Merge branch 'innerloopinc/REQ-2-on-demand-interview-prep-pipeline'

Author: mcherif

alembic/versions/b16bb02fc7ed_add_interview_prep_sheet.py

@@ -0,0 +1,49 @@
+"""add_interview_prep_sheet
+
+Revision ID: b16bb02fc7ed
+Revises: a18ea51b76fd
+Create Date: 2026-05-02 11:19:16.704168
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = 'b16bb02fc7ed'
+down_revision: Union[str, Sequence[str], None] = 'a18ea51b76fd'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('interview_prep_sheets',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('job_application_id', sa.Integer(), nullable=False),
+    sa.Column('status', sa.String(), nullable=False),
+    sa.Column('company_snapshot', sa.Text(), nullable=True),
+    sa.Column('role_requirements_summary', sa.Text(), nullable=True),
+    sa.Column('likely_technical_questions', sa.Text(), nullable=True),
+    sa.Column('likely_behavioral_questions', sa.Text(), nullable=True),
+    sa.Column('talking_points', sa.Text(), nullable=True),
+    sa.Column('gaps_or_risks', sa.Text(), nullable=True),
+    sa.Column('prep_plan_30_min', sa.Text(), nullable=True),
+    sa.Column('error_message', sa.Text(), nullable=True),
+    sa.Column('generated_at', sa.DateTime(), nullable=True),
+    sa.Column('created_at', sa.DateTime(), nullable=True),
+    sa.ForeignKeyConstraint(['job_application_id'], ['jobs.id'], ),
+    sa.PrimaryKeyConstraint('id'),
+    sa.UniqueConstraint('job_application_id')
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_table('interview_prep_sheets')
+    # ### end Alembic commands ###

models/database.py

@@ -1,5 +1,5 @@
 import datetime
-from sqlalchemy import Column, Integer, String, Text, DateTime, Date, UniqueConstraint
+from sqlalchemy import Column, Integer, String, Text, DateTime, Date, UniqueConstraint, ForeignKey
 from sqlalchemy.orm import declarative_base
 
 Base = declarative_base()
@@ -62,3 +62,20 @@ class ApplicationHistory(Base):
     __table_args__ = (
         UniqueConstraint("company", "job_title", name="_company_job_uc"),
     )
+
+class InterviewPrepSheet(Base):
+    __tablename__ = "interview_prep_sheets"
+
+    id = Column(Integer, primary_key=True)
+    job_application_id = Column(Integer, ForeignKey("jobs.id"), unique=True, nullable=False)
+    status = Column(String, nullable=False, default="processing")
+    company_snapshot = Column(Text, nullable=True)
+    role_requirements_summary = Column(Text, nullable=True)
+    likely_technical_questions = Column(Text, nullable=True)
+    likely_behavioral_questions = Column(Text, nullable=True)
+    talking_points = Column(Text, nullable=True)
+    gaps_or_risks = Column(Text, nullable=True)
+    prep_plan_30_min = Column(Text, nullable=True)
+    error_message = Column(Text, nullable=True)
+    generated_at = Column(DateTime, nullable=True)
+    created_at = Column(DateTime, default=datetime.datetime.utcnow)

run_pipeline.py

@@ -1352,6 +1352,49 @@ def cover_letter_cmd(job_id: int, profile: str, model: str, regenerate: bool):
         session.close()
 
 
+@cli.command(name='interview-prep')
+@click.argument('job_application_id', type=int)
+@click.option('--profile', default='profile.yaml', show_default=True, help='Path to candidate profile YAML')
+def interview_prep_cmd(job_application_id: int, profile: str):
+    """Generate a personalized interview prep sheet for a job application."""
+    from utils.interview_prep import run_interview_prep
+
+    profile_data = _load_profile(profile)
+    if not profile_data:
+        click.echo(click.style(f"Could not load profile from {profile}", fg="red"), err=True)
+        sys.exit(1)
+
+    def _progress(step: int, msg: str) -> None:
+        click.echo(f"[{step}/6] {msg}")
+
+    session = SessionLocal()
+    try:
+        sheet = run_interview_prep(job_application_id, profile_data, session, progress_callback=_progress)
+        click.echo(click.style(f"\n✅ Interview prep sheet saved (ID: {sheet.id})", fg="green"))
+
+        import json as _json
+        sections = [
+            ("Company Snapshot", sheet.company_snapshot),
+            ("Role Summary", sheet.role_requirements_summary),
+            ("Technical Questions", sheet.likely_technical_questions),
+            ("Behavioral Questions", sheet.likely_behavioral_questions),
+            ("Talking Points", sheet.talking_points),
+            ("Gaps / Risks", sheet.gaps_or_risks),
+            ("30-Min Prep Plan", sheet.prep_plan_30_min),
+        ]
+        for title, raw in sections:
+            click.echo(click.style(f"\n--- {title} ---", fg="cyan"))
+            try:
+                click.echo(_json.dumps(_json.loads(raw), indent=2, ensure_ascii=False))
+            except (TypeError, ValueError):
+                click.echo(str(raw))
+    except (ValueError, RuntimeError) as exc:
+        click.echo(click.style(str(exc), fg="red"), err=True)
+        sys.exit(1)
+    finally:
+        session.close()
+
+
 @cli.command(name='perf')
 @click.option('--job', default=None, help='Filter to runs matching this job substring (e.g. "Coinbase")')
 @click.option('--last', default=10, type=int, show_default=True, help='Number of most-recent runs to plot')

tests/test_interview_prep.py

@@ -0,0 +1,197 @@
+"""
+Tests for utils/interview_prep.py.
+
+Uses an in-memory SQLite session and mocks requests.post so no real
+Ollama instance is needed.
+"""
+import json
+from unittest.mock import MagicMock, call, patch
+
+import pytest
+import requests
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker
+
+import config
+from models.database import Base, InterviewPrepSheet, Job
+from utils.interview_prep import run_interview_prep
+
+
+# ---------------------------------------------------------------------------
+# Fixtures
+# ---------------------------------------------------------------------------
+
+@pytest.fixture
+def engine():
+    e = create_engine("sqlite:///:memory:")
+    Base.metadata.create_all(e)
+    yield e
+    Base.metadata.drop_all(e)
+
+
+@pytest.fixture
+def session(engine):
+    S = sessionmaker(bind=engine)
+    s = S()
+    yield s
+    s.close()
+
+
+@pytest.fixture
+def profile():
+    return {
+        "personal": {"name": "Jane Doe", "current_title": "Backend Engineer"},
+        "skills": ["Python", "SQL", "Docker"],
+        "work_history": [
+            {
+                "company": "Acme",
+                "title": "Senior Engineer",
+                "from": "2020",
+                "to": "present",
+                "highlights": ["Built scalable APIs", "Led team of 5"],
+            }
+        ],
+    }
+
+
+@pytest.fixture
+def job(session):
+    j = Job(
+        external_id="interview-test-001",
+        source="remotive",
+        company="TechCorp",
+        title="Senior Backend Engineer",
+        location="Remote",
+        url="https://example.com/jobs/1",
+        description="<p>We need a Python expert with SQL and Docker skills.</p>",
+        description_text="We need a Python expert with SQL and Docker skills.",
+        status="shortlisted",
+    )
+    session.add(j)
+    session.commit()
+    session.refresh(j)
+    return j
+
+
+def _ollama_response(content: dict) -> MagicMock:
+    mock = MagicMock()
+    mock.raise_for_status.return_value = None
+    mock.json.return_value = {"message": {"content": json.dumps(content)}}
+    return mock
+
+
+CONTEXT_RESPONSE = {
+    "company_snapshot": {
+        "industry": "SaaS",
+        "likely_size": "50-200",
+        "culture_signals": ["fast-paced"],
+        "red_flags": [],
+    },
+    "role_summary": {
+        "core_responsibilities": ["Build APIs"],
+        "must_have_skills": ["Python", "SQL"],
+        "nice_to_have_skills": ["Docker"],
+        "seniority_signals": "Senior",
+    },
+}
+
+QUESTIONS_RESPONSE = {
+    "technical_questions": ["Describe your Python API experience.", "How do you optimize SQL queries?"],
+    "behavioral_questions": ["Tell me about a time you led a project.", "How do you handle deadlines?"],
+}
+
+MAPPING_RESPONSE = {
+    "talking_points": [
+        {
+            "jd_requirement": "Python expertise",
+            "candidate_evidence": "Built scalable APIs at Acme",
+            "suggested_story": "Discuss the API work at Acme",
+        }
+    ],
+    "gaps_or_risks": [],
+}
+
+ACTION_PLAN_RESPONSE = {
+    "minutes_0_10": ["Review company website"],
+    "minutes_10_20": ["Practice Python API questions"],
+    "minutes_20_30": ["Rehearse STAR stories"],
+    "priority_note": "Emphasize Python API experience",
+}
+
+
+# ---------------------------------------------------------------------------
+# Tests
+# ---------------------------------------------------------------------------
+
+def test_run_interview_prep_success(session, job, profile):
+    responses = [
+        _ollama_response(CONTEXT_RESPONSE),
+        _ollama_response(QUESTIONS_RESPONSE),
+        _ollama_response(MAPPING_RESPONSE),
+        _ollama_response(ACTION_PLAN_RESPONSE),
+    ]
+    with patch("utils.interview_prep.requests.post", side_effect=responses):
+        sheet = run_interview_prep(job.id, profile, session)
+
+    assert sheet.status == "completed"
+    assert sheet.job_application_id == job.id
+    assert sheet.generated_at is not None
+
+    # All 7 sections populated
+    assert json.loads(sheet.company_snapshot)["industry"] == "SaaS"
+    assert json.loads(sheet.role_requirements_summary)["seniority_signals"] == "Senior"
+    assert len(json.loads(sheet.likely_technical_questions)) == 2
+    assert len(json.loads(sheet.likely_behavioral_questions)) == 2
+    assert len(json.loads(sheet.talking_points)) == 1
+    assert json.loads(sheet.gaps_or_risks) == []
+    assert json.loads(sheet.prep_plan_30_min)["priority_note"] == "Emphasize Python API experience"
+
+
+def test_run_interview_prep_invalid_job_id(session, profile):
+    with pytest.raises(ValueError, match="not found"):
+        run_interview_prep(99999, profile, session)
+
+    # No sheet should have been created
+    count = session.query(InterviewPrepSheet).count()
+    assert count == 0
+
+
+def test_run_interview_prep_missing_description(session, profile):
+    empty_job = Job(
+        external_id="interview-nodesc-001",
+        source="remotive",
+        company="TechCorp",
+        title="Engineer",
+        location="Remote",
+        url="https://example.com/jobs/2",
+        description=None,
+        description_text=None,
+        status="shortlisted",
+    )
+    session.add(empty_job)
+    session.commit()
+    session.refresh(empty_job)
+
+    with pytest.raises(ValueError, match="no description"):
+        run_interview_prep(empty_job.id, profile, session)
+
+    sheet = session.query(InterviewPrepSheet).filter(
+        InterviewPrepSheet.job_application_id == empty_job.id
+    ).first()
+    assert sheet is not None
+    assert sheet.status == "failed"
+    assert "empty" in sheet.error_message.lower()
+
+
+def test_run_interview_prep_timeout_retries(session, job, profile):
+    with patch("utils.interview_prep.requests.post", side_effect=requests.Timeout("timed out")):
+        with patch("utils.interview_prep.time.sleep"):
+            with pytest.raises(RuntimeError, match="timed out"):
+                run_interview_prep(job.id, profile, session)
+
+    sheet = session.query(InterviewPrepSheet).filter(
+        InterviewPrepSheet.job_application_id == job.id
+    ).first()
+    assert sheet is not None
+    assert sheet.status == "failed"
+    assert "timed out" in sheet.error_message.lower()