repositories and follows GitLab Duo Agent Platform conventions.

🧩 Challenges

  • Understanding GitLab Duo Agent architecture
  • Working with graph-native data models
  • Designing reusable intelligence engines
  • Mapping graph evidence into actionable insights
  • Creating Flow-based orchestration
  • Managing GitLab repository permissions and onboarding constraints

🏆 Accomplishments

  • ✅ Graph-native reasoning instead of keyword search
  • ✅ Modular intelligence architecture
  • ✅ Automated reviewer recommendations
  • ✅ Change impact assessment
  • ✅ Blueprint generation engine
  • ✅ GitLab Duo Agent integration
  • ✅ GitLab Flow integration
  • ✅ Orbit Knowledge Graph utilization

💎 What We Learned

The most valuable AI systems are not those with the largest models. The most valuable AI systems are those grounded in accurate organizational context.

By combining AI reasoning with GitLab Orbit Knowledge Graph relationships, Oracle produces recommendations that are evidence-based, explainable, and actionable.

🚀 Future Roadmap

🔌 MCP Server Integration

Expose Oracle capabilities as callable MCP tools.

🔄 Live GitLab Duo Workflows

Participate directly in Merge Request workflows.

💥 Predictive Blast Radius Modeling

Visualize cascading impacts before implementation.

📝 Automated Work Item Generation

Convert blueprints into GitLab Issues, Tasks, and Epics.

🔁 Continuous Repository Intelligence

Generate proactive recommendations as repositories evolve.

🌐 Organization-Wide Knowledge Graph Intelligence

Expand Oracle from project-level analysis to enterprise architecture intelligence.

🛠️ Technology Stack

Python · FastAPI · GitLab Orbit APIs · GitLab Duo Agent Platform · GitLab Flows · Pydantic · Uvicorn · REST APIs · Graph Traversal Logic · AI-Powered Reasoning

Built With

Share this project:

Updates