legacyexe

legacy.exe — Compare & Refactor

Minimal LLM-powered code comparison / refactoring tool.
It includes:

FastAPI server (server/app/main.py) — GitHub integration, filters, LLM calls.
Frontend client (frontend/) — static HTML/JS, runs against your local API.
Tests (server/tests/) — unit + e2e with Uvicorn.
BYO-key client (old) (old/) — experimental web UI for direct provider calls.

✨ Features

Fetches code from GitHub repos (with PAT or GitHub App).
Filters by file extension + path globs.
Sends numbered code to an LLM for refactor/review.
Provider-agnostic: works with Groq (OpenAI API) or Gemini (v1beta).

Returns structured JSON:

{ "report": "...", "summary": ["..."], "updated_files": [{ "path": "...", "content": "..." }] }

Frontend UI to run comparisons and download results.

🚀 Quick Start (Windows/PowerShell)

Backend (FastAPI)

cd server

# (optional) pick your Python; 3.11/3.12 both fine
py -V
py -m venv .venv
.venv\Scripts\Acticvate.ps1

# deps
py -m pip install --upgrade pip
py -m pip install -r requirements.txt

Env file

Copy-Item .env.example .env
notepad .env

Set at least:

ALLOWED_ORIGINS=http://127.0.0.1:5501,http://localhost:5501

# Prefer Gemini (works out-of-the-box)
LLM_API_URL=https://generativelanguage.googleapis.com/v1beta
LLM_API_KEY=YOUR_GEMINI_KEY
LLM_MODEL=gemini-1.5-flash

# Optional (Groq, OpenAI-compatible)
GROQ_API_KEY=YOUR_GROQ_KEY
GROQ_MODEL=llama-3.1-8b-instant

# Optional GitHub
GITHUB_PAT=
GITHUB_APP_ID=
GITHUB_INSTALLATION_ID=
GITHUB_PRIVATE_KEY_PATH=./keys/github_app.pem

Run API

uvicorn app.main:app --host 127.0.0.1 --port 8000 --reload
# → http://127.0.0.1:8000

Smoke-test API

# health
Invoke-RestMethod -Uri "http://127.0.0.1:8000/health" -Method GET

$body = @{
  repo          = "fromLELI/storycompare-with-LLM";
  branch        = "main";
  include_ext   = @(".py", ".js", ".html", ".css");
  include_paths = @("/");                            # scan whole repo
  requisitos    = "Refatore o codigo da melhor forma possivel";
  debug_no_llm  = $false;
  max_files     = 10;
  max_bytes     = 200000;

  # get the raw LLM output back for debugging
  debug_echo_raw = $true
} | ConvertTo-Json -Depth 6

$r = Invoke-RestMethod -Uri "http://127.0.0.1:8000/compare" -Method POST -ContentType "application/json" -Body $body
$r | ConvertTo-Json -Depth 10

Frontend (static)

Option A — quick local server (Python):

cd ..\frontend
py -m http.server 5501
# → http://127.0.0.1:5501  (open index.html)

Option B — VS Code Live Server:

Open frontend/ folder in VS Code → “Go Live” (ensure port 5501 if you didn’t change CORS).

End-to-end flow

Start backend (step 1).
Start frontend (step 2).
In the UI, fill: repo, branch, provider (Gemini or Groq), model, API key.
Click comparar. Result JSON shows report, summary, updated_files.

Running tests (Windows)

cd server
.\.venv\Scripts\Activate.ps1

# Unit + debug tests (no live LLM)
pytest -q

# Live LLM (needs LLM_API_KEY in your .env)
pytest -q -m live_llm

PowerShell curl equivalent (if you prefer curl.exe)

curl.exe -s "http://127.0.0.1:8000/compare" -H "content-type: application/json" -d ^
"{`"repo`":`"ddb0000/legacyexe`",`"branch`":`"main`",`"include_ext`":[`".py`",`".js`"],`"include_paths`":[`"server/`",`"frontend/`"],`"requisitos`":`"Sem codigo duplicado!`",`"debug_no_llm`":false}"

That’s it. Run API, run static server, hit /compare.

🚀 Quick Start (Linux)

1. Clone & install

cd server
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

2. Environment

Copy .env.example → .env and edit:

# CORS (local frontend)
ALLOWED_ORIGINS=http://127.0.0.1:5501,http://localhost:5501

# === Generic LLM config (preferred) ===
LLM_API_URL=https://generativelanguage.googleapis.com/v1beta
LLM_API_KEY=your_gemini_key
LLM_MODEL=gemini-1.5-flash

# === Back-compat (Groq) ===
GROQ_API_KEY=your_groq_key
GROQ_MODEL=llama-3.1-8b-instant

# === GitHub auth (optional) ===
GITHUB_PAT=
GITHUB_APP_ID=
GITHUB_INSTALLATION_ID=
GITHUB_PRIVATE_KEY_PATH=./keys/github_app.pem

3. Run server

uvicorn app.main:app --reload
# API at http://127.0.0.1:8000

4. Run frontend

Open frontend/index.html in a browser. (LiverServer preferred) Fill in repo, branch, model, key → click comparar.

📡 API

POST /compare

{
  "repo": "org/repo",
  "branch": "main",
  "include_ext": [".py",".js"],
  "include_paths": ["src/"],
  "requisitos": "Sem código duplicado!",
  "debug_no_llm": false,
  "max_files": 10,
  "max_bytes": 200000
}

Response:

{
  "report": "Refatoração aplicada",
  "summary": ["..."],
  "updated_files": [
    { "path": "src/util.py", "content": "def util(): ..." }
  ]
}

Health check: GET /health → {"ok": true}

🧪 Tests

From server/:

pytest -q               # unit + debug tests
pytest -q -m live_llm   # live LLM tests (requires LLM_API_KEY in env)

test_e2e_uvicorn_debug.py: spins up API with debug_no_llm.
test_e2e_uvicorn_live_llm.py: full roundtrip with real LLM.
Unit tests: filters, safe_json, repo fallback, etc.

🖥️ Project Layout

frontend/       # static UI
server/app/     # FastAPI app
server/tests/   # pytest suite

🔑 Providers

Gemini (preferred):

LLM_API_URL=https://generativelanguage.googleapis.com/v1beta
LLM_MODEL=gemini-1.5-flash

Groq (OpenAI-compat):

LLM_API_URL=https://api.groq.com/openai/v1/chat/completions
LLM_MODEL=llama-3.1-8b-instant

Pass via .env or override in request body.

⚡ Roadmap

Multi-file patch/diff export
Better error surfaces in frontend
More providers (OpenAI, Anthropic, etc.)
Richer unit test coverage