Interview Sessions

Overview

Interview sessions are the core of the platform’s practice experience. Each session is tracked in the InterviewSession model, which records questions, answers, scores, feedback, and performance metrics.

Session Types

The platform supports multiple session types identified by the session_type field:

Session Type	Description	Use Case
`mock`	General mock interviews	Behavioral and situational questions
`mock_resume`	Resume-based mock interviews	Questions derived from your specific experience
`technical`	Technical knowledge interviews	Deep-dive into technical concepts
`coding`	Data science coding challenges	Coding problems and algorithm questions
`hr`	HR screening interviews	Culture fit and soft skills
`agentic`	AI-powered adaptive interviews	Dynamic question generation based on performance
`debugging`	Debugging challenges	Code debugging and problem-solving

Interview Session Model

Each session stores the following data:

class InterviewSession(db.Model):
    id                # Unique session identifier
    user_id           # Foreign key to User
    session_type      # Type of interview (see table above)
    questions         # JSON array of questions asked
    score             # Overall score (0-100)
    feedback          # Detailed feedback text
    duration          # Session duration in seconds
    created_at        # Session start timestamp
    completed_at      # Session completion timestamp
    speech_metrics    # JSON with research-grade speech analysis

Starting a Session

Choose session type

Select the type of interview you want to practice:Mock Interview (Resume-based):

POST /api/mock_interview/questions
Authorization: Bearer <your-token>
{
  "question_count": 5,
  "difficulty": "medium"
}

Requires uploaded resume. Generates questions based on your experience.Technical Interview:

POST /api/technical/questions
Authorization: Bearer <your-token>
{
  "topic": "machine_learning",
  "count": 8
}

Coding Interview:

POST /api/coding/questions
{
  "count": 3,
  "difficulty": "medium"
}

Each endpoint returns a set of questions tailored to the session type.

Initialize session

When you start a live interview (with voice), the system:

Creates a new InterviewSession record
Initializes speech tracking metrics
Starts AssemblyAI streaming (or mock streamer)
Begins recording session start time

WebSocket Event:

socket.emit('start_interview', {
  user_id: <your-user-id>,
  session_type: 'technical',
  topic: 'python'
});

Response:

{
  "status": "started",
  "message": "Interview started successfully",
  "use_mock": false,
  "session_id": 123
}

Session initialization details

The system creates session tracking objects:

Session Key: {user_id}_{session_type}
Speech Metrics: Tracks speaking time, pauses, fluency
Running Statistics: Research-grade metrics for analysis
Silence Detection: Monitors long pauses (>5 seconds)

Located in app.py:1218-1301

During the Session

Answering Questions

The interview flow depends on the session type:

Live Voice Interview
Text-based Interview
Coding Session

Real-time Speech Processing:

Speak your answer: Audio is streamed to AssemblyAI
Transcription: Your speech is converted to text in real-time
Metrics tracking: Speaking time, pauses, and fluency monitored
Question completion: System detects when you finish answering

WebSocket Events:

// Receive transcription
socket.on('transcription', (data) => {
  console.log('Partial:', data.text);
});

// Final transcript
socket.on('final_transcription', (data) => {
  console.log('Final:', data.text);
});

The system tracks:

Speech start/end times
Silence periods
Total words spoken
Response latency

Submit written answers:

POST /api/mock_interview/evaluate_answer
Authorization: Bearer <your-token>
{
  "question": "Tell me about a challenging project",
  "answer": "In my previous role, I led a team...",
  "expected_answer": "Expected to mention leadership, problem-solving..."
}

The system evaluates:

Semantic similarity to expected answer
Keyword coverage
Content quality
Completeness

Submit code solutions:

POST /api/coding/evaluate
{
  "question": "Write a function to find duplicates in an array",
  "user_code": "def find_duplicates(arr): ...",
  "language": "python"
}

Evaluation includes:

Code correctness
Efficiency and optimization
Best practices
Edge case handling

Real-time Feedback

During the session, you receive: Immediate Evaluation:

Semantic similarity score (0-1)
Keyword coverage percentage
Specific improvement suggestions
Concept mastery updates

Speech Metrics (live interviews):

Speaking ratio (speaking time / total time)
Words per minute (WPM)
Long pause count
Fluency score
Average response latency

Located in app.py:1000-1179

Running Statistics

The RunningStatistics class tracks:

class RunningStatistics:
    session_start           # Session start timestamp
    questions_answered      # Number of questions completed
    total_words            # Total words spoken
    speaking_time          # Cumulative speaking time
    silence_time           # Cumulative silence time
    long_pause_count       # Pauses longer than 5 seconds
    forced_silence_time    # Time waiting for next question
    response_latencies     # List of response start times
    semantic_similarities  # List of similarity scores per answer

These metrics are computed at the end for the final analysis.

Completing the Session

Stop the interview

End your session when you’ve completed all questions:WebSocket:

socket.emit('stop_interview', {
  user_id: <your-user-id>
});

HTTP Fallback:

POST /api/stop_interview
Authorization: Bearer <your-token>
{
  "user_id": <your-user-id>
}

Compute final metrics

The system performs comprehensive analysis:Research-grade Metrics:

Session duration (total time)
Effective duration (excluding forced silence)
Speaking vs. silence ratio
Words per minute (WPM)
Average semantic similarity
Fluency score (0-1)
Long pause analysis

Function: compute_research_metrics() in app.py:1053-1055Example Output:

{
  "session_duration": 600.0,
  "effective_duration": 580.0,
  "speaking_time": 420.0,
  "silence_time": 160.0,
  "speaking_ratio": 0.724,
  "total_words": 850,
  "wpm": 121.4,
  "long_pause_count": 3,
  "avg_response_latency": 2.3,
  "fluency_score": 0.867,
  "questions_answered": 8,
  "avg_semantic_similarity": 0.78
}

Save to database

The completed session is saved with:

session = InterviewSession(
    user_id=user_id,
    session_type='technical',
    questions=json.dumps(questions_list),
    score=avg_semantic_similarity * 100,
    duration=int(session_duration),
    completed_at=datetime.utcnow(),
    speech_metrics=json.dumps(metrics)
)

Located in app.py:1158-1165

Generate session summary

For mock interviews, request an overall summary:

POST /api/mock_interview/session_summary
Authorization: Bearer <your-token>
{
  "answers": [
    {
      "question": "...",
      "answer": "...",
      "evaluation": {...}
    }
  ]
}

Returns comprehensive feedback on:

Overall performance
Strongest areas
Areas for improvement
Specific recommendations
Weakest concepts to practice

The summary is saved to the session’s feedback field.

Reviewing Sessions

View Session History

Retrieve your past interview sessions:

GET /api/sessions?type=technical
Authorization: Bearer <your-token>

Response:

{
  "sessions": [
    {
      "id": 123,
      "session_type": "technical",
      "score": 78.5,
      "duration": 600,
      "created_at": "2026-03-03T14:30:00",
      "completed_at": "2026-03-03T14:40:00",
      "questions_count": 8
    }
  ]
}

Filter by:

type: Session type (mock, technical, coding, etc.)
Automatically sorted by most recent first

Located in app.py:3170-3186

Session Details

Get full details of a specific session including:

Complete list of questions
Your answers and transcripts
Detailed feedback
All speech metrics
Concept mastery updates

Performance Analytics

Session data feeds into: UserMastery Updates:

Mastery level per topic
Weak vs. strong concepts
Learning velocity
Difficulty progression

See Performance Tracking for detailed analytics.

Session Best Practices

For best results:

Upload your resume before starting resume-based sessions
Ensure quiet environment for voice interviews
Answer thoughtfully rather than rushing
Review feedback immediately after completing
Practice weak areas identified in session summaries

Sessions must have at least one answered question to generate valid metrics. The system validates with metrics.get('questions_answered', 0) > 0 before analysis.

Next Steps

Track Your Progress

View mastery levels and learning analytics

Interview Endpoints

Explore all interview endpoints

Resume Analysis

Generate personalized questions from your resume

Technical Q&A

Deep-dive into technical concept practice

Overview

Getting Started

Core Features

User Guide

Overview

Session Types

Interview Session Model

Starting a Session

During the Session

Answering Questions

Real-time Feedback

Running Statistics

Completing the Session

Reviewing Sessions

View Session History

Session Details

Performance Analytics

Session Best Practices

Next Steps

Track Your Progress

Interview Endpoints

Resume Analysis

Technical Q&A

Build docs developers (and LLMs) love

Overview

Getting Started

Core Features

User Guide

​Overview

​Session Types

​Interview Session Model

​Starting a Session

​During the Session

​Answering Questions

​Real-time Feedback

​Running Statistics

​Completing the Session

​Reviewing Sessions

​View Session History

​Session Details

​Performance Analytics

​Session Best Practices

​Next Steps

Track Your Progress

Interview Endpoints

Resume Analysis

Technical Q&A

Build docs developers (and LLMs) love

Overview

Session Types

Interview Session Model

Starting a Session

During the Session

Answering Questions

Real-time Feedback

Running Statistics

Completing the Session

Reviewing Sessions

View Session History

Session Details

Performance Analytics

Session Best Practices

Next Steps