AOSSIE-Org / AOSSIE-Org/DebateAI
[FEATURE]: AI-generated Post-Debate Performance Report
- 主要语言
- TypeScript
- 星标
- 84
- 派生
- 198
- 平均合并
- 2 天 19 小时
- 30 天内合并 PR
- 30
描述
### Feature and its Use Cases
## Problem
Right now once a debate ends, the user just sees win/loss and optionally a
transcript export (#403). There's no actual feedback on how well they
argued — no scoring, no idea which points were weak, no guidance on what
to improve next time.
## Proposed Feature
Show an AI-generated performance report right after the debate ends (both
vs-bot and vs-human/team modes):
1. **Overall scores** — persuasion, clarity, rebuttal effectiveness (0-100 each)
2. **Per-argument breakdown** — each statement tagged Strong/Moderate/Weak
with a one-line reason (e.g. "no supporting evidence")
3. **Fallacy flags** — can reuse #377's fallacy detection if that lands
first, otherwise a simple LLM prompt on the transcript works as v1
4. **Improvement tips** — 2-3 specific suggestions based on the actual
transcript, not generic advice
## Approach
- New endpoint that generates the report once after a debate ends and
caches it (no regenerating on every view)
- Reuse the existing LLM client already used for debate responses
- New results component to show scores + breakdown + tips instead of
just win/loss
## Why this matters
This is what actually makes DebateAI useful for improving as a debater,
not just playing. #403 only exports the raw transcript and #377 is just
the detection model — neither gives the user a readable, actionable report.
### Additional Context
_No response_
### Code of Conduct
- [x] I have joined the [Discord server](https://discord.gg/hjUhu33uAn) and will post updates there
- [x] I have searched existing issues to avoid duplicates
贡献指南
这个仓库没有索引到贡献指南
评估
这个 Issue 还没有评估数据。