story-retrieval

Select targeted stories from the persistent story corpus for immediate interview use.

Workflow

1. Resolve {user_dir} and gate

Require {user_dir}/CareerNavigator/StoryCorpus.json.

If missing or empty:

• Ask to run mine-stories first (or run it now with user approval if appropriate).
• Continue with reduced-confidence fallback from ExperienceLibrary.json only if the user wants to proceed immediately.

2. Build retrieval intent

Use available interview context:

• application_id or company + role
• interview_stage
• JD text (if available)
• user-stated focus areas (leadership, conflict, ambiguity, technical depth, etc.)

Derive desired competencies/themes for this specific interview.

3. Rank stories

For each story in corpus, score by:

• competency overlap
• theme overlap
• stage relevance (e.g. executive -> strategy/influence; technical -> architecture/debugging)
• quality signals (clarity/specificity/credibility)
• result + ownership signals
• recency/diversity (avoid returning 10 near-duplicates)

4. Return compact context set (Layer 3)

Return only a small subset:

• default: 8-12 stories
• for short prep: 5-8
• for deep prep: up to 15 if user explicitly asks

Output each item as:

• story_id
• one-line raw_summary
• mapped competencies/themes
• why selected for this interview
• STAR readiness status (star_ready)
• short coaching note if STAR gaps exist

5. Optional STAR promotion

For top stories with incomplete STAR fields:

• draft concise STAR skeletons
• set star_ready: true only when S/T/A/R are each concrete and evidence-backed
• write updates back to StoryCorpus.json

6. Handoff contract for interview-coach

When invoked by prep/mock flows, provide a compact handoff payload:

{
  "retrieval_context": {
    "company": "...",
    "role": "...",
    "interview_stage": "..."
  },
  "selected_stories": [
    {
      "story_id": "...",
      "summary": "...",
      "competencies": ["..."],
      "themes": ["..."],
      "star_ready": true
    }
  ]
}