NEURAX v0.6.3

NEURAX v0.6.3

The Analytical Compiler for Neural Architectures - predict cost, memory and performance before training.

What is new in v0.6.3

• Add Multimodal (VLM) model support: new ModelType::Multimodal for vision+language models (CLIP, LLaVA-style, mobile VLMs), including parser, agent planning template and tests.
• Fix parameter calculation: fall back to ffn_dim/num_heads when intermediate_size/num_attention_heads are absent.
• Replace generic emojis with the Notionists avatar family across the UI.
• Add .cargo/config.toml pointing MLIR crates at LLVM 18.
• Add MIT LICENSE, roadmap v2.0, deployment scripts and reference model configs.

Highlights (project-wide)

• 10-pass analytical IR pipeline
• MLIR compiler backend (13 dialects, LLVM 18)
• Visual design canvas with 680+ blocks
• AI copilot agent (multi-provider)
• Inference Intelligence (22 parameters)
• Time Machine multi-year projections
• 88 reference templates across 11 architecture families

Documentation

• Design: docs/DESIGN.md
• API Reference: docs/API_REFERENCE.md
• Deployment: docs/DEPLOYMENT.md
• Changelog: CHANGELOG.md