Skip to content

Releases: wuwei-struct/RepoSense

RepoSense v0.1.0 — Backend transaction, side-effect, and upgrade-context system

16 May 03:26
@wuwei-struct wuwei-struct
v0.1.0
52f8b05

Choose a tag to compare

View all tags
RepoSense v0.1.0 — Backend transaction, side-effect, and upgrade-context system Latest
Latest

RepoSense v0.1.0

RepoSense is an open-source backend transaction, side-effect, and upgrade-context system for AI-assisted projects.

This first public release focuses on one problem: AI can generate the first version of a backend quickly, but developers still need a reliable way to understand what the backend actually does before maintaining, reviewing, or upgrading it.

RepoSense turns a repository into evidence-backed engineering facts, backend side-effect signals, reports, and upgrade-ready context for developers and AI assistants.

What is included

Backend verifier core

  • Extracts Findings, Events, and Evidence from a repository.
  • Builds an event graph for backend-relevant signals.
  • Detects API routes, transaction signals, database operations, queue dispatch/consume signals, and cache operations where supported.
  • Generates a Backend Verifier Report with:
    • API Surface Summary
    • Backend Events Summary
    • Transaction Signals
    • Queue Dispatch Signals
    • Cache Operation Signals
    • Side-effect Map
    • High-risk Findings
    • Evidence Index
    • Limitations

Pattern Engine

  • Includes deterministic, evidence-backed pattern detection.
  • Emits patterns.json and pattern_summary.json.
  • Provides an open Pattern Contract and Rule Authoring Guide for future community rules.

Context Pack

  • Generates a Context Pack for the next AI-assisted maintenance or upgrade step.
  • Packages repository facts, API surface, findings, evidence, quality gate outputs, and run metadata into a reproducible handoff bundle.
  • Designed to help tools like Cursor, Claude Code, Codex, ChatGPT, or human maintainers understand the project before making changes.

Local reports and UI

  • Generates report.html for local inspection.
  • Includes Overview, Events, API Surface, and Findings pages.
  • Includes a local Learn UI for concept/case browsing.
  • Includes release screenshots and a canonical release demo script.

Local AI-derived outputs

The open-source version includes local, evidence-backed, limited AI-derived outputs:

  • ai_summary.md
  • ai_risks/risks.md
  • ai_explain/*/explain.md
  • limited ai ask flow
  • controlled source drilldown based on evidence references

These are local, conservative, and grounded in RepoSense artifacts. They are not a hosted AI Insight platform.

CLI / CI / exports

  • CLI runner
  • Quality Gate
  • Baseline & Diff support
  • SARIF export
  • Run Manifest
  • Backend Verifier Report
  • Context Pack export
  • Release demo script

Quickstart

Run the local demo:

powershell -ExecutionPolicy Bypass -File tools/demo_run.ps1

Or generate the canonical release demo:

powershell -ExecutionPolicy Bypass -File tools/release_demo.ps1

After running the demo, you will see outputs such as:

  • report.html
  • backend_verifier_report.md
  • backend_verifier_report.json
  • patterns.json
  • pattern_summary.json
  • ai_summary.md
  • ai_risks/risks.md
  • ai_explain/*/explain.md
  • exports/context_pack.zip
  • run_manifest.json

Trust boundaries

RepoSense uses conservative, evidence-backed detection.

It does not guarantee full backend correctness.
It does not prove that all transactions are correct.
It does not guarantee safe upgrades.
It does not default to unrestricted whole-repository AI source roaming.

The intended workflow is:

Code -> Facts -> Patterns -> Insights -> Context Pack

Facts first. Source on demand.

Open-source boundary

This repository includes the open-source core:

  • Fact Engine / Scanner
  • Pattern Engine detection logic
  • CLI / CI runner
  • SARIF export
  • Context Pack spec
  • Backend Verifier Report
  • Local report UI
  • Local Learn UI
  • Event / Finding / Evidence schemas
  • Docs and tests

Hosted or commercial layers may later provide:

  • guided explanation enrichment
  • repair suggestions and repair playbooks
  • multi-run history
  • team/project workspace
  • long-term upgrade advisor
  • enterprise reporting

The open-source project is usable on its own. Hosted layers are intended to improve collaboration depth and workflow quality, not to make the local verifier unusable.

Screenshots

Core screenshots are included in the repository under:

docs/assets/screenshots/

The main README embeds the first screenshot set for:

  • Backend Events
  • API Surface
  • Backend Verifier Report

Known limitations

  • Current analysis remains conservative and heuristic in some areas.
  • Side-effect maps are not full runtime call-chain proofs.
  • Some framework-specific patterns may require additional community rules.
  • Learn demo data is currently minimal.
  • AI-derived outputs are local and limited, not a full hosted AI Insight system.

Recommended next steps

  • Add more real-world fixtures.
  • Expand language/framework rules.
  • Improve Learn case density.
  • Add more pattern authoring examples.
  • Continue hardening Context Pack as an AI-assisted maintenance standard.
Assets 2
Loading