Online Debate Versus AICopy Anchor Link

Lets a debater start a full practice round against an AI opponent — pick a debate-timer format and side, then submit their own speeches one at a time, in turn order.

• Route: none of its own — the panel renders inside /coaching. /practice/versus-ai now serves Practice vs AI, the ported full-round feature.
• Nav: the Coach hub
• Package: debate-round

What it showsCopy Anchor Link

A form to start a round: a round ID, a debate-timer format, and which side the user is speaking (formats with no secondary side, like Congress, only offer the primary side). Starting a round derives its full speech order via buildAiVersusSpeechOrder and persists an empty round through saveAiVersusRound.

Once a round is active, the panel renders its full turn order — each slot tagged "You" or "AI" and marked Delivered/Next/Pending — and, when it's the user's turn, a text area to type and submit the next expected speech. Submission calls validateSpeechSubmission before saving. On the AI's turn, a "Generate AI speech" button calls the real AI speech-generation request and saves the returned text as that slot's speech.

Every delivered AI speech in the turn-order list gets its own "Regenerate" button. Clicking it re-requests that slot, from the same prior-speeches context originally used to generate it, and replaces it in place — every other speech is untouched.

A "Download transcript" button appears once a round's nextSlot is null (every speech delivered), rendering a plain-text transcript labeled "You"/"AI" per slot.

The speech-submission text field also has a "🎤 Record"/"Stop recording" button dictating directly into the same speechText state via the browser's Web Speech API, with a disabled fallback when unsupported.

Data flowCopy Anchor Link

round/ai-versus-speech-order.ts
  → buildAiVersusSpeechOrder(styleKey, userSide)   — derives the turn order
  → validateSpeechSubmission(order, submittedCount, speechName)

state/aiVersusRounds.ts (localStorage: aiVersusRounds)
  → buildAiVersusRoundsPanelView()
  → getAiVersusRoundStatus(roundId)
  → panels/AiVersusRoundPanel.tsx
  → apps/debate-ai.com/app/versus-ai/page.tsx

Generating the AI's next speech:
panels/AiVersusRoundPanel.tsx
  → buildAiResponseRequest(order, submittedCount, submittedSpeeches)
  → round/ai-versus-speech-client.ts's requestAiVersusSpeech(request)
      → round/ai-versus-speech-ai.ts's AI_VERSUS_SPEECH_SYSTEM_PROMPT + buildAiVersusSpeechUserPrompt(request)
      → POST /api/reason-ai
      → parseAiVersusSpeechResponse(text)
  → saveAiVersusRound({ ...record, submittedSpeeches: [...] })

Known gapsCopy Anchor Link

None open. Every delivered AI speech can be regenerated independently in place; a completed round's transcript can be downloaded as plain text; speech submission stays text-only beyond microphone dictation, and there is no transcription path for an already-recorded audio/video file, matching every other panel in this repo that shares that same gap.