Example Interview Session

This directory contains a complete example of a WREN interview session and the generated outputs.

Session Details

• Session ID: example_session
• Interview Type: Mock interview demonstrating adaptive questioning
• Turn Count: 8 turns (early termination)
• Completion Status: User-initiated early exit

Directory Structure

example_session/
├── logs/
│   └── conversation_20251108_150303.json    # Full transcript with reasoning
└── profiles/
    ├── profile_20251108_150303.json         # Machine-readable profile
    ├── profile_20251108_150303.md           # Human-readable markdown
    └── profile_20251108_150303_SHAREABLE.txt # Formatted for sharing

What This Example Shows

1. Adaptive Interviewing

The agent demonstrates:

• Deep listening: Questions reference specific phrases from previous answers
• Style matching: Adjusts question complexity to match user's vocabulary
• Coverage tracking: Ensures all dimensions are explored
• Natural flow: Conversation feels organic, not scripted

Example exchange:

USER: "I welcome obstruction when it serves the subject's resistance—
Beckett's deliberate impoverishment mirrors existential depletion."

AGENT: "You describe emotional accessibility as something that 'pulses 
through fissures in the obstruction'—a beautiful metaphor that suggests 
you're attuned to the rhythm of revelation and concealment."

The agent directly echoes the user's language ("obstruction," "fissures") and asks about pacing based on their sophisticated response.

2. Profile Generation

The generated profile contains:

Style Signature (0-100 scales):

• Prose density: 70 (dense, literary)
• Pacing: 60 (deliberate but forward-moving)
• Tone: 10 (dark, serious)
• Worldbuilding: 20 (minimal, interior-focused)
• Character focus: 90 (deeply psychological)

Taste Anchors:

• Loves: The Remains of the Day, Beloved, The Metamorphosis, Beckett, Lispector
• Avoids: Finnegans Wake (linguistic difficulty as subject)

Reader Archetype: "Fracture Dweller"

Reading Philosophy:

"You read as an act of emotional archaeology, seeking texts that fracture their own forms to admit the unsayable. You value provisional grammar over restored wholeness, and you dwell permanently inside the break."

3. Implicit Signals

Calculated from response patterns:

• Vocabulary richness: 0.95 (highly sophisticated, literary)
• Response brevity: 0.2 (long, essay-like responses)
• Engagement index: 0.95 (deeply engaged, philosophical)

These metrics are extracted without explicitly asking, showing how the system reads between the lines.

4. Kimi K2 Reasoning

The conversation log includes Kimi K2's internal thinking process (if reasoning was enabled), showing:

• How the agent analyzes each response
• What patterns it notices
• Why it chooses specific follow-up questions

5. Actionable Output

The profile can be directly used with any LLM:

"Write a short story for me with these parameters:
- Prose density: 70/100 (literary but not impenetrable)
- Tone: 10/100 (dark, restrained)
- Character focus: 90/100 (psychological over plot)
- Theme: Language failing to contain private catastrophe
- Ending: Transcendent through remaining broken
- Avoid: Linguistic cleverness as the subject itself"

How to View These Files

JSON Profile (Machine-Readable)

cat user_profiles/example_session/profiles/profile_20251108_150303.json

Or in Python:

import json
with open('user_profiles/example_session/profiles/profile_20251108_150303.json') as f:
    profile = json.load(f)
    print(profile['reader_archetype'])  # "Fracture Dweller"

Conversation Log

python view_conversation_log.py user_profiles/example_session/logs/conversation_20251108_150303.json

Shows formatted conversation with turn-by-turn breakdown.

Shareable Profile

cat user_profiles/example_session/profiles/profile_20251108_150303_SHAREABLE.txt

Formatted text ready to share on social media or documentation.

Key Takeaways

1. Interview feels natural: Not a questionnaire, but a conversation
2. Agent listens deeply: References specific words and concepts from previous turns
3. Profile is actionable: Precise numeric scores, not vague descriptions
4. Multiple formats: Machine-readable JSON + human-readable text
5. Transparent: Includes reasoning, explanations, and methodology

Running Your Own Interview

./run_interview.sh

Your session will be saved to user_profiles/cli_TIMESTAMP/ with the same structure as this example.

---

Note: This is a real interview session conducted with WREN, demonstrating the system's capabilities with a sophisticated reader. Your own interview will adapt to your style and preferences.