Hi, I’m Shaurya , a developer passionate about building unique projects, exploring niche ideas, and constantly learning

Shaurya Sharma  • Peerlist

Long-horizon LLM agents fail in predictable ways: they loop on the same tool calls, storm-edit one file, and flood their own context until the task collapses — and you usually notice only after the tokens are already spent.Sotis lives inside your agent's loop and catches it live.&nbsp;It detects the spiral (sliding-window entropy + exact/semantic loop detection), rolls the workspace back to a&nbsp;verified-good&nbsp;state — one proven to still parse, not just the last snapshot — distills the bloated history into a compact resumption prompt, and lets the agent keep going.Proven on real agents.&nbsp;Validated on live runs (Groq Llama-3.3-70B, OpenRouter Gemini) detecting real meltdowns and restoring clean, resumable workspaces; 100% meltdown-detection across a 6-scenario gauntlet. Detection + rollback add &lt;0.2 ms per step.Drop-in.&nbsp;pip install sotis&nbsp;→ wrap your loop in ~3 lines, or add the native LangGraph guard node. Works with OpenAI, Anthropic, DeepSeek, Google, and any OpenAI-compatible endpoint.Key features:<ul><li>Real-time meltdown detection — entropy + exact/semantic loop detection + workspace edit-storm guard</li><li>Verified-good rollback — restores a state proven safe, not just the most recent (possibly corrupted) one</li><li>Context distillation — a compact resumption prompt instead of replaying the spiral (~86% fewer tokens)</li><li>Native LangGraph guard node, plus a drop-in wrapper for custom ReAct / tool-calling loops</li><li>Any LLM provider (OpenAI / Anthropic / DeepSeek / Google / OpenAI-compatible)</li><li>Observability — Streamlit dashboard + structured JSON logs</li></ul>Community-driven roadmap&nbsp;— feature requests from people who actually ran it became shipped features.Honest scope:&nbsp;Sotis&nbsp;bounds&nbsp;the failure mode and keeps your agent resumable — it doesn't make a weak model smart. It's a library for developers building their own agents, not a plugin for closed agents (Claude Code, Codex) that expose no hook into their loop.

Sotis

Breath-SOM is a topological representation of a Self-Organizing Map (Kohonen network) designed to behave less like a static data grid, and more like a living organism.

Breath-SOM

• Engineered anomaly detection on simulated AWS Cost Explorer data, surfacing $9K+ in recoverable cloud spend and flagging mock cost spikes up to $4.5K per instance — validating the detection logic across 5 regions and 20+ resource types. • Built an automated remediation engine (stop / delete / rightsize) with audit logging and approval workflows, reducing manual cloud ops toil to a single Slack button press. • Designed a FastAPI backend handling async Slack interactions and AWS remediation triggers, integrating EC2, EBS, RDS, CloudFront, and Lambda via boto3 for real-time infrastructure control.

CloudCFO – AWS Cost Optimization & Automation

A transparent, local, explainable Linux task assistant. Local-first. No cloud. Runs entirely on your machine via Ollama. Explicit over automatic. Tex always shows what it is going to do before doing it. Explain before execute. You understand what is happening, always. Education is a feature. Tex teaches while it works. User stays in control. No background daemons. No silent actions. No raw shell from model output.

A real-time DDoS attack visualization platform featuring an interactive 3D globe, live threat metrics, and deep space aesthetics. Powered by Cloudflare Radar and AbuseIPDB data with machine learning threat classification.

ThreatPulse Monitor

A minimal AI-powered 3D scene generator for Blender — converts natural language prompts into 3D object renders. Uses a fine-tuned LLM to parse text prompts → structured JSON actions Executes actions safely inside Blender Automatically renders cinematic images No cloud dependency — runs fully locally

Tiny-LLM-3D

TUI-Asciify is a fast terminal-based tool that converts images and videos into real-time, colored ASCII art using a Textual TUI and a C-accelerated backend for smooth, high-performance rendering.