01
Repo Map
How is the repository organized?
A bounded directory hierarchy across analyzed files and modules.
Observed structure
FAI product - Repository evidence
FAI Repo Intelligence
Understand a repository before you adopt, modify, or connect it.
Deterministic, no-execution analysis turns bounded source evidence into five connected repository views. Observed facts, inferred lifecycle, and projected flow remain visibly distinct.
Cloud contract: GitHub + ZIPJSON + MarkdownNo source bodies in results
Repository console
5 FACETS - NO EXECUTIONAuthenticated workbench
Analyze all five facets here
Submit a public GitHub repository or bounded ZIP, follow the asynchronous job, inspect analyzer-returned nodes and edges, and export the schema-validated result without leaving FrootAI.
Loading your account…
Five facets, one evidence model
Different questions need different repository views
Every facet contains nodes, analyzer-returned edges, hubs, confidence, evidence, and limitations. Nothing is upgraded from inference to fact for presentation.
02
Repo Graph
How are local files related?
File hierarchy plus static, line-evidenced local import relationships.
Not runtime semantics
03
Repo Flow
How does source move toward delivery?
Observed build, test, and delivery artifacts arranged into an inferred lifecycle.
Not an execution trace
04
Code Flow
Where can execution enter?
Conventional entrypoints and external service references with confidence labels.
Projected edges stay labeled
05
Agent Flow
How is the Agentic OS composed?
Agents, instructions, prompts, skills, workflows, and evaluation topology.
Sparse when artifacts are absent
Cloud analysis pipeline
From repository URL to portable evidence
The cloud runtime is asynchronous, tenant isolated, and designed to return the five-facet JSON/Markdown contract. VS Code remains a separate local read-only workflow for repository signals and Play recommendations.
- 01ValidateAccept a public GitHub URL or a bounded ZIP archive.
- 02PinResolve an immutable source revision before analysis.
- 03CollectRead bounded files without executing repository code.
- 04AnalyzeBuild all five deterministic facets with confidence and evidence.
- 05DeliverReturn schema-validated JSON and Markdown without source bodies.
Truth boundaries are part of the product
Observed
Directory structure, files, line-evidenced imports, and artifact presence.
Inferred
Lifecycle and coordination relationships supported by conventional artifacts.
Projected
Architectural hypotheses such as external input routes; never presented as runtime traces.
Production-safe result contract
- Immutable revision recorded
- No repository code executed
- No source file bodies returned
- Paths and truncated hashes as provenance
- Bounded file and duration budgets
- Schema-validated JSON and Markdown
Azure-backed runtime
Built for bounded asynchronous analysis
The API and worker run on Azure Container Apps with managed identity, private data services, immutable container digests, OpenTelemetry, and autoscaling. The authenticated browser workbench uses this runtime through the same-origin FrootAI facade.
Authenticated workbench enabled
API
JWT validation, idempotency, tenant ownership, quota, job status, and cancellation.
Queue + worker
Service Bus decouples bounded collection and deterministic analysis from the browser request.
Private state
Cosmos DB tracks jobs; Blob Storage retains private inputs and result artifacts under policy.
Start with a repository
Inspect first. Change second.
Run the complete five-facet analysis in the authenticated workbench, or open the repository in VS Code for local signals and Play recommendations.