Free for open source projects and schools. AI Council costs nothing for OSS maintainers and educational institutions. Apply →

Home / Docs / Analysis

Project analysis

Analysis builds the code graph, metrics and architecture view, binds knowledge and conventions to code, and fills the search index. Until the first analysis, the project screens are empty.

Running analysis

  1. Open the project and click Analyze.
  2. Choose a preset: Quick (L1–L3), Standard or Full. Use Full for a project's first analysis. An external analyzer such as SonarQube is turned on with a toggle in the Analysis engines block. The built-in engine always runs.
  3. Click Start Analysis.

Stages

GroupWhat happens
PrepareLoading code, parsing sources, walking files
CodeStructure (L1), dependency graph (L2), metrics (L3), data model, conventions and style (L6)
ArchitecturePatterns (L5), groups and layers (L4), domain (L7), co-changed files, code smells (L8), quality scores (L9), diagrams (L10)
KnowledgeRebinding knowledge and conventions to the parsed code
ScannerRunning the external analyzer, if enabled
SearchSemantic search index over code and knowledge. On a large project this is the longest stage

Leaving the page doesn't stop the run. Cancel stops the run, and results already written are kept. Include the run's Job ID when you report a problem.

What to check after a run

  • Inventory counters PACKAGES, CLASSES, FUNCTIONS, FILES on the Dashboard. Noticeably lower numbers mean only part of the code was parsed.
  • Warnings in Recent Analysis. The warnings label expands to show which layer wasn't written and why.
  • Preset. If it was a reduced preset, the search index and knowledge rebinding didn't run, so AI Chat and knowledge delivery rely on the previous run.

Next: Conventions and knowledge

Made on
Tilda