Skip to content

## Repository description (recommended) AI-powered automated debugging with iterative fix loops. GitHub/CI integration. Supabase + Next.js. Planning repo. Optional shorter fallback (if the UI truncates): AI debugging with iterative fix loops. GitHub/CI + Supabase + Next.js. Planning repo.

Notifications You must be signed in to change notification settings

TheMailmans/FixLoopDev

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 

Repository files navigation

FixLoopDev (Planning Repository)

This repository is for development planning and documentation of FixLoop — an AI-powered automated debugging service that analyzes codebases, proposes patches, runs tests, and iterates until issues are resolved. This content will be refined as the roadmap is finalized and implementation proceeds.

Project Overview

FixLoop provides a web-based service that:

  • Connects to GitHub repositories or accepts direct code uploads
  • Indexes and analyzes project code to understand structure and context
  • Runs an automated debug loop: analyze → fix → test → iterate (with iteration cap)
  • Creates branches/PRs with proposed fixes and attaches logs/artifacts
  • Operates with secure, isolated execution and clear auditability

Core Features

  • AI Integration
    • Primary model: GPT‑5 (with web search tool for up-to-date information)
    • Optional fallbacks/alternates for resilience
  • Code Analysis / Indexing
    • Language-aware chunking (tree-sitter)
    • Vector search (pgvector) for retrieval-augmented reasoning
  • Automated Debug Loop
    • Analyze failure + context
    • Generate minimal patch with guardrails (patch size, file whitelist)
    • Apply patch on a branch and run tests/build
    • Evaluate results and iterate until success or cap
  • GitHub Integration
    • GitHub App for repo access, branches, PRs, webhooks
    • GitHub Actions used as the default secure execution environment
  • Direct Uploads
    • Users can upload source to run in a sandboxed container
    • Enforced timeouts and resource limits

Technical Architecture (High Level)

  • Frontend: Next.js 15 + TypeScript on Vercel
  • Backend/API: Next.js API routes (or tRPC) + Inngest/Trigger.dev for background orchestration
  • Data: Supabase (Auth + Postgres + Storage + pgvector). Uses existing FixLoop project (not ThinkBack)
  • Execution:
    • GitHub repos → GitHub Actions workflow template for install/test/build
    • Direct uploads → Isolated runner VM/containers (Node/Python first)
  • AI Tools:
    • webSearch tool (e.g., Tavily) for current information
    • fileOps for creating diffs/patches (git-based)
    • tests tool delegates to Actions/runner and aggregates results
  • Observability: Sentry (errors), PostHog (product), Logflare/Axiom (logs)
  • Caching: Upstash Redis for short-lived orchestration state

Recommended Tech Stack

  • Next.js 15 + TypeScript, Tailwind, shadcn/ui
  • Supabase (Auth + Postgres + Storage + pgvector)
  • Stripe (subscriptions + metered usage + optional success bounty top-ups)
  • Inngest or Trigger.dev for orchestrating debug loops
  • GitHub App + Actions for repo-based execution; container runner for uploads
  • Tavily (or SerpAPI) for web search

Automated Debug Loop Design

  1. Inputs: repo ref or uploaded project, commands (install/test/build), iteration cap
  2. Analyze: failure logs, code context via RAG, user repro steps
  3. Propose Fix: minimal, safe patch with explanation; limit patch size
  4. Apply Patch: create branch/commit (or patch local upload workspace)
  5. Test: use GitHub Actions (CI) or isolated container; capture artifacts/logs
  6. Evaluate: success if tests pass/target failure resolved; else iterate
  7. Stop Conditions: success, iteration cap, catastrophic regression

Outputs: PR link (repos) or downloadable patch (uploads), final run report (iterations, diffs, timings, costs).

Data Model (Initial)

  • users(id, email, role)
  • projects(id, owner_user_id, name)
  • repos(id, project_id, provider, provider_repo_id, installation_id)
  • runs(id, project_id, repo_id/upload_id, status, iteration_cap, commands, created_at/completed_at)
  • iterations(id, run_id, number, plan, patch_summary, outcome, passed)
  • patches(id, iteration_id, diff_text, branch_name, pr_number)
  • artifacts(id, run_id, type, url)
  • usage(id, user_id, tokens_in/out, compute_seconds, run_id)
  • plans(id, name, price, included_credits, overage_per_unit)

Monetization Strategy

  • Hybrid model (recommended):
    • Subscription tiers with included monthly credits (tokens + compute + runs)
    • Metered overage pricing beyond included credits
    • Optional success bounty on specific issues (charged when CI passes)
  • Example tiers (tune later):
    • Hobby: $29/mo — 2 runs/day, 2h compute, 1 repo
    • Pro: $99/mo — 10 runs/day, 10h compute, 5 repos, priority queue
    • Team: $299/mo — pooled 40h compute, 25 repos, SSO beta
  • Cost control: iteration caps, patch size limits, leverage GitHub Actions minutes, cache embeddings/deps

Development Roadmap & Timeline (Aggressive MVP)

  • Day 1–2: Supabase wiring (FixLoop DB), Next.js scaffold, Auth; Stripe basic; GitHub App + install flow, repo connect UI
  • Day 3–5: Runs/iterations schema + dashboard; Actions workflow template; indexer (JS/TS, Python) + pgvector; orchestrator skeleton
  • Day 6–8: Agent loop (analyze→patch→branch→test→evaluate); logs/artifacts UI; web search tool; direct upload with single runner
  • Day 9–11: Billing (metered + quotas); admin metrics, error reporting; UX polish (timeline, PR links, final report)
  • Day 12–14: Guardrails/hardening; security pass; dogfooding on sample repos; private beta on Vercel

MVP acceptance: connect repo → run → AI opens PR with passing fix for a curated bug in ≤10 iterations; direct upload works for Node/Python sample projects; billing/limits enforced; artifacts visible.

Testing & Validation

  • Layered validation: CI test suite (primary), static checks (lint/typecheck/build), optional sanity command
  • Uploads require install/test commands (fixloop.yml optional)
  • Regressions: track failing tests pre/post; auto-revert catastrophic changes
  • Observability: record test outputs, coverage summaries, timing, token usage per iteration

Risks & Mitigations

  • Untrusted code execution → GitHub Actions / isolated containers only; strict time/CPU/memory limits
  • Cost overruns → iteration caps, quotas, caching, usage tracking
  • Sparse tests → allow user-provided commands; provide test authoring suggestions
  • Multi-language complexity → start with JS/TS + Python; expand later

Repository Purpose

This is a development/planning repository. Content will evolve as we finalize the roadmap and transition to the implementation repository. No application code is scaffolded here yet.

About

## Repository description (recommended) AI-powered automated debugging with iterative fix loops. GitHub/CI integration. Supabase + Next.js. Planning repo. Optional shorter fallback (if the UI truncates): AI debugging with iterative fix loops. GitHub/CI + Supabase + Next.js. Planning repo.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published