feat: complete voice session safety and confirmation flow

backend/app/schemas/voice_session_schemas.py

@@ -42,6 +42,12 @@ class VoiceSessionFinalizeRequest(BaseModel):
     generate_final_audio: bool = False
 
 
+class VoiceTurnConfirmRequest(BaseModel):
+    """Resolve one pending confirmation before the story continues."""
+
+    action: Literal["accept", "retry_recording", "switch_to_text"]
+
+
 class VoiceSessionAbandonRequest(BaseModel):
     """Explicitly abandon one in-progress session."""
 
@@ -75,8 +81,12 @@ class VoiceTurnSummaryResponse(BaseModel):
     intent_confidence: float | None = None
     understanding_summary: str | None = None
     requires_confirmation: bool = False
+    confirmation_state: str = "not_needed"
     confirmation_reason: str | None = None
     confirmation_message: str | None = None
+    safety_flags: list[str] = Field(default_factory=list)
+    safety_blocked: bool = False
+    safety_message: str | None = None
     assistant_text: str | None = None
     assistant_audio_ready: bool = False
     assistant_audio_url: str | None = None
@@ -105,7 +115,10 @@ class VoiceSessionSummaryResponse(BaseModel):
     latest_detected_intent: str | None = None
     latest_understanding_summary: str | None = None
     latest_requires_confirmation: bool = False
+    latest_confirmation_state: str | None = None
     latest_confirmation_message: str | None = None
+    latest_safety_flags: list[str] = Field(default_factory=list)
+    latest_safety_message: str | None = None
     latest_assistant_audio_ready: bool = False
     last_turn_status: str | None = None
     transcription_mode_hint: str | None = None
@@ -131,6 +144,25 @@ class VoiceTurnAcceptedResponse(BaseModel):
     status: str
 
 
+class VoiceSessionAnalyticsResponse(BaseModel):
+    """Aggregated voice co-creation analytics for one user."""
+
+    window_days: int | None = None
+    total_sessions: int = 0
+    active_sessions: int = 0
+    finalized_sessions: int = 0
+    abandoned_sessions: int = 0
+    total_turns: int = 0
+    successful_turns: int = 0
+    failed_turns: int = 0
+    asr_failures: int = 0
+    tts_failures: int = 0
+    low_confidence_turns: int = 0
+    safety_interventions: int = 0
+    turn_success_rate: float = 0.0
+    finalize_conversion_rate: float = 0.0
+
+
 class VoiceSessionFinalizeResponse(BaseModel):
     """Finalize response after a session is converted into a story."""