Synalux — Memory-Augmented AILocal + Cloud AI with 20 multimodal tools, multi-agent orchestration, Prism memory, and 12-language support. Works with Ollama (offline) or OpenRouter (cloud). HIPAA-compliant healthcare workflows. Installation Launch VS Code Quick Open ( Ctrl+P), paste the following command, and press enter.Copied to clipboard |
✦ Synalux for VS CodeThe High-Performance Clinical Development & Practice Automation Extension
Downloads
Install the VS Code extension: 🏗️ Core Architecture & Capabilities🏥 1. Unified Clinical WorkspaceMaintain operational continuity by accessing your primary practice database within the development environment.
🛡️ 2. Zero-Cloud HIPAA EnforcementSynalux prioritizes a Local-First security model to ensure Protected Health Information (PHI) remains isolated on the local machine.
⚡ 3. O(1) Cognitive Memory (Zero-Search Retrieval)Utilizing the Prism Coder engine, Synalux maintains a mathematically direct "Mind Palace" of development and clinical context.
🔬 4. Deep Research Intelligence (v11.0 Elite)The v11.0 engine introduces a multi-provider discovery pipeline that transforms the assistant into a clinical scientist.
🖥️ Quick Reference: IDE Panels (v0.15.0+)Synalux ships 4 specialized webview panels accessible via the Command Palette ( 🎨 Drag-and-Drop Builder (
|
| Category | Capabilities |
|---|---|
| Clinical | list_patients, manage_session_note, list_appointments, generate_report |
| Workspace | read_file, list_files, search_files, get_open_editors |
| Shell & Git | run_command, git_status, git_commit, git_push |
| Testing | vitest, deps_scanner, security_audit |
| Media | image_gen, tts, voice_input, ocr |
| IDE Panels | openBuilder, openAuthDB, openDevContainer, openCustomerBoard |
| Compliance | Integrated BoldSign & DocuSign templates for intake and BIP signing |
🚀 Deployment
- Installation: Search for
Synaluxin the VS Code Extensions Marketplace. - Authentication:
Cmd+Shift+P → Synalux: Sign In with Google. This synchronizes with your existingsynalux.aiprofile. - Initialization:
Cmd+Shift+P → Synalux: Open Chat. - Air-Gapped Mode: Configure Ollama for 100% on-device clinical reasoning.
🔧 Technical Specification (Admins & Engineers)
🧠 Prism Memory Engine
Synalux leverages Prism Coder to maintain a persistent, traversable graph of development history and architectural decisions.
🛡️ Role-Based Access Control (RBAC)
The system enforces 9 distinct clinical and administrative roles. Tool permissions are fail-closed; lack of verified role authorization results in immediate tool deactivation.
📊 Model Fleet & Cascade Routing
Synalux uses the prism-coder Qwen3.5 fleet for zero-cloud tool routing and local-first inference. All models score 100% on BFCL function-calling (3-seed mean). The 27B additionally scores 100% on a 15-problem coding eval.
| Model | Size | BFCL | Runs on |
|---|---|---|---|
| prism-coder:27b | 16.8 GB | 100% × 3 seeds | Mac 24GB+ |
| prism-coder:9b | 5.8 GB | 100% × 3 seeds | Mac 16GB+ |
| prism-coder:4b | 3.4 GB | 100% × 3 seeds | Mac/iPad 8GB+ |
| prism-coder:2b | 2.3 GB | 99.1% × 3 seeds | Any device |
Cascade architecture: requests route through the smallest viable model by free RAM. In chat/code modes, thinking is enabled for complex tasks. Quality gate failures (empty, think-only, truncation, loop) escalate to cloud on paid tiers.
User request → 27B (think:true) → quality gate → serve ($0)
↓ gate fail
Cloud (Claude Sonnet 4) → serve (paid tiers only)
🔍 L3 Grounding Verifier
When prism_infer is called with an evidence payload, the L3 grounding verifier automatically checks the model's response against the provided evidence. Unverified or hallucinated claims are flagged before returning to the caller — critical for clinical decision support where accuracy is non-negotiable.
🛡️ Behavioral Verification (v19)
The verify_behavior tool challenges AI agents with domain-specific scenarios before they edit behavioral code. 17 built-in domains (KDS, billing, auth, voice ordering, clinical, HR, accounting, and more) catch bad edits before they happen — like an ABA antecedent intervention for code quality.
Works automatically via skill injection — no hooks or local configuration needed. Custom domains can be added per workspace on Enterprise plans.
📊 Plans & Pricing
| Free | Standard $19/mo | Advanced $49/mo | Enterprise $99/mo | |
|---|---|---|---|---|
| Seats included | 1 | 1 | up to 5 | up to 25 |
| Local model ceiling | up to 4b | up to 9b | up to 27b | up to 27b |
| Daily inference limit | 50 | 200 | 2,000 | 100,000 |
| Max output tokens | 512 | 1,024 | 2,048 | 4,096 |
| Cloud fallback | — | Claude Sonnet 4 | Claude Sonnet 4 | Priority + Sonnet 4 |
| L3 grounding verifier | — | ✓ | ✓ | ✓ |
| Behavioral verification | advisory | 17 domains | 17 domains | custom domains |
| Knowledge search | limited | unlimited | unlimited | unlimited |
| Session memory | limited | unlimited | unlimited | unlimited |
| Analytics dashboard | — | ✓ | ✓ | ✓ |
| HIPAA BAA | — | — | — | ✓ |
All plans include full local Ollama inference (offline, $0 cost), 30+ IDE tools, clinical workspace, and the Prism memory engine. 14-day free trial on all paid plans.
Why Synalux vs Competitors
| Feature | Synalux | GitHub Copilot | Cursor | Windsurf | Amazon Q |
|---|---|---|---|---|---|
| Local inference (2B–27B) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Works offline (local-only mode) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Open-weight models (HuggingFace) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Data stays on machine (local tier) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Persistent cross-session memory | ✅ | ✅ | ❌ | ❌ | ❌ |
| Cognitive routing (episodic/semantic) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Session drift detection (HRR) | ✅ | ❌ | ❌ | ❌ | ❌ |
| L3 grounding verifier | ✅ | ❌ | ❌ | ❌ | ❌ |
| Behavioral verification (pre-edit) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Clinical workspace (HIPAA/SOAP/BIP) | ✅ | ❌ | ❌ | ❌ | ❌ |
| MCP server (tools + memory for agents) | ✅ | ❌ | ❌ | ❌ | ❌ |
| Web IDE | ✅ | ✅ | ❌ | ❌ | ✅ |
| Flat-rate pricing (not per-seat) | ✅ | ❌ | ❌ | ❌ | ❌ |
License
BSL-1.1
© 2024–2026 Dmitri Costenco. All rights reserved.