CodeRadius Documentation

CodeRadius builds a live, queryable graph of your services, APIs, databases, and teams from your actual code. It uses that graph to catch cross-repo breakage before it ships: enforce architectural policy in CI, evaluate blast radius on a proposed change, and give AI coding agents real architectural context instead of a guess.

User Guide

Core Workflows

• Use Cases: Three concrete scenarios where CodeRadius prevents production incidents that linters and code search cannot catch.
• Governance: Enforce architectural standards across your fleet in CI and in dashboards. Declarative YAML policies evaluated against the live graph.
• Impact Evaluation: Predict breaking changes and blast radius before merge. Like terraform plan for architecture.
• Vulnerability Scanning: Fleet-wide CVE intelligence on every analysis run. Know which services run a vulnerable version, who owns them, and what ships the fix.
• MCP Server: Give AI coding agents live architectural context so they stop breaking downstream systems.
• Grounding & Trust Tiers: What the colored dots on the dashboard mean, how to triage flagged entities, and how to filter by trust tier from the CLI.

Setup & Configuration

• Introduction: What CodeRadius is, the problem it solves, and how to get started in under 10 minutes.
• CLI Commands: Complete reference of all available terminal commands and their flags.
• coderadius.yaml: Teach the AI about proprietary SDKs and map database identities without writing code.
• .crignore: Control which files are excluded from analysis during ingestion.
• Supported Frameworks: Languages, frameworks, protocols, databases, and message brokers that CodeRadius analyzes out-of-the-box.

Explore

• Architecture Dashboard: Self-contained HTML report for architectural health, dependency governance, and SPOF analysis.
• System Registry: Auto-generated service catalog of every repository, service, and team.
• SPOFs & Data Gravity: Single Points of Failure detection and architectural bottleneck ranking.
• Agent Harness: AI tooling adoption metrics, maturity matrix, and context gap analysis.
• Blast Radius Scoring: How CodeRadius classifies the risk of changing or breaking an architectural node using the Downstream Gravity Score and Impact Tiers (T0-T4).
• Context Engineering: How to use CodeRadius to build and distribute organizational AI coding standards through context engineering.

Architecture Deep-Dives

• System Architecture: How CodeRadius constructs the architectural graph, from ingestion pipeline to graph storage and query layer.
• Code Ingestion Pipeline: How the pipeline transforms a repository into a knowledge graph, filtering out as much noise as possible before invoking any LLM.
• Grounding, Evidence, Quality: How CodeRadius attributes every node and edge in the graph to its origin, supporting evidence, and trust tier.
• Impact Explorer Scoring System: Mathematical specification of the Downstream Gravity Score that feeds the T0-T4 impact tier classifier.
• API Endpoint Dedup & Cross-Service Matchmaking: How CodeRadius prevents duplicate APIEndpoint nodes when the same logical route is described by multiple producers, and joins consumer->provider edges across services.
• Service Topology Architecture: The code-first identity model: filesystem autodiscovery decides what services exist, catalogs decide who owns them, and the topology resolver welds the two together.
• Catalog Drift: Grounded-Identity Reconciliation: How cr drift reconciles a catalog's declared facts against the code-observed graph; drift is only asserted between facts that resolve to the same real graph node.
• Data Domain Model: The logical/physical split for datastore topology: one logical identity, N physical surfaces, one per deployment environment.
• Messaging Domain Model: Three-layer ontology for message broker topology, scaling from a single cluster to multi-region, multi-tenant enterprise deployments.
• Graph URN Taxonomy: The canonical URN templates that key the CodeRadius graph; resources that should converge to the same node must produce the same URN.
• Library vs Package: Component Ontology: Disambiguates workspace-internal code (:Library) from declared dependencies (:Package) in the graph domain.
• Contrib Plugin System & Crossplane PubSub Extraction: Extending the structural extraction layer with domain-specific plugins that extract infrastructure topology from Helm chart Crossplane CRD templates.
• Incremental Cache Versioning: The engine-versioned Merkle tree: how CLI upgrades with improved detection logic invalidate only the affected slice of the graph cache, without full re-ingestion.
• Graph Database Optimizations: Traversal vs label scans: how the fundamental graph queries (like the Merkle index query) avoid performance bottlenecks at multi-tenant scale.
• Team Mapping & Repository Discovery: Internal logic for team detection, organization mapping, and repository path resolution within the architectural graph.