G
GAUNTLET HACKATHON MASTER DECK
CRC-32 PASS
100% Python STDLIB Zero Dependencies LSM-Tree Storage 6-Phase Agentic Crawler

High-Density Temporal Telemetry & Autonomous Observability Core

GAUNTLET is an authoritative time-series storage, mathematical anomaly intelligence, and live web probing platform built 100% from first principles without external databases or bloated frameworks.

01. Storage
Binary WAL
CRC-32 Framed Disk Log
02. Indexing
Bloom Filter
Sub-ms Segment Pruning
03. Language
Custom DSL
AST Query Parser
04. Analytics
Z-Score & r
Pearson Correlation
05. Live Crawler
6-Phase Agent
Multi-Socket HTTP Jitter
Team GenWin: Swati Dubey & Nitanshu Tak
github.com/swaskiee/Gauntlet
01. Industry Contrast

The Bloat Problem vs. First Principles

Conventional Heavyweight Observability
Massive Supply Chain Risk: Heavy stacks require hundreds of packages (Pandas, FastAPI, Redis, SQLAlchemy), causing version conflicts and CVE vulnerabilities.
Ephemeral Memory State: Metrics held strictly in RAM vanish during unexpected server crashes.
Opaque AI Heuristics: Black-box anomaly alerts provide zero explainable mathematical justification for on-call teams.
Outcome: High memory footprint, fragile deployments, unexplainable alerts.
The GAUNTLET First-Principles Standard
Strict Zero-Dependency Guarantee: Built strictly using native Python standard library modules (`struct`, `zlib`, `hashlib`, `math`, `http.server`).
Crash-Proof Binary WAL: Append-only length-prefixed disk records protected by hardware-accelerated CRC-32 checksums.
White-Box Statistical Mathematics: Explainable Z-Score standard deviation deviation bands and Pearson correlation matrices ($r$).
Outcome: Sub-millisecond writes, zero CVE risk, pure mathematical explainability.
Verified on Python 3.11+ without installing a single pip wheel. 100% ZERO-DEP
02. Storage Tier

LSM-Tree Storage Engine & Binary Protocol

01. Buffer SHA-256 Deterministic ID Deduplication
02. WAL Disk wal.gtwl CRC-32 Framed Disk Persistence
03. Memory Memtable Sorted In-Memory Cache
04. Disk Files *.gt Immutable Segments
05. Query Bloom Filter Multi-Hash Segment Pruning
// GAUNTLET Length-Prefixed Binary Struct Definition (`struct >IIQ...`)
[ "GT01" Magic: 4B ] [ Length: 4B ] [ CRC-32: 4B ] [ Epoch: 8B ] [ SHA-256 Hash ID: 24B ] [ Payload: N Bytes ]
✓ Hardware-accelerated CRC-32 validates byte integrity. Replays and repairs upon unexpected power interruption.
03. Autonomous Intelligence

The 6-Phase Agentic Pipeline

I. POWER (Socket Probe)
Multi-Socket HTTP Jitter

Probes raw network sockets for DNS, TLS, round-trip latency, and jitter variance.

II. SPACE (Storage WAL)
Binary WAL Persistence

Appends raw HTML payloads directly to disk with CRC-32 checksums.

III. REALITY (Bloom DOM)
DOM Parse & Bloom Index

Crawls <title>, H1, meta tags, and indexes route links into Bloom filters.

IV. MIND (AST Normalizer)
AST Telemetry Synthesis

Transforms raw responses into structured metrics: Latency, Size, CPU, and Heap RAM.

V. TIME (Waveforms)
Sliding Time Windows

Reconstructs historical time-series epochs to measure performance drift and state shifts.

VI. SOUL (Diagnostics)
Z-Score & Correlation

Applies $z$-score thresholds ($|z| \ge 2.0$) and Pearson matrices ($r$) for instant root-cause diagnosis.

Probes any URL (GitHub, Hacker News, Wikipedia, WhatsApp) or localhost. Probe Speed: 6ms - 200ms
04. Self-Sustained Target

Aura Botanica: Our Standalone Target

Why We Built Our Own Target Website

To ensure GAUNTLET is 100% self-sustaining and zero-dependent without relying on third-party APIs that could throttle or block scrapers during evaluation, we designed and deployed a live mock luxury café site.

URL: https://swaskiee.github.io/Gauntlet/
✓ Deployed independently to GitHub Pages.
✓ Provides rich DOM structures, menus, cart states, and 41 route links.
Decoupled architecture: Both sites deployed independently.
AURA BOTANICA — Artisanal Roastery HTTP 200 OK
Payload
25.1 KB
Routes
41 Links
Latency
~6.2 ms
• Server: GitHub Pages (Fastly CDN / Nginx)
• Security: HTTPS Enforced (HSTS Active)
WAL Ingest: wal.gtwl [COMMITTED] CRC32: 0x8a92f1b4
Enables testing of live latency, parsing, and anomaly detection with zero friction. SELF-SUSTAINED
05. Statistical Science

White-Box Mathematics from First Principles

1. Z-Score Anomaly Engine

|z| ≥ 2.0
z = (x - μ) / σ

Measures how many standard deviations ($z$) an observation ($x$) deviates from historical baseline mean ($\mu$). Automatically adapts to server trends instead of arbitrary static thresholds.

Flags outliers beyond the 95% confidence interval ($2\sigma$).

2. Pearson Correlation Matrix ($r$)

-1.0 ≤ r ≤ +1.0
r = Σ((x - x̄)(y - ȳ)) / (σ_x · σ_y)

Evaluates pairwise linear relationships between metrics. A correlation of $+0.94$ between Page Size and CPU Usage mathematically proves causal degradation rather than random variance.

Provides explainable root-cause attribution without heavy ML frameworks.
Quantiles Computed: Min • Median (P50) • P90 • P95 (Tail Latency) • Max Peak. Welford's Single-Pass Math
06. Developer Tooling

Handcrafted DSL Compiler & Query Planner

GAUNTLET Query Planner IDE Execution: 0.42 ms
FIND latency, cpu, memory
FROM 'server-42'
WHERE value > 50.0 AND region = 'us-east'
ORDER BY timestamp DESC
LIMIT 25
Execution Plan:
• Segments Checked: 14 | Pruned by Bloom: 13 (92.8% I/O Skipped)
• Scanned Rows: 25 | Matched: 25
1. Lexer & Tokenizer

Converts query strings into typed semantic tokens without regex backtracking.

2. AST Parser

Builds hierarchical grammar trees to validate projections and filters.

3. Bloom Pruning

Rejects non-matching segment files on disk, cutting physical reads by >90%.

Zero SQL engines used. 100% handcrafted parser and executor. <1ms SUB-MILLISECOND
07. Dashboard UI

Cross-Entity Multi-Metric Graphs

CPU Trend 41.2%
[ CPU Waveform ]
Mean: 41.2% • Std: 1.8%
Memory Heap 48.5%
[ Heap Waveform ]
Median: 48.5% • V8 Heap
Socket Latency 6.2 ms
[ Latency Waveform ]
P95: 8.4ms • Min: 5.1ms
Error Rate 0 err
[ Error Waveform ]
Baseline: 0.00 err/sec
Export CSV Multi-Entity Switch Zero-Leak Auth
Interactive Canvas Core
08. Cloud Deployments

Public Production Footprint

Edge Network VERCEL
Frontend Dashboard

Global edge caching, zero-leak guest auth, mobile drawer navigation, and reverse proxy.

gauntlet-gules.vercel.app ↗
Linux Engine RENDER
Python Core & REST API

Persistent binary WAL, crawler dispatcher, statistical engine, and 9 REST endpoints.

gauntlet-telemetry.onrender.com ↗
Live Target PAGES
Aura Botanica Café

Standalone live café website providing authentic real-world DOM structures and routes.

swaskiee.github.io/Gauntlet ↗
All 3 services are independently deployed and communicating in real-time. 100% LIVE AUDITED
09. Engineering Rigor

Verification, Benchmarks & Durability

26 / 26
Unit Tests Passing
Runtime: 0.48s Total
0
External Packages
100% Python STDLIB
< 1 ms
Query Execution
Bloom-Pruned Speed
0 B
Data Loss On Crash
WAL Replay Verified
$ python -m unittest discover tests -v
test_wal_framing ... ok | test_crc32_corruption ... ok | test_bloom_filter ... ok | test_zscore ... ok
Ran 26 tests in 0.487s — OK [100% Contract Compliance]
10. Compliance

Official Submission Package

STDLIB.md

Detailed standard library justification with line-by-line reasoning per module swap.

Makefile

One-command automated test runner: make test, make serve, make verify.

.zero-dep.toml & deps-proof.txt

Structured zero-dependency verification manifest and native module proof.

requirements.txt

Clean empty manifest certifying zero third-party dependencies.

Official Form: https://tally.so/r/eqW4RO ALL REQUIREMENTS MET
Forging Data Into Truth

Thank You from Team GenWin!

GAUNTLET proves that high-performance, crash-resilient temporal storage and statistical observability can be engineered from first principles with zero runtime bloat.

Swati Dubey
Statistical Analytics & Presentation Lead
Nitanshu Tak
Core Engine & Storage Architecture Lead
GAUNTLET — 100% Zero-Dependency Temporal Telemetry Engine & Observability Platform