Contribution idea: wire the deck critique endpoint
Hi! I would like to help with a small, reviewable contribution.
While exploring PitchFight AI, I noticed that ENABLE_DECK_CRITIQUE=true is documented, the frontend/product direction mentions deck critique, and /api/deck-critique currently returns a placeholder. A safe first contribution could be:
- define the backend request/response contract for deck critique;
- support a text-only fallback first, so it works even when a vision model is not configured;
- reuse the existing coaching/scorecard style so feedback feels native to PitchFight;
- keep all model/API keys backend-only and avoid frontend secret exposure.
If this direction fits, I can prepare a small patch around the existing handle_deck_critique_placeholder path.
Hi, thanks for pointing this out and for offering to help!
You’re right — deck critique was mentioned in the docs, but that was from an earlier idea and not part of the current hackathon submission scope. For now I’m freezing the build around the core PitchFight flow: founder briefing, pitch battle, voice mode, deal phase, scorecard, and retry coaching.
I’ll clean up the README/Field Notes so it doesn’t confuse contributors. Really appreciate the thoughtful suggestion — deck critique is something I’d be happy to revisit after the submission window.
I prepared this as a small PR here: https://huggingface.co/spaces/build-small-hackathon/PITCHFIGHT_AI/discussions/2
Scope kept intentionally narrow:
- wires
/api/deck-critiqueas a text-only backend MVP; - accepts
deck_text,slides, or existingstartupcontext; - uses NVIDIA JSON output when available;
- falls back to local heuristics if the model is unavailable;
- adds focused
unittestcoverage for validation, fallback, and model JSON paths.
Local checks I ran:
python3 -m unittest tests/test_deck_critique.py -vpython3 -m py_compile app.py core/*.py tests/test_deck_critique.py