Skip to main content

What's New in Version 0.681.0

Released on June 7, 2026

This release introduces inference mode selection for select models, seed controls in advanced settings, TF Code enhancements with native plugins, improved diff view and smarter context management, configurable agentic tooling limits, enhanced Agents Arena with rich media support, HTML-to-PDF generation, and improved serverless model reliability! 🚀

🤖 Inference Modes for Select Models​

Fast and Priority Inference​

For select models, you can now choose between different inference modes to balance speed, stability, and cost. Both Fast and Priority modes incur a higher UoI cost than the default Standard mode.

What's New:

  • Standard (default) - Balanced performance and cost, suitable for most use cases
  • Fast - Optimised for speed. Responses are generated quickly at a higher UoI cost than Standard, ideal for high-volume or time-sensitive use cases where speed matters more than output stability
  • Priority - Optimised for stability and reliability. Responses are generated with higher priority access to compute resources, ensuring more consistent outputs and lower latency variance at a higher UoI cost than Standard. Ideal for production workloads where consistency and uptime are critical

Note: Inference modes are only available for select models. If a model does not support inference mode selection, the default mode will be used automatically. This feature is new and subject to change.

📖 Documentation: Agents - Hosting and Models

🤖 Increased Serverless Model Reliability​

More Consistent Performance Across All Models​

We've increased the reliability of all models served in serverless mode across the board. This means more consistent response quality, fewer failures, and improved uptime — regardless of which inference mode you choose.

What's Improved:

  • Higher consistency - More predictable response quality across all serverless models
  • Fewer failures - Reduced error rates and retry occurrences
  • Improved uptime - Better availability across the entire serverless model fleet

📖 Documentation: Agents - Hosting and Models

Reproducible Outputs with Seed​

Agent configuration now exposes a Seed parameter under Advanced Settings, giving you precise control over response randomness and reproducibility.

What's New:

  • Seed - Set a seed value for reproducible outputs. When the same seed and temperature are used, the model will produce more consistent responses across multiple runs. Useful for testing, debugging, or scenarios where deterministic behaviour is required

Availability: Available for all subscriptions under Settings > Agents > [Your Agent] > Advanced Settings.

📖 Documentation: Agents - Advanced Settings 📖 API Documentation: Create Agent

🎯 Configurable Agentic Tooling Limits​

Full Control Over Tool Call Thresholds​

You can now modify all three agentic tooling limit fields for operator agents, giving you full control over how agents behave during complex, multi-step tasks:

What's New:

  • Max tool calls per turn - Configure the maximum number of tool calls the agent can make in a single turn (default: 200)
  • Max same-tool calls - Configure the maximum number of consecutive calls to the same tool before triggering stuck detection (default: 20)
  • Max consecutive failures - Configure the maximum number of consecutive tool call failures before the circuit breaker triggers (default: 10)

Previously these values were fixed. Now you can fine-tune them to match your workflow needs — whether you need more room for iterative processing or tighter guardrails for production reliability.

📖 Documentation: Agents - Orchestrator Tools

💻 HTML-to-PDF Generation​

Generate PDF Documents from HTML​

Agents with code execution can now generate PDF documents from HTML content using the built-in weasyprint library. This supports full HTML/CSS rendering including layouts, fonts, and styling — ideal for programmatically generating reports, invoices, and formatted documents.

What's New:

  • HTML-to-PDF conversion - Convert HTML content to PDF directly within code execution sessions
  • Full CSS support - Layouts, fonts, colours, and styling are rendered faithfully
  • No configuration required - weasyprint is included in the default code execution environment

📖 Documentation: Execution Hooks

🖥️ TF Code Enhancements​

Native Plugins Integration​

TF Code now supports native plugins integration, allowing you to extend its capabilities with custom plugins. Plugins can add new commands, integrate with external services, or provide specialised workflows tailored to your development environment.

Better Diff View​

TF Code now provides an improved diff view when reviewing file changes. The enhanced diff display makes it easier to see exactly what was added, removed, or modified — with clearer syntax highlighting and side-by-side comparison for code changes.

Compact with On-Demand Instructions​

The compact command now accepts on-demand instructions, allowing you to tell TF Code exactly what context matters most when summarising the conversation. Instead of relying on automatic context trimming, you can specify what to preserve:

/compact Keep the API endpoint definitions and the authentication logic

This ensures that when the conversation gets long and context needs to be compressed, the information you care about is never lost.

Profile Switching​

TF Code now supports profile switching, allowing you to move between workspaces quickly without leaving the terminal. Whether you manage multiple workspaces or switch between environments, you can now swap profiles seamlessly.

📖 Documentation: TF Code CLI

🏟️ Enhanced Agents Arena​

Rich Media Battles and Artifact Review​

The Agents Arena now supports passing files, audio, images, and videos to agents during battles, enabling you to test and compare agent performance with rich media inputs — not just text. You can also review agent artifacts (generated files, images, documents) directly within the Arena interface after each battle.

What's New:

  • File uploads - Pass documents, code files, and data to both agents during a battle
  • Audio input - Upload audio files for agents with audio analysis capabilities
  • Image input - Upload images for agents with vision or image analysis enabled
  • Video input - Upload videos for agents with video understanding capabilities
  • Artifact review - View and compare generated artifacts (files, images, documents) from each agent side by side

📖 Documentation: Agents Arena

📋 Summary​

This update includes:

  • Fast and Priority Inference - Choose between speed-optimised and stability-optimised inference for select models (higher UoI cost than Standard)
  • Increased Serverless Reliability - More consistent performance, fewer failures, and improved uptime across all serverless models
  • Seed in Advanced Settings - Set seed values for reproducible outputs directly from agent advanced settings
  • Configurable Agentic Tooling Limits - All three tool call thresholds (per turn, same-tool, consecutive failures) are now user-modifiable
  • HTML-to-PDF Generation - Agents with code execution can generate PDF documents from HTML using the built-in weasyprint library
  • Enhanced Agents Arena - Pass files, audio, images, and videos during battles and review agent artifacts
  • TF Code Native Plugins - Extend TF Code with custom plugins for additional functionality
  • TF Code Better Diff View - Improved diff display with clearer syntax highlighting and side-by-side comparison
  • TF Code Compact with On-Demand Instructions - Specify what context to preserve when compacting conversations
  • TF Code Profile Switching - Move between workspaces quickly from the terminal

Thank you for using ToothFairy! We're constantly working to improve your experience. If you have any feedback or encounter any issues, please don't hesitate to reach out to our support team at support@toothfairyai.com.

Keep building magic with AI! 🧚‍♀️✨

v0.681.0