Chakshu Gautam ChakshuGautam

Keycloak Anti-Corruption Layer for DIGIT Sandbox

Date: 2026-03-05 Status: Approved PRD: Sandbox Sign Up/Login (Aarushi, v1, 23 Feb 2026)

TL;DR

MoM Quality Comparison: Sarvam 105b vs Gemini 2.5 Pro

Transcript: 04_bichitra_jan20.mp3 — 82-entry Odia diarized transcript (Chalacchitra Jagat 50th anniversary) Generated: 2026-03-03

Stats

Metric	Sarvam 105b	Gemini 2.5 Pro
Output language	Odia script	English
Prompt tokens	5,510	~5,500

egov-workflow-v2: Hardcoded `STATE_LEVEL_TENANT_ID` Prevents Multi-Tenant Support

Problem

egov-workflow-v2 has a hardcoded state.level.tenant.id configuration property (set via STATE_LEVEL_TENANT_ID env var) that is used for MDMS master data lookups at startup and during escalation. This means the workflow service can only serve one state tenant root — any workflow definitions or MDMS data stored under a different root tenant are invisible.

Affected Code

WorkflowConfig.java:94-95 — Config injection:

DIGIT Core: Exception Handling Audit

Date: 2026-02-23 Scope: /root/code/Digit-Core and /root/code/digit-2.9lts-core-storm Total Instances Found: 40+

Overview

Category	Count	Severity

DIGIT Service Dependency Tree

Layer 0: Infrastructure (no dependencies)
├── postgres-db
├── redis
└── redpanda

Layer 1: Connection Pooling
└── pgbouncer

Solution → Document Mapping

1. SI Enablement Package

Solution	Docs
1.1 RnR of SI and Platform team	D
1.2 Onboarding Guide for SIs	D, H
1.3.1.1 Documentation and Examples	C, D
1.3.1.2 Auto-generated config files	C, B
1.3.1.3 Progressive Disclosure of configs	C, E

Country Ready Starter Pack

SI Enablement Package 1.1. Docs that define RnR of SI and Platform team 1.2. Onboarding Guide for SIs 1.3. What do SIs do? 1.3.1. Configure 1.3.1.1. Documentation and Examples 1.3.1.2. Auto-generated config files for common usecases 1.3.1.3. Progressive Disclosure of configs based on complexity

MECE Knowledge Graph System with Continuous Refactoring

A self-improving knowledge graph system that maintains MECE (Mutually Exclusive, Collectively Exhaustive) compliance through automated refactoring, semantic search, and event-driven monitoring.

Core Architecture

Data Structures

MECE-Compliant Knowledge Graph Construction: A Multi-Mechanism Approach

Abstract

We present a novel agentic framework for constructing Mutually Exclusive, Collectively Exhaustive (MECE) knowledge graphs from unstructured government scheme data. The system employs four complementary mechanisms—vector-based semantic search, dynamic synonym mapping, strict naming conventions, and post-processing consolidation—to achieve 53.6% entity compression while maintaining complete graph connectivity. Our approach demonstrates significant improvements over baseline methods, reducing duplicate entities by 60% and achieving 95+ quality scores in MECE compliance metrics.

	{
	"adminFormId": "att-001",
	"moduleId": "attendance",
	"formSchema": {
	"name": "attendance",
	"label": "Attendance",
	"subModules": [
	{
	"name": "student-attendance",
	"label": "Student Attendance",