How crpf-mitadt/hiring-screening-engine works
Chandra Eval is a specialized resume/JD screening engine, not the recruitment platform. Its main workflow accepts a JD and resume, extracts structured records (using full-document Chandra OCR for raw resume PDFs and Gemma-based structuring), evaluates eligibility and evidence deterministically, and writes decision-support reports that require human review. The graph preserves the documented OCR, safety, provenance, caching/artifact, and reporting capabilities. Some links are supported by the SOURCE INDEX’s import references rather than inspected call sites; unsampled implementation details are not asserted.
Main entry points: match_workflow.py Match workflow · engine.py Deterministic scoring · resume_structuring.py Resume structuring · candidate_evaluation.py Candidate evaluation
Components 26
Workflow entry
- Match CLI
src/chandra_eval/cli.py - Match workflow
src/chandra_eval/pipeline/match_workflow.py - Runtime settings
src/chandra_eval/config.py
Document processing
- JD ingestion
src/chandra_eval/ingestion/jd_document.py - JD structuring
src/chandra_eval/pipeline/jd_structuring.py - Full-document OCR
src/chandra_eval/ocr/full_document.py - Chandra client
src/chandra_eval/ocr/chandra_client.py - Resume ingestion
src/chandra_eval/ingestion/resume_document.py - Resume structuring
src/chandra_eval/pipeline/resume_structuring.py - Gemma client
src/chandra_eval/llm/gemma_client.py - Candidate sanitizer
src/chandra_eval/safety/candidate_sanitizer.py
Evaluation
- Typed schemas
src/chandra_eval/schemas/common.py - JD schema
src/chandra_eval/schemas/jd.py - Resume schema
src/chandra_eval/schemas/resume.py - Evaluation schema
src/chandra_eval/schemas/evaluation.py - Candidate evaluation
src/chandra_eval/pipeline/candidate_evaluation.py - Deterministic scoring
src/chandra_eval/scoring/engine.py - Eligibility checks
src/chandra_eval/scoring/eligibility.py - Evidence matching
src/chandra_eval/scoring/matching.py - Scoring dimensions
src/chandra_eval/scoring/dimensions.py - Evidence provenance
src/chandra_eval/scoring/provenance.py
Results and configuration
- Report writers
src/chandra_eval/reports - Run artifacts
Other components
- Engineer / reviewer
- Chandra OCR server
- Gemma service
Connections 36
- Engineer / reviewer → Match CLI: invokes
README.md - Match CLI → Match workflow: dispatches
src/chandra_eval/cli.py - Match CLI → Runtime settings: loads settings
src/chandra_eval/cli.py - Match workflow → Runtime settings: reads settings
src/chandra_eval/pipeline/match_workflow.py - Match workflow → JD ingestion: loads JD
src/chandra_eval/pipeline/match_workflow.py - Match workflow → JD structuring: requests structuring
src/chandra_eval/pipeline/match_workflow.py - Match workflow → Full-document OCR: runs OCR
src/chandra_eval/pipeline/match_workflow.py - Full-document OCR → Chandra client: requests pages
src/chandra_eval/ocr/full_document.py - Chandra client → Chandra OCR server: sends OCR request
README.md - Match workflow → Resume ingestion: loads resume
src/chandra_eval/pipeline/match_workflow.py - Match workflow → Resume structuring: requests structuring
src/chandra_eval/pipeline/match_workflow.py - JD structuring → JD ingestion: reads document
src/chandra_eval/pipeline/jd_structuring.py - JD structuring → Gemma client: requests extraction
src/chandra_eval/pipeline/jd_structuring.py - Resume structuring → Resume ingestion: reads OCR artifacts
src/chandra_eval/pipeline/resume_structuring.py - Resume structuring → Gemma client: requests extraction
src/chandra_eval/pipeline/resume_structuring.py - Gemma client → Gemma service: sends model request
- Resume structuring → Candidate sanitizer: sanitizes profile
src/chandra_eval/pipeline/resume_structuring.py - Resume ingestion → Candidate sanitizer: builds safe profile
src/chandra_eval/ingestion/resume_document.py - JD structuring → JD schema: validates output
src/chandra_eval/pipeline/jd_structuring.py - Resume structuring → Resume schema: validates output
src/chandra_eval/pipeline/resume_structuring.py - Candidate evaluation → JD ingestion: loads JD
src/chandra_eval/pipeline/candidate_evaluation.py - Candidate evaluation → Resume ingestion: loads candidate
src/chandra_eval/pipeline/candidate_evaluation.py - Candidate evaluation → Deterministic scoring: evaluates candidate
src/chandra_eval/pipeline/candidate_evaluation.py - Match workflow → Candidate evaluation: runs evaluation
src/chandra_eval/pipeline/match_workflow.py - Deterministic scoring → Eligibility checks: checks eligibility
src/chandra_eval/scoring/engine.py - Deterministic scoring → Evidence matching: matches evidence
src/chandra_eval/scoring/engine.py - Deterministic scoring → Scoring dimensions: scores dimensions
src/chandra_eval/scoring/engine.py - Deterministic scoring → Evidence provenance: tracks evidence
src/chandra_eval/scoring/engine.py - Deterministic scoring → Evaluation schema: produces result
src/chandra_eval/scoring/engine.py - Typed schemas → JD schema: supports
src/chandra_eval/schemas/jd.py - Typed schemas → Resume schema: supports
src/chandra_eval/schemas/resume.py - Typed schemas → Evaluation schema: supports
src/chandra_eval/schemas/evaluation.py - Candidate evaluation → Report writers: writes reports
src/chandra_eval/pipeline/candidate_evaluation.py - Match workflow → Report writers: writes run reports
src/chandra_eval/pipeline/match_workflow.py - Report writers → Run artifacts: writes JSON CSV HTML
README.md - Engineer / reviewer → Run artifacts: reviews evidence
README.md