Skip to content

v1.0.7 — Markdown Data Tables, Gemini Flash Defaults & Self-Healing Sessions

Choose a tag to compare

View all tags
@Aniket-a14 Aniket-a14 released this 30 Aug 11:05
· 126 commits to master since this release
v1.0.7
346c1a7

🧙‍♂️ Wizard v1.0.7 Release Notes

Wizard v1.0.7 delivers essential reliability enhancements, rich Markdown data table rendering, KaTeX mathematical typesetting, real-time data mode synchronization across sessions, enhanced CLI workspace auto-discovery, default Gemini Flash frontier model acceleration, automatic code fence & stray language header stripping, intelligent LLM error sanitization, self-healing WebSocket handshakes, and comprehensive security hardening.

🚀 Key Improvements in v1.0.7

1. 🖥️ CLI Workspace Auto-Discovery & Global Homebrew Execution

  • Multi-Vector Workspace Resolution: The wizard CLI (wizard init, wizard start, wizard status, wizard skills, etc.) automatically detects your Wizard repository across:
    1. WIZARD_ROOT / WIZARD_HOME environment variables.
    2. Current working directory and any parent directory.
    3. Executable install directory (Homebrew Cellar / opt / share).
    4. Standard user project workspaces (~/Projects/Wizard-w2, ~/.wizard).
  • Actionable Guidance: Clear multi-step instructions if run in an unrecognized directory.

2. 🧮 KaTeX Mathematical Typography & Scientific Formula Rendering

  • Full Mathematical Rendering: Integrated KaTeX in the chat interface for both inline equations ($r = 0.5360$, $R^2 = 0.5616$) and display equation blocks ($$...$$).
  • Statistical Notation Support: Regression statistics, correlation coefficients, $p$-values, and matrix notations render cleanly.

3. 🛡️ Grounding Verification for Scientific & Statistical Notation

  • Scientific Notation Parsing: Grounding engine extracts and validates LaTeX scientific notation expressions ($4.3650 \times 10^{-94}$) without misidentifying exponents as uncomputed numbers.
  • Inequality Bound Alignment: Supports hypothesis testing thresholds ($p < 0.0001$ against $0.0000$) to eliminate false-positive uncomputed figure warnings.

4. 🌐 Data Mode & Runtime Privacy Synchronization

  • Real-Time Session Sync: Switching between Local-Only, Cloud-Only, and Hybrid modes immediately synchronizes across active sessions, .env, and the Models Workbench.
  • One-Click Mode Switcher: When browsing models in the Models Workbench, one-click switching is provided directly on restricted provider cards.
  • CORS Preflight Hardening: Configured PATCH and HEAD in CORSMiddleware, enabling browser settings management from external frontends.

5. 📊 Markdown Table Rendering Engine

  • Full Data Table Rendering: Fixed a block parser bug where table headers preceded by paragraphs without explicit blank lines were swallowed into plain text. Tables now parse accurately and render with zebra striping, sticky headers, and tabular numeric fonts.
  • Descriptive Statistics Presentation: Summary statistics, data quality triage matrices, distribution quantiles, and missing value indicators render in high-contrast data tables.

6. ⚡ Frontier System Prompts & Code Repair

  • 6-Stage Analytical Scaffolding: Prompts upgraded across Data Cleaning, Quantitative Python Worker, Hypothesis Planning, Step Decision, Independent Verification, and Answer Synthesis.
  • Automatic Code Fence Stripping: Enhanced CodeGuard.repair() and orchestrator._extract_code() to cleanly strip markdown backtick blocks and stray language identifiers (python\n, py\n) on line 1, eliminating self-healing retry loops and reducing turn execution time from 2+ minutes down to seconds.

7. 🧠 Gemini Flash Model Prioritization

  • Zero-Friction Free Tier Defaults: Default model recommendation heuristics prioritize gemini-2.5-flash and gemini-3.7-flash over quota-exhausted preview models, ensuring fast out-of-the-box performance on standard Google AI Studio API keys.
  • Frontier Multimodal Reasoning: Flash models automatically inherit "reasoning", "code", and "vision" role capabilities.

8. 🔄 Self-Healing WebSocket Handshakes

  • Session Auto-Minting: When connecting to /ws/chat with an expired or invalid session token, the server transparently mints a fresh session and transmits the session frame instead of hard-rejecting with WebSocket 1008.
  • Auto-Clearing Stale Tokens: Frontend automatically purges expired localStorage session references upon receiving 404 or connection resets, preventing infinite reconnect loops.

📦 Verified Release Binaries & SHA-256 Checksums

Platform Architecture SHA-256 Checksum
macOS Apple Silicon (arm64) 89f1cc063ffa820fc59cdeb3ea1674bb60628ac46b9dfb95981ee674bc723056
macOS Intel (amd64) 2bfdf4bc6245421b806953c99fcd86e038549331c77476f5d35907c713dc30d0
Linux AMD64 (x86_64) 01830a2d658392207bb83666b50c244ee004d89be3e4ee53981d96fa9cc9c3ac
Linux ARM64 (aarch64) c5915d8a468aa177b3456bf00dcae994dbf87019a5063dbf7c9a51cd9e412eee
Windows AMD64 (x86_64) 96c2574e776bb48a6e74f1bef49e7a24f5404e99e8eb69cc2e542d3a15a46ac7

💻 Quick Installation

Homebrew (macOS / Linux)

brew tap Wizard-AIA/homebrew-wizard
brew install wizard

One-Line Install Script

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/Wizard-AIA/Wizard-w2/master/scripts/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/Wizard-AIA/Wizard-w2/master/scripts/install.ps1 | iex
Assets 9
Loading