⚡ SMARAN.AI Coder
🧠 Enterprise Cognitive Intelligence • 🔥 Autonomous Code Synthesis • ⚡ Multi-LLM Orchestration
🔥 What is SMARAN.AI Coder?
🧠 SMARAN.AI Coder is a transparent multi-LLM cognitive intelligence platform that shows you exactly which models, plugins, memory systems, and compression engines are active on every response.
Runs a live AI pipeline with 19 routing strategies, Headroom 60–90% prompt compression, Claude-Mem persistent cross-session memory, and a transparent skill execution engine — all visible in real-time execution receipts above every response.
✨ Feature Arsenal
⚡ Multi-LLM Dynamic Routing
🎯 19 Intelligent Routing Strategies → Priority, Weighted, Round-Robin, Cost-Optimized, Latency-First, Auto-Combo
| 🏷️ Provider |
🤖 Model |
🌟 Speciality |
| 🟣 DeepSeek |
V4 Pro (671B MoE) / R1 |
◈ Ultra-deep reasoning & code generation |
| 🟢 Nvidia NIM |
Nemotron 3 Ultra 70B |
◈ Hardware-accelerated inference |
| 🔵 Google |
Gemma 4 32B / Gemini 2.5 Flash |
◈ Free tier + multimodal |
| 🟡 Moonshot |
Kimi K3 Ultra |
◈ 200K token context window |
| 🔴 Anthropic |
Claude Opus 4.6 / Sonnet |
◈ Enterprise-grade reasoning |
| ⚡ Groq Cloud |
LLaMA 3.3 70B |
◈ 500+ tokens/sec ultra-fast |
| 🟠 OpenRouter |
Free Routes |
◈ Zero-cost community models |
| 🟤 Zhipu AI |
GLM 5.3 Code |
◈ Code-specialist agent |
| 💻 Local |
DeepSeek-Coder / Qwen 2.5 |
◈ Offline Ollama / vLLM |
🗜️ Headroom Compression Engine
🔋 60–90% Token Reduction → Slashes latency & cost while preserving full semantic fidelity
┌──────────────────────────────────────────────┐
│ 📥 Raw Prompt (10,000 tokens) │
│ ↓ RTK Filter Pipeline │
│ ↓ Caveman Prose Reduction │
│ ↓ Stacked Compression Modes │
│ 📤 Compressed (1,500 tokens) → 85% Saved! │
└──────────────────────────────────────────────┘
🧠 Claude-Mem Persistent Memory
💾 Cross-session cognitive memory → Remembers your architecture, past fixes, and design patterns
- 🔁 Retains debugging history across conversations
- 🏗️ Remembers project structure and tech stack decisions
- 🎯 Auto-retrieves relevant past context for new queries
- 🔗 Syncs with long-term knowledge graph
🛠️ Live Transparency Receipts
👁️ Every response shows exactly what ran → Complete execution visibility
┌─────────────────────────────────────────────────────────────┐
│ ⚡ Dynamic Auto-Combo │ 🗜️ Headroom: 65–90% Saved │
│ 🧠 Memory: Synced │ 🛠️ task-observer, ui-ux-pro │
│ 📁 Workspace: Active │ │
└─────────────────────────────────────────────────────────────┘
📎 Attach Files & Logs
🖇️ Click the paperclip icon in the prompt bar to attach code files, logs, or snapshots directly for targeted debugging
- 📄 Supports
.ts, .js, .py, .json, .html, .css, .md, .txt, .csv, .yaml, .sql, .sh
- 🖼️ Image files:
.png, .jpg, .webp
💬 Multi-Session History
📚 Manage sessions with individual delete and smart auto-titles
- ➕ New Chat → Start fresh conversation
- 🕒 Session History → Browse & restore past sessions
- 🗑️ Individual Delete → Remove specific sessions
- 🧹 Clear All → Wipe all history
🔑 BYO Provider API Keys
🔧 Configure your own keys or use the zero-cost free model catalog
| Provider |
Key Format |
Cost |
| 🟠 OpenRouter |
sk-or-v1-... |
Free + Paid tiers |
| ⚡ Groq Cloud |
gsk_... |
Free (500+ T/s) |
| 🟣 DeepSeek |
sk-... |
Pay-per-use |
| 🔴 Anthropic |
sk-ant-... |
Pay-per-use |
| 🟢 OpenAI |
sk-... |
Pay-per-use |
| 🔵 Google Gemini |
AIzaSy... |
Free + Paid |
🎯 Quick Action Chips
| Chip |
Action |
What It Does |
| ⚡ |
Explain |
Deep structural and logical breakdown of selected code |
| 🛠️ |
Refactor |
Performance, readability, and architecture optimization |
| 🧪 |
Unit Tests |
Comprehensive test generation covering all edge cases |
| 🐞 |
Fix Bugs |
Security vulnerability detection and performance leak repair |
💻 Quick Start Guide
Step 1 → Install the extension from VS Code
Step 2 → Open SMARAN.AI Coder tab in the sidebar (🧠 icon)
Step 3 → Type your coding task or click a Quick Action chip
Step 4 → Attach files with 📎 for targeted context
Step 5 → Click ⚡ Apply to File or 📋 Insert at Cursor
🔒 Security & Privacy
| 🛡️ Feature |
✅ Status |
| 🔑 API Keys stored locally only |
✅ Never transmitted |
| 📡 Zero telemetry collection |
✅ Your code stays private |
| 🔐 bcrypt password hashing |
✅ Industry standard |
| 🚫 CORS-restricted backend |
✅ Origin-validated |
| ⏱️ Rate limiting (brute-force protection) |
✅ slowapi enforced |
| 🔒 Email verification flow |
✅ Active |
👨💻 Developer
Shashwat Mishra — Full-Stack AI Engineer & Software Architect
Connect on LinkedIn
📄 License
MIT License — Free to use, modify, and distribute.
Built with ❤️ by Shashwat Mishra
Empowering developers with autonomous cognitive intelligence
