RL4 Kernel — Dev Continuity System
Never lose your cognitive thread.
Captures your dev context, generates structured prompts for AI agents, and maintains your reasoning history across sessions.
🎯 What Is This?
RL4 (Reasoning Layer 4) is a VS Code extension that solves the hardest problem in software development: context loss.
When you:
- 🔄 Switch between multiple projects
- ⏸️ Take a break and come back days later
- 👥 Hand off work to teammates or AI agents
- 🧠 Can't remember "Why did I make this decision?"
RL4 captures everything you do and turns it into structured, AI-ready context.
✨ Core Features
1. Cognitive Capture Engine
Automatically records:
- ✅ Git commits (hash, message, diff summary)
- ✅ File changes (what, when, why)
- ✅ Dependencies updates (package.json, requirements.txt)
- ✅ Configuration changes (YAML, TOML, ENV)
- ✅ Test executions and results
No manual documentation required. RL4 watches silently in the background.
2. Smart Context Snapshots
Generate instant cognitive snapshots with one command:
Command: RL4 › 🧠 Where Am I? — Cognitive Snapshot
What you get:
- 📊 What you've been working on (last N commits, file changes)
- 🎯 Active goals and decisions in progress
- 🔗 Related issues, PRs, and evidence
- 📈 Pattern detection and recommendations
Perfect for:
- Returning to a project after weeks
- Handing off context to a teammate
- Providing full context to AI coding assistants (Cursor, Claude, GPT)
3. Structured Prompts for AI Agents
RL4 generates copy-paste prompts optimized for LLM agents:
# Context for AI Agent
## Current State
- Active Branch: feature/auth-refactor
- Last 5 Commits: [...]
- Modified Files: 12 (auth/, middleware/)
## Goals
- Refactor authentication middleware
- Add JWT token validation
- Update tests
## Decisions Made
- [ADR-042] Chose JWT over sessions (2025-11-10)
- Evidence: Security review, performance benchmarks
## What I Need
- Review security implications of current JWT implementation
- Suggest improvements for token refresh flow
Why this matters:
- 🚀 AI agents get perfect context without hallucinating
- 🔒 Full transparency: You see and validate everything
- 🔁 Works with any LLM (Cursor, Claude, ChatGPT, Windsurf)
4. Reasoning History & ADRs
RL4 auto-generates Architecture Decision Records (ADRs) from your work:
Example ADR:
id: ADR-042
title: "Adopt JWT for authentication"
status: accepted
date: 2025-11-10
context: |
Security review identified session-based auth as bottleneck.
Performance benchmarks showed JWT reduces DB queries by 60%.
decision: |
Migrate to JWT-based authentication with RS256 signing.
consequences: |
- Improved performance (+60% faster auth)
- Stateless authentication enables horizontal scaling
- Requires secure key management (RSA keys in vault)
evidence:
- commits: [a1b2c3d, e4f5g6h]
- issues: [#123, [#456](https://github.com/Soynido/reasoning-layer-v3/issues/456)]
- benchmarks: [perf-report-2025-11.json]
Benefits:
- 📚 Never lose the "why" behind your decisions
- 🔍 Searchable reasoning history
- 🤝 Onboard new team members 10x faster
5. Kernel-Based Architecture
RL4 runs as a cognitive kernel with:
- 🧠 Adaptive Modes: Standard, Focused, Exploratory, Free
- 💾 Append-only ledger: Immutable event history (JSONL)
- 🔐 Cryptographic integrity: SHA256 hashing, RSA signing
- 📊 Health monitoring: Auto-recovery, watchdog, diagnostics
Performance:
- ⚡ Minimal overhead (<1% CPU usage)
- 💾 Efficient storage (~10MB per 1000 events)
- 🔄 Non-blocking async architecture
🚀 Quick Start
Installation
- Install the extension:
code --install-extension reasoning-layer-rl4-3.3.0.vsix
Restart VS Code — RL4 activates automatically
Open any Git repository — Capture starts immediately
First Commands
📊 Check Status
Command Palette → RL4 Kernel: 📊 Kernel Status
See what RL4 has captured and system health.
🧠 Generate Snapshot
Command Palette → RL4 Kernel: 🧠 Where Am I? — Cognitive Snapshot
Get instant context for your current state.
🔄 Run Cognitive Cycle
Command Palette → RL4 Kernel: 🧠 Run Cognitive Cycle
Analyze patterns, detect decisions, generate ADRs.
🖥️ Open Dashboard
Command Palette → RL4 Kernel: 🖥️ Show Dashboard
Visual interface for goals, patterns, and correlations.
📂 What Gets Captured?
RL4 creates a .reasoning_rl4/ folder in your workspace:
.reasoning_rl4/
├── traces/
│ ├── file_changes.jsonl # File modifications
│ ├── git_commits.jsonl # Git history
│ └── ide_activity.jsonl # IDE events
├── adrs/
│ ├── active.json # Current ADRs
│ └── auto/ # Auto-generated proposals
├── context.json # Latest cognitive snapshot
├── patterns.json # Detected patterns
├── forecasts.json # Predictive insights
└── kernel/
└── state.json.gz # Kernel state
Privacy: All data stays local. No telemetry, no external calls.
🎮 Use Cases
1. Context Switching
Problem: You work on 5 projects simultaneously and lose track.
Solution:
# Before switching projects
RL4 › 🧠 Where Am I?
# Copy snapshot → Paste in notes
# When returning
RL4 › 🧠 Restore Context
# Instant recalibration
2. AI Agent Integration
Problem: Your AI assistant (Cursor, Claude) doesn't know your project history.
Solution:
# Generate context prompt
RL4 › 🧠 Where Am I?
# Copy output → Paste in Cursor Chat
# Agent now has full context
Result: 10x better AI suggestions, zero hallucinations.
3. Team Handoffs
Problem: Teammate takes over your work but doesn't understand decisions.
Solution:
# Generate handoff doc
RL4 › 📊 Generate Handoff Report
# Includes:
# - What was done (commits, files)
# - Why decisions were made (ADRs)
# - What's next (goals, blockers)
4. Post-Mortem Analysis
Problem: Production issue — need to understand what changed.
Solution:
# Query event history
RL4 › 📜 Show Recent Changes (last 7 days)
# See exactly:
# - Which commits went out
# - Config changes
# - Dependency updates
# - Related decisions
🧠 Adaptive Intelligence Modes
RL4 adapts its behavior based on what you're doing:
Standard Mode (Default)
- Balanced capture and analysis
- Moderate context depth
- Good for normal development
Focused Mode (High Precision)
- Deep context capture
- Maximum evidence linking
- Best for critical decisions
Exploratory Mode (Discovery)
- Pattern detection enabled
- Correlation analysis active
- Great for refactoring
Free Mode (Minimal)
- Lightweight capture only
- Low overhead
- Ideal for quick edits
Switch modes:
Command: RL4 › ⚙️ Switch Mode → [Choose mode]
📊 Technical Architecture
Kernel Components
┌─────────────────────────────────────────────┐
│ RL4 KERNEL (Extension Host) │
├─────────────────────────────────────────────┤
│ │
│ ┌─────────────────────────────────────┐ │
│ │ Capture Engines │ │
│ │ - FileWatcher │ │
│ │ - GitObserver │ │
│ │ - ConfigTracker │ │
│ │ - DependencyMonitor │ │
│ └─────────────────────────────────────┘ │
│ │
│ ┌─────────────────────────────────────┐ │
│ │ Cognitive Layer │ │
│ │ - PatternDetector │ │
│ │ - DecisionSynthesizer │ │
│ │ - ADR Generator │ │
│ │ - ContextBuilder │ │
│ └─────────────────────────────────────┘ │
│ │
│ ┌─────────────────────────────────────┐ │
│ │ Persistence Layer │ │
│ │ - JSONL Ledger (append-only) │ │
│ │ - Gzip compression │ │
│ │ - Cryptographic signing │ │
│ └─────────────────────────────────────┘ │
│ │
│ ┌─────────────────────────────────────┐ │
│ │ API Layer │ │
│ │ - UnifiedPromptBuilder │ │
│ │ - SnapshotGenerator │ │
│ │ - QueryEngine │ │
│ └─────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────┘
🔧 Advanced Features
Integrity Verification
RL4 › 🔐 Verify Integrity
# Validates cryptographic signatures
# Ensures no data tampering
Historical Reconstruction
RL4 › 🕰️ Reconstruct History
# Scans Git history (up to 1000 commits)
# Generates synthetic traces for past events
# Fills cognitive gaps from late installation
Cognitive Graph Visualization
RL4 › 📈 Show Cognitive Graph
# Visual network of:
# - Decisions → Evidence
# - Patterns → Occurrences
# - Goals → Progress
🛡️ Privacy & Security
Local-First Architecture
- ✅ All data stays on your machine
- ✅ No telemetry, no external calls
- ✅ No account required
Cryptographic Guarantees
- 🔒 SHA256 hashing for integrity
- 🔑 RSA signatures for authenticity
- 📜 Append-only ledger (immutable)
Transparent Operation
- 👁️ All captured data in plain JSONL
- 🔍 Readable with any text editor
- 🗑️ Easy to delete (just remove
.reasoning_rl4/)
📈 Metrics & Monitoring
Real-Time KPIs
Check extension status bar for:
- 📊 Events captured (total count)
- 🎯 Active goals (progress %)
- 🔗 Patterns detected (count)
- 💚 System health (0-100%)
Diagnostic Commands
# Health check
RL4 › 🏥 System Diagnostics
# Performance stats
RL4 › ⚡ Performance Report
# Storage usage
RL4 › 💾 Storage Analysis
🎯 Roadmap
Current: Phase E2 (v3.3.0)
- ✅ Kernel architecture stable
- ✅ Adaptive modes implemented
- ✅ Unified prompt builder
- ✅ Smart UI dashboard
Next: Phase E3 (Q1 2026)
- 🎯 AI agent direct integration (MCP protocol)
- 🎯 Semantic search (vector embeddings)
- 🎯 Multi-workspace support
- 🎯 Team collaboration features
Future: Phase E4+ (Q2 2026)
- 🔮 Predictive insights (forecast decisions)
- 🤖 Autonomous recommendations
- 📊 Analytics dashboard (web UI)
- 🌐 Cloud sync (optional, encrypted)
Documentation
Troubleshooting
Contributing
This is an active research project. Contributions welcome!
Philosophy: Local-first, privacy-preserving, developer-friendly.
📄 License
PROPRIETARY — Copyright © 2025 Valentin Galudec. All rights reserved.
For licensing inquiries, contact: valentin@galudec.com
🎓 Credits
Author: Valentin Galudec
Project: Reasoning Layer V4 (RL4)
Version: 3.3.0 (Intelligent Deviation Modes)
Repository: https://github.com/Soynido/reasoning-layer-v3
🌟 Why RL4 Exists
"The hardest part of software development isn't writing code — it's understanding why the code was written that way in the first place."
RL4 solves this by capturing the reasoning behind every decision and making it instantly accessible.
Result:
- 🚀 Faster onboarding
- 🧠 Better AI assistance
- 📚 Living documentation
- 🔍 Searchable decision history
- 🤝 Seamless team collaboration
Try it today. Your future self will thank you.
Last updated: November 13, 2025
