CodeFlowIQ converts repository source files into searchable evidence. The UI does not invent runtime knowledge directly; it asks backend query handlers for indexed evidence and relationship context.
flowchart TD
Repo["Target repository files"]
Ignore[".codeflowiqignore\nand built-in exclusions"]
FileScan["File scanner"]
Language["Language detection"]
CSharp["C# analyzer\nclasses, methods, APIs, DI, calls"]
Sql["SQL analyzer\ntables, procedures, reads/writes"]
JsTs["JS/TS analyzer\nroutes, fetch calls, frontend links"]
Azure["Azure signal analyzer\nblob, storage, queue, functions, config"]
Evidence["Evidence normalization\nfiles, symbols, relationships, metadata"]
Store["Local SQLite evidence store"]
Explorer["Repository Explorer"]
RuntimeMap["Runtime Map"]
FlowChains["End-to-End Flows"]
CSharpTrace["C# Backend Trace"]
Overview["Start Here / Overview"]
Repo --> Ignore --> FileScan --> Language
Language --> CSharp
Language --> Sql
Language --> JsTs
Language --> Azure
CSharp --> Evidence
Sql --> Evidence
JsTs --> Evidence
Azure --> Evidence
Evidence --> Store
Store --> Explorer
Store --> RuntimeMap
Store --> FlowChains
Store --> CSharpTrace
Store --> Overview
- File evidence: path, extension, language, folder, and source preview.
- Symbol evidence: classes, methods, constructors, interfaces, endpoints, procedures, tables, and variables where supported.
- Relationship evidence: method calls, interface implementations, DI registrations, API route handling, SQL reads/writes, stored procedure execution, Azure service usage, and cross-file handoffs.
- Trace evidence: ordered C# backend execution steps, source previews, unresolved boundaries, hidden framework calls, and configured depth limits.
- Quality metadata: exact versus inferred evidence, duplicate occurrence count, source-backed line preview, relationship group, and missing downstream coverage.
- Repository source is read locally and indexed into local storage.
- UI screens query the local API; they should not parse source files directly.
- Repository Explorer is the central evidence workspace.
- Curated screens such as Runtime Map and Overview should act as readable entry points into full evidence.
- Large datasets should be paged, virtualized, grouped, or drilled into instead of rendered all at once.