Analysis Integration

backend/services/llm/router.py

@@ -16,8 +16,10 @@
 from pydantic import ValidationError
 
 from services.llm.schemas import (
+    FullSessionFeedbackResponse,
     ReviewContextPayload,
     SessionFeedbackResponse,
+    SessionReviewPayload,
 )
 
 FEEDBACK_MODEL = "openai/gpt-oss-20b"
@@ -98,6 +100,68 @@
 Be specific and grounded in the provided transcript and segments. Do not invent \
 facts the candidate did not say. Output only the JSON object."""
 
+_SESSION_RESPONSE_SCHEMA = {
+    "type": "object",
+    "properties": {
+        "overallSummary": {"type": "string"},
+        "overallStrengths": {"type": "array", "items": {"type": "string"}},
+        "overallImprovements": {"type": "array", "items": {"type": "string"}},
+        "overallDeliveryNotes": {"type": "string"},
+        "questionReviews": {
+            "type": "array",
+            "items": {
+                "type": "object",
+                "properties": {
+                    "question": {
+                        "type": "object",
+                        "properties": {
+                            "id": {"type": "string"},
+                            "text": {"type": "string"},
+                        },
+                        "required": ["id", "text"],
+                    },
+                    "transcriptScores": _RESPONSE_SCHEMA["properties"][
+                        "transcriptScores"
+                    ],
+                    "feedback": _RESPONSE_SCHEMA["properties"]["feedback"],
+                    "modelAnswer": _RESPONSE_SCHEMA["properties"]["modelAnswer"],
+                },
+                "required": ["question", "transcriptScores", "feedback", "modelAnswer"],
+            },
+        },
+    },
+    "required": [
+        "overallSummary",
+        "overallStrengths",
+        "overallImprovements",
+        "overallDeliveryNotes",
+        "questionReviews",
+    ],
+}
+
+_SESSION_SYSTEM_PROMPT = """You are an expert interview coach reviewing a \
+candidate's full mock interview session.
+
+You receive multiple question-and-answer pairs. Each answer includes the \
+question text, full transcript, and timestamped segments with delivery signals \
+(arousal, dominance, valence — each 0..1). These are tone-and-delivery signals \
+only, not clinical labels.
+
+Produce a single JSON object with:
+1. overallSummary: 3-5 sentence overview of how the candidate performed across \
+the whole session (content and delivery trends).
+2. overallStrengths: 3-5 session-level strengths.
+3. overallImprovements: 3-5 session-level, actionable improvements.
+4. overallDeliveryNotes: one paragraph on vocal delivery patterns across answers.
+5. questionReviews: one entry per question answered, in the same order provided. \
+Each entry must include:
+    - question: echo back the exact id and text from the input
+    - transcriptScores: clarity/structure/relevance/conciseness (0..1) for that answer
+    - feedback: summary, strengths, improvements, deliveryNotes for that answer
+    - modelAnswer: a strong example answer for that question
+
+Be specific and grounded in the transcripts. Do not invent facts. Output only JSON."""
+
 
 def _require_groq() -> None:
     if not os.environ.get("GROQ_API_KEY"):
@@ -121,53 +185,109 @@ def _build_user_prompt(payload: ReviewContextPayload) -> str:
     )
 
 
-def _call_groq(user_prompt: str) -> str:
+def _call_groq(user_prompt: str, *, session: bool = False) -> str:
     completion = _groq.chat.completions.create(
         model=FEEDBACK_MODEL,
         messages=[
-            {"role": "system", "content": _SYSTEM_PROMPT},
+            {
+                "role": "system",
+                "content": _SESSION_SYSTEM_PROMPT if session else _SYSTEM_PROMPT,
+            },
             {"role": "user", "content": user_prompt},
         ],
         temperature=0.4,
         response_format={
             "type": "json_schema",
             "json_schema": {
-                "name": "session_feedback",
+                "name": "session_feedback" if session else "answer_feedback",
                 "strict": False,
-                "schema": _RESPONSE_SCHEMA,
+                "schema": _SESSION_RESPONSE_SCHEMA if session else _RESPONSE_SCHEMA,
             },
         },
     )
     return completion.choices[0].message.content or ""
 
 
-@router.post("/generate", response_model=SessionFeedbackResponse)
-async def generate(payload: ReviewContextPayload) -> SessionFeedbackResponse:
-    """Generate transcript scores + qualitative feedback + a model answer for a
-    recorded interview answer using a single Groq call."""
-    _require_groq()
-
-    if not payload.transcript.text.strip():
-        raise HTTPException(status_code=422, detail="Transcript is empty.")
+def _build_session_user_prompt(payload: SessionReviewPayload) -> str:
+    blocks = []
+    for n, answer in enumerate(payload.answers, start=1):
+        blocks.append(f"=== ANSWER {n} ===\n{_build_user_prompt(answer)}")
+    return "\n\n".join(blocks)
 
-    user_prompt = _build_user_prompt(payload)
 
+def _generate_with_retry(
+    user_prompt: str,
+    *,
+    session: bool = False,
+) -> dict:
     last_error: Exception | None = None
-    for _ in range(2):  # one retry on transport or validation failure
+    for _ in range(2):
         try:
-            raw = _call_groq(user_prompt)
+            raw = _call_groq(user_prompt, session=session)
         except Exception as exc:
             last_error = exc
             continue
 
         try:
-            data = json.loads(raw)
-            return SessionFeedbackResponse.model_validate(data)
-        except (json.JSONDecodeError, ValidationError) as exc:
+            return json.loads(raw)
+        except json.JSONDecodeError as exc:
             last_error = exc
             continue
 
     raise HTTPException(
         status_code=502,
         detail=f"Feedback generation failed: {last_error}",
     )
+
+
+@router.post("/generate", response_model=SessionFeedbackResponse)
+async def generate(payload: ReviewContextPayload) -> SessionFeedbackResponse:
+    """Generate transcript scores + qualitative feedback + a model answer for a
+    recorded interview answer using a single Groq call."""
+    _require_groq()
+
+    if not payload.transcript.text.strip():
+        raise HTTPException(status_code=422, detail="Transcript is empty.")
+
+    user_prompt = _build_user_prompt(payload)
+
+    try:
+        data = _generate_with_retry(user_prompt)
+        return SessionFeedbackResponse.model_validate(data)
+    except HTTPException:
+        raise
+    except ValidationError as exc:
+        raise HTTPException(
+            status_code=502,
+            detail=f"Feedback generation failed: {exc}",
+        ) from exc
+
+
+@router.post("/generate-session", response_model=FullSessionFeedbackResponse)
+async def generate_session(
+    payload: SessionReviewPayload,
+) -> FullSessionFeedbackResponse:
+    """Generate holistic session feedback plus per-question reviews in one Groq call."""
+    _require_groq()
+
+    if not payload.answers:
+        raise HTTPException(status_code=422, detail="No answers provided.")
+
+    for answer in payload.answers:
+        if not answer.transcript.text.strip():
+            raise HTTPException(
+                status_code=422, detail="One or more transcripts are empty."
+            )
+
+    user_prompt = _build_session_user_prompt(payload)
+
+    try:
+        data = _generate_with_retry(user_prompt, session=True)
+        return FullSessionFeedbackResponse.model_validate(data)
+    except HTTPException:
+        raise
+    except ValidationError as exc:
+        raise HTTPException(
+            status_code=502,
+            detail=f"Session feedback generation failed: {exc}",
+        ) from exc

backend/services/llm/schemas.py

@@ -61,3 +61,26 @@ class SessionFeedbackResponse(BaseModel):
     transcriptScores: TranscriptScores
     feedback: QualitativeFeedback
     modelAnswer: ModelAnswer
+
+
+class SessionReviewPayload(BaseModel):
+    """Request body for POST /feedback/generate-session — multiple Q&A pairs."""
+
+    answers: list[ReviewContextPayload]
+
+
+class QuestionReview(BaseModel):
+    question: Question
+    transcriptScores: TranscriptScores
+    feedback: QualitativeFeedback
+    modelAnswer: ModelAnswer
+
+
+class FullSessionFeedbackResponse(BaseModel):
+    """Response body for POST /feedback/generate-session."""
+
+    overallSummary: str
+    overallStrengths: list[str]
+    overallImprovements: list[str]
+    overallDeliveryNotes: str
+    questionReviews: list[QuestionReview]

frontend/app/InterviewClient.module.css

@@ -197,6 +197,39 @@
   cursor: not-allowed;
 }
 
+.feedbackModeFieldset {
+  border: none;
+  margin: 0;
+  padding: 0;
+  display: flex;
+  flex-direction: column;
+  gap: 0.55rem;
+}
+
+.feedbackModeFieldset:disabled {
+  opacity: 0.45;
+}
+
+.feedbackModeLegend {
+  font-size: 0.78rem;
+  text-transform: uppercase;
+  letter-spacing: 0.07em;
+  opacity: 0.5;
+  margin-bottom: 0.15rem;
+}
+
+.feedbackModeOption {
+  display: flex;
+  align-items: center;
+  gap: 0.55rem;
+  font-size: 0.88rem;
+  cursor: pointer;
+}
+
+.feedbackModeOption input {
+  accent-color: #fff;
+}
+
 .questionPlaceholder {
   font-size: 1rem;
   opacity: 0.35;