campus-placement-engine / docs /RELEASE_NOTES.md
deeep-sec's picture
feat: institutional campus placement eligibility & interview scheduling engine
788eab4 verified
|
Raw History Blame Contribute Delete
5.03 kB

Institutional Placement Engine: Release Notes (v1.0.0)

Release Version: 1.0.0-institutional
Target Specification: Problem Core Engine ("One student, three drives, one Tuesday")
Deployment Mode: 100% Offline, Air-Gapped Local Area Network (LAN)


Highlights & Capabilities

1. Pure-Domain Academic Rule Engine & AST Evaluator

  • Typed JSON AST: 13 rule node primitives (CGPAMinRule, PercentageMinRule, BranchInRule, ActiveBacklogsMaxRule, HistoricalBacklogsMaxRule, GapYearsMaxRule, AgeMaxRule, SemesterMinRule, ProgramInRule, CustomFieldRule, AndRule, OrRule, NotRule).
  • High-Throughput Determinism: Evaluates 1,800 students against 10 recruitment drives (18,000 evaluations) in 0.231 seconds (24 ms per drive).
  • Two-Tier Explanations: Generates concise, empathetic plain-English single-line summaries for students alongside complete source-traced execution logs for grievance committees.
  • Cryptographic Audit Snapshot: Generates SHA-256 hashes over candidate records and rule ASTs for bit-for-bit replayability.

2. Multi-Policy Institutional Offer Enforcement

  • Canonical Policies Included:
    1. Strict One-Offer Policy with Dream Tier exception ($\ge$ 10.0 LPA).
    2. Progressive Salary Multiplier Policy (1.5x minimum upgrade rule).
    3. Tier-Based Quota Policy (Standard, Dream, Super-Dream quotas).
  • Configurable Ambiguity Toggles: Explicit toggles for inclusive threshold boundaries, fixed pay vs total CTC, off-campus offer inclusion, declined/revoked offer handling, and cooling periods.
  • Dispute Explainer: Automated side-by-side comparative explainer resolving candidate grievances in writing without coordinator bias.

3. Deterministic Announcement Extraction Pipeline

  • Multi-Format Parsing: Extracts requirements from pasted email text, .eml, digital/tabular PDFs (pdfplumber), and Microsoft Word .docx documents.
  • Accuracy Benchmark: Achieved 25 of 25 announcements (100.0%) extracted with zero manual re-typing, exceeding the target threshold ($\ge$ 20/25).
  • Zero Cloud & Zero AI Hallucination: Pure deterministic pattern matching operating at 14.6 ms/document offline.

4. OR-Tools CP-SAT Clash-Free Scheduler

  • Constraint Satisfaction Model: Discretized 15-minute slot intervals with hard NoOverlap constraints on rooms, company panels, and candidates.
  • Benchmark Performance:
    • Baseline (600 students / 12 drives / 8 rooms): 1.05 seconds (Target: < 60s).
    • Stress Case (1,800 students / 40 drives / 20 rooms): 5.04 seconds (Target: < 120s).
  • Physical Buffer Enforcement: Automatically schedules 15-minute inter-block walking buffers between consecutive student interviews.

5. What-If Impact Engine & Publish Workflow

  • Interactive Simulation: Simulates prospective cut-off and schedule changes in 18 ms with zero database side effects.
  • Data-Driven Dynamic Headlines: Auto-generates explanatory headlines (e.g., "Lowering cut-off from 7.0 to 6.5 adds 38 students; 6 clash with Northwind test at 10:00").
  • Immutable Version Snapshots: Official publish workflow with one-click rollback capabilities.
  • Enforcement ChangeSets: Automatically propagates hiring results to cancel downstream clashes and free interview panels.

6. Air-Gapped Security & LAN Infrastructure

  • Single-Port Serving: Boots on port 8000 serving both REST API endpoints and bundled React SPA assets. Automatically detects and displays student Wi-Fi URLs.
  • Strict Zero-Outbound Air-Gap: Automated socket interception test asserts zero external connections.
  • Cryptographic Protections: Argon2id password hashing, AES-256 Fernet field-level encryption, PBKDF2 passphrase-encrypted .enc database backups, and one-click institutional wipe.
  • Scoped Student Access: Candidates authenticate via roll number + OTP, accessing solely their own status and clash-free timetable.

Verification & Audit Metrics Summary

Verification Suite Target Result Status
Backend Test Suite (Pytest) 100% pass 149 / 149 tests passed PASSED
Hidden Edge Cases Regression 60 scenarios 60 / 60 scenarios passed PASSED
Air-Gap Guardrail Tests 0 outbound connections 5 / 5 socket intercept tests passed PASSED
Code Style & Linting (Ruff) 0 errors 0 errors / 0 warnings PASSED
Static Type Checking (Mypy) 0 errors across 113 files 0 errors (Strict Mode) PASSED
Frontend Production Build Clean build 1,973 modules transformed (dist/) PASSED
End-to-End Suite (Playwright) 100% pass 4 / 4 tests passed (16.7s) PASSED

Installation & Launch

# Windows
start.bat

# Linux / macOS
./start.sh

Access the application at http://localhost:8000 or your local LAN address printed in the terminal.