Changelog

This changelog highlights significant releases. For complete release notes, see the GitHub releases page.

v0.7.20 (Latest)

Release Date: December 7, 2024

Features

Evaluation Framework — Complete model evaluation system with benchmark, task-specific, operational, safety, and comparison evaluation types
LLM-as-Judge — Automated evaluation using LLM judges for pairwise model comparison
Evaluation UI — Multi-step wizard for creating evaluations with inputs management and run controls

Improvements

Full UI wiring to backend evaluation APIs
Test inputs persistence across evaluation runs

v0.7.x Series

v0.7.18

Evaluation Inputs & Run UI

Add test input management for evaluations
Run controls for starting and monitoring evaluation progress

v0.7.17

LLM-as-Judge Configuration

Task-specific evaluation configuration UI
Custom judge prompts and criteria definition

v0.7.16

Create Evaluation Modal

Multi-step wizard for evaluation creation
Target selection (model, stack, scenario)
Benchmark configuration

v0.7.15

Evaluation Management UI

Evaluation list view in Studio panel
Status indicators and progress tracking

v0.7.10

Evaluation Comparison Visualization

Visual comparison of evaluation results across targets
Delta highlighting for performance differences

v0.7.9

LLM-as-Judge Service

Backend service for automated model evaluation
Pairwise comparison algorithm implementation

v0.7.8

Evaluation Framework Foundation

Database schema for evaluations with migrations
Complete API endpoints for CRUD operations
Pydantic models for evaluation types, targets, and results

v0.7.7

Theme Transitions & Accessibility

Theme transition animations
Accessibility improvements for theme switching
ThemeToggle component

v0.7.6

Semantic Design Tokens

Component migration to semantic design tokens
Consistent theming across UI components

v0.7.5

Design System Tokens

Design token infrastructure for semantic theming
Color, spacing, and typography tokens

v0.7.4

Theme Infrastructure

ThemeProvider context for app-wide theme state
useTheme hook for component-level theme access

v0.7.3

Extended Training Workflows

Additional training workflow templates
Support for more training scenarios (RLHF, DPO, GRPO, SFT)

v0.7.2

UI Polish

Tool modal improvements
ToolsSection refinements
Bug fixes for UI consistency

v0.7.1

Parallelism Strategy Advisor

New tool for parallelism configuration recommendations
UI integration with Tools panel
Artifact type: parallelism_advisor

v0.7.0

Training Memory Calculator

Memory estimation for training workloads
Tools panel integration
Artifact type: memory_calculator

v0.6.x Series

v0.6.23

Model Capabilities Registry

Unified provider configuration registry
Centralized model capability definitions

v0.6.22

Reasoning Mode Configuration

Settings UI for reasoning mode selection
Extended thinking support configuration

v0.6.21

Provider Health Monitoring

Health check endpoints for all LLM providers
Automatic fallback on provider failures
API key validation from database

v0.6.20

Ollama Local Models

Ollama provider integration for local LLM deployment
Stability improvements
Bug fixes for scenario configuration

v0.6.19

Google Gemini Provider

Google Gemini API integration
Multi-model support for Gemini family

v0.6.17

Lab Panel Model Indicator

Visual indicator for active model selection
OpenAI provider fixes

v0.6.11

Studio Accessibility & GitHub Sync

Studio panel accessibility improvements
GitHub reverse sync for issue tracking

v0.6.10

Single Default Stack Enforcement

Enforce single default stack when applying blueprints
Prevent stack configuration conflicts

v0.6.9

Studio Stack Integration

Stack cards in Studio panel
Stack detail modal for configuration review

v0.6.8

Stack Integration & StackPanel Deprecation

Consolidated stack management into main interface
Deprecated separate StackPanel component
Improved stack configuration flow

v0.6.7

Studio - Scenario Integration

Scenarios now visible in Studio panel
Link artifacts to scenarios for context
Enhanced scenario-driven analysis

v0.6.6

Studio Foundation - Collapsible Sections

Collapsible sections for better artifact organization
Improved navigation for large workspaces
Foundation for upcoming Studio enhancements

v0.5.x Series

v0.5.x Highlights

Scenario System — Initial implementation of workload configuration
Stack Recommendations — AI-powered infrastructure suggestions
Blueprint Gallery — Pre-curated knowledge bundles

v0.4.x Series

v0.4.x Highlights

Hybrid Search — Combined keyword and semantic search with RRF
Citation Tracking — Numbered references in AI responses
Artifact Detection — Automatic extraction of tables and structured data

v0.3.x Series

v0.3.x Highlights

Multi-Provider LLM — Support for Anthropic, OpenAI, Google, Ollama
Source Classification — Automatic categorization of documents
Workspace Isolation — Separate contexts for different projects

v0.2.x Series

v0.2.x Highlights

LangGraph Integration — Stateful agent orchestration
Thinking Steps — Transparent AI reasoning
PDF Support — Document upload and parsing

v0.1.x Series

v0.1.x Highlights

Initial Release — Core workspace and chat functionality
URL Indexing — Web page ingestion and search
Basic UI — Three-panel layout foundation

Versioning Scheme

Lattice follows semantic versioning:

MAJOR.MINOR.PATCH

MAJOR — Breaking changes (rare, with migration guides)
MINOR — New features, backward compatible
PATCH — Bug fixes and minor improvements

Upgrade Notes

Database Migrations

New versions may include database migrations. These run automatically on startup but can be run manually:

alembic upgrade head

API Compatibility

API changes are documented in release notes. Breaking changes are introduced in new API versions with a deprecation period for the old version.

Stay Updated

Watch Releases — Star the repository and enable release notifications
GitHub Discussions — Join conversations about new features
Roadmap — See what’s coming in future releases