OTHERS

ObservabilityOS — AI-Native DevOps Intelligence & Observability Platform

AI-native log analytics and incident response platform with real-time streaming, statistical anomaly detection, and multi-provider LLM diagnostics.

View Live Docs NPM

ObservabilityOS — AI-Native DevOps Intelligence & Observability Platform

Client

Personal

Timeline

2025 – Present

My Role

Sole Developer & Architect

Project Overview

ObservabilityOS is a production-grade, AI-native observability and incident response platform built as a Turborepo monorepo. It features a publishable TypeScript SDK (@observability-os/sdk) with batched async ingestion and retry logic, a statistical anomaly engine using Z-score deviation across 12 rolling time windows, a multi-provider LLM failover pipeline (AICredits → Claude → GPT-4o-mini → Mock heuristics) with circuit breakers and per-token cost tracking, real-time log streaming via Redis Pub/Sub, multi-channel alerting (Slack, Discord, Microsoft Teams) using the Adapter pattern, BullMQ job queues for async anomaly processing, SaaS billing via Stripe + Razorpay with plan-gated quota enforcement, PII scrubbing before data persistence, GitHub OAuth authentication, dual-layer caching (Redis + in-memory), and an auto-generated post-mortem report generator producing structured Markdown incident reports.

Key Features

Feature Groups

Features

1 / 0

Key Capabilities

Publishable TypeScript SDK with batched async ingestion, configurable flush intervals, and concurrency-safe retry logic

#01

Statistical Z-score anomaly detection engine across 12 rolling 5-minute windows for error rate, CPU, and latency

#02

Multi-provider LLM failover pipeline (AICredits → Anthropic Claude → OpenAI GPT-4o-mini → Mock heuristics) with per-provider circuit breakers

#03

Real-time log streaming via Redis Pub/Sub channels

#04

Multi-channel alerting (Slack Block Kit, Discord Embeds, MS Teams MessageCard) with Adapter pattern and circuit breakers

#05

BullMQ async job queue for anomaly processing with graceful Redis-absent fallback

#06

Redis sliding-window rate limiting using sorted-set atomic pipelines

#07

Dual-layer caching (in-memory L1 + Redis L2) with auto-backfill

#08

Plan-gated multi-tenant ingest API with Redis-tracked byte quotas per billing tier

#09

PII scrubbing at ingestion boundary (emails, JWTs, credit cards, auth headers, DB URIs)

#10

Multi-provider SaaS billing (Stripe + Razorpay) with subscription lifecycle management

#11

Auto-generated post-mortem reports with TTD/TTR metrics, AI narrative, and SRE checklists

#12

Morning digest email system with per-project AI summary and health stats

#13

Chaos simulator app for end-to-end failure scenario testing

#14

Case Study

Problem & Approach

The Challenge

Problem Statement

Most observability tools are glorified log viewers — they lack intelligent anomaly detection, provide no automated root-cause analysis, and cost thousands per month. Building an observability SaaS that competes with Datadog requires real-time log ingestion, statistical anomaly detection, AI-powered diagnostics, and multi-channel alerting — all while keeping latency sub-100ms and maintaining graceful degradation when dependencies fail.

The Solution

Our Approach

Architected a full observability platform in a Turborepo monorepo with 4 layers: (1) Ingestion — Next.js API routes with SHA-256 auth, Zod validation, PII scrubbing, Redis rate limiting, and MongoDB bulk insert; (2) Anomaly — Z-score engine computing rolling standard deviations over 12×5-min windows with configurable thresholds and deploy correlation; (3) AI — 4-tier LLM failover chain with per-provider circuit breakers (CLOSED/OPEN/HALF_OPEN), AbortController timeouts, and plan-aware routing; (4) Notification — Adapter-pattern dispatcher for Slack, Discord, Teams with circuit-breaker-protected webhook delivery.

Details

Project Information

Client

Personal

Timeline

2025 – Present

Role

Sole Developer & Architect

Technologies

20 Technologies

Tech Stack

20 Technologies

TypeScript

Next.js

React

MongoDB

Redis

BullMQ

Anthropic Claude

OpenAI GPT

Radix UI

Tailwind CSS

Zod

Docker

Turborepo

Razorpay

Stripe

Resend

ioredis

Recharts

Vitest

Playwright

Outcomes

Results & Impact

#01

Built and published an npm-installable TypeScript SDK (@observability-os/sdk) with concurrency-safe buffering and automatic retry

#02

Z-score anomaly engine distinguishes genuine incidents from normal traffic fluctuations — eliminating false pager noise

#03

Multi-provider AI failover ensures zero-downtime diagnostics even when individual LLM providers are unavailable

#04

Redis sliding-window rate limiting with atomic sorted-set pipelines — no naive 'reset every minute' approach

#05

Dual payment provider integration (Stripe + Razorpay) with plan-gated quota enforcement

Project Gallery

Platform Demo Video

Dashboard Overview

Platform Overview

Log Ingestion Pipeline

Anomaly Detection Engine

AI Incident Analysis

Multi-Channel Alerts

Real-Time Streaming

Dashboard Analytics

More Projects

WEB APPLICATION

Drishti Marketing OS — Enterprise WhatsApp AI Agent Platform

Enterprise WhatsApp Business inbox replacing third-party BSPs with a custom Meta Cloud API integration, multi-provider AI agent (Claude/GPT-4o/Gemini), RAG knowledge base, and real-time Zoho CRM data — achieving 89.7% AI self-service rate and 96% reduction in lead response time.

OPEN SOURCE

Sentinel — Adaptive Distributed Load Control Microservice

Open-source adaptive HTTP reverse proxy in TypeScript/Bun, sustaining ~6,800 req/s at p99 41ms baseline. Protects distributed backends from traffic spikes and cascading failures via hybrid rate limiting, distributed circuit breaking, and real-time pressure-based throttling.