Skip to content

roadmap.md

Latest commit

 

History

History
106 lines (81 loc) · 4.95 KB

roadmap.md

File metadata and controls

106 lines (81 loc) · 4.95 KB
Raw

RepoAtlas Roadmap

This document contains active future work only. Shipped changes belong in CHANGELOG.md, current behavior in spec.md, and durable decisions in adr/.

Product direction

RepoAtlas produces deterministic, evidence-backed Candidate Briefs from static repository signals. The product should deepen evidence, analyzer accuracy, and platform safety without executing repository code or introducing AI-generated analysis.

Now

Analyzer trust

Priority Goal Key files
Evaluation gold set Expand human-labeled fixtures and measure entrypoint, relationship, command, and ranking agreement eval/, src/analyzer/eval/
Python and Java semantic depth Narrow the evidence-depth gap with the TypeScript/JavaScript Compiler API pack src/analyzer/packs/
Small-repository calibration Stabilize structural-hotspot rankings for repositories with very few scored files src/analyzer/scoring/
End-to-end cancellation Propagate one cancellation signal through acquisition, extraction, indexing, and persistence src/app/api/analyze/route.ts, src/lib/ingest.ts, src/analyzer/index.ts

Evidence and output integrity

Priority Goal Key files
Interview-guide e2e realignment Rewrite guide/interview-preparation e2e expectations against the current page structure (headings removed during landing redesign) e2e/interview-preparation.spec.ts, e2e/landing-page-starts.spec.ts

Shipped since this list was last curated (verified 2026-08-22, see docs/audits/2026-08-22-adversarial-audit.md): bounded source-file snippets, Markdown export escaping, technical-decision evidence, and cross-language complexity decision-point alignment. Small-repository calibration ships via blendedMetricRank shrinkage in src/analyzer/scoring/metrics.ts.

Verification and frontend scale

Priority Goal
Component coverage Add behavior-focused tests before raising global coverage thresholds
Snapshot stability Normalize volatile fields in Candidate Brief golden tests
Large-report budgets Establish measured rendering and export budgets before increasing report size
Client bundle budget Lazy-load export dependencies and prevent unmeasured homepage growth
Operational metrics Add privacy-safe service metrics without logging repository content

Next

Analyzer depth

  • Build a combined reduced architecture for repositories containing multiple supported languages.
  • Improve framework and test-tool detection using manifests and lockfiles.
  • Extend branch-aligned history with bounded co-change evidence.
  • Define compatible migrations for older stored reports beyond validate-or-reject.

Storage and export

  • Verify filesystem and Blob retention behavior against the same adapter tests.
  • Review private Blob access for stored reports and share-token records.
  • Evaluate a structured vector PDF path after current raster export budgets are measured.

Later / exploratory

Area Direction
Progressive results Stream real analyzer checkpoints only after cancellation and persistence semantics are defined
Private repositories Require user-scoped OAuth and a new access model; never reuse a server-owned token
Folder-map filtering Hide generated and binary paths in the UI without hiding analysis skips
Attribute overrides Consider .gitattributes only after deterministic pack-selection rules are defined

Explicit non-goals

  • LLM-generated analysis or external AI calls
  • Executing or profiling repository code
  • Full SAST or vulnerability scanning
  • Public report deletion without an ownership model
  • Caller-controlled filesystem paths on the network API
  • Claims of confirmed bugs, correctness, production readiness, business purpose, or dynamic runtime behavior

Work-selection guidance

  1. Start with the first incomplete item in Now unless a verified production incident changes priority.
  2. Check CHANGELOG.md and current source before assuming work is unimplemented.
  3. Update spec.md with behavioral, limit, schema, storage, or API changes.
  4. Add or amend an ADR for durable security or architecture decisions.
  5. Keep one boundary family or behavior change per pull request and run the narrowest relevant tests plus required repository checks.

Related documents

Document Responsibility
README.md Product overview, quick start, limits, and developer entry points
CHANGELOG.md Shipped release history
spec.md Current product and behavioral contract
guardrails.md Non-negotiable implementation constraints
SECURITY.md Vulnerability reporting and security guarantees
adr/ Durable architecture decisions and rationale

Last updated: 2026-08-22