v0.0.0

Release Notes v0.0.0

v0.0.0 (MVP)

Initial MVP release of InterviewGraph.

Highlights

• Built end-to-end LangGraph pipeline for resume-based interview question generation.
• Added FastAPI service for text input and PDF upload workflows.
• Added containerization and GitHub Container Registry (GHCR) publish workflow.

Implemented

• Pipeline nodes implemented in casts/resume_ingestor/modules/nodes.py:
  • extract_text
  • parse_sections
  • extract_signals
  • generate_questions
  • rate_difficulty
  • format_output
• Graph flow wired in casts/resume_ingestor/graph.py:
  • extract_text -> parse_sections -> extract_signals -> generate_questions -> rate_difficulty -> format_output
• State/data contracts implemented in casts/resume_ingestor/modules/state.py.
• API endpoints added in app/main.py:
  • GET /health
  • POST /api/v1/interview-questions
  • POST /api/v1/interview-questions/upload

Packaging and Delivery

• Added Dockerfile for API container image build.
• Added .dockerignore to reduce image build context.
• Added GHCR publish workflow: .github/workflows/publish-ghcr.yml.

Quality and Validation

• Added and updated tests:
  • tests/node_tests/test_node.py
  • tests/cast_tests/resume_ingestor_test.py
  • tests/api_tests/test_api.py
• MVP regression baseline verified with passing test suite.

Known Limitations (MVP)

• Question generation is currently template-oriented and can feel generic.
• Difficulty scoring is deterministic and not fully content-aware.
• OCR for scanned/image PDFs is not included.
• RAG, multi-agent orchestration, and answer evaluation are out of scope.

Next Focus

• Improve question depth with stronger resume-grounded signal extraction.
• Add quality validation for duplicate/generic/ungrounded questions.
• Upgrade generation to LLM-driven depth while preserving strict output schema.

New Contributors

• @swj9707 made their first contribution in #1

Full Changelog: https://github.com/swj9707/interview-graph/commits/v0.0.0