- Replace urllib with requests-compatible pattern, add timeout=120s (issues #5,#15) - Fix docstring --model-magic -> --model (issue #11) - Fix questions.json path resolution via __file__ (issue #10) - Dedent experiment_results.append() outside except block (issue #9) - Add sliding window (last 5) for accumulated context (issue #8) - Add OLLAMA_API_KEY auth support via header (issue #7) - sys.exit(1) on Ollama failure instead of fake response (issue #6) - Make Ollama URL configurable: --ollama-url / OLLAMA_URL env (issue #4) - Add HTTP cleartext warning (issue #3) - Validate --project name (alphanumeric only), protect --output-dir (issues #1,#2,#14) - Remove PDFs and pyvenv.cfg from git tracking (issues #12,#13) - Update .gitignore for PDFs, pyvenv.cfg, output/
Description
Proposal-to-plan engine: run project proposals through the IDIOT method via a local LLM (Ollama), producing phased plans before any code is written.
Languages
Python
100%