Release v0.9.0

CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## 0.9.0 (2025-07-21)
+[Release Notes](releases/0.9.0.md)
+
+
 ## 0.8.0 (2025-07-09)
 [Release Notes](releases/0.8.0.md)
 

VERSION

@@ -1 +1 @@
-0.8.0
+0.9.0

cli/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "trainloop-cli"
-version = "0.8.0"
+version = "0.9.0"
 description = "TrainLoop command-line interface"
 authors = ["Mason Pierce <masonpierce@trainloop.ai>"]
 readme = "README.md"

releases/0.9.0.md

@@ -0,0 +1,66 @@
+Summary: Major documentation overhaul with comprehensive examples, new upgrade command, enhanced SDK reliability, and critical configuration changes
+
+## TrainLoop Evals v0.9.0
+
+### ⚠️ **BREAKING CHANGES**
+- **Configuration Default Changed**: `flush_immediately` now defaults to `true` across all SDKs and config templates
+  - **Impact**: Data will be written to disk immediately instead of being buffered
+  - **Why**: Prevents data loss when processes terminate unexpectedly or files are moved/deleted
+  - **Migration**: If you need the old behavior, explicitly set `flush_immediately: false` in your `trainloop.config.yaml`
+  - **Affected Components**: All SDK configurations, scaffold templates, and example projects
+
+### 🚀 New Features
+- **Upgrade Command**: Added `trainloop upgrade` command for seamless CLI updates via pip
+- **Comprehensive Examples**: Added full working examples for Python, TypeScript, and Go with custom metrics and evaluation suites
+- **Documentation Site**: Launched complete Docusaurus-based documentation at evals.docs.trainloop.ai
+- **Setup Script**: Created unified `./scripts/setup.py` for one-command monorepo setup
+- **Standalone Integration Testing**: Added dedicated integration test runner for Python SDK to avoid pytest import conflicts
+
+### 📚 Documentation
+- **Complete Rewrite**: Added 50+ documentation pages covering tutorials, guides, and API references
+- **Interactive Examples**: Provided runnable code samples for all SDK languages with complete evaluation workflows
+- **Architecture Guides**: Detailed explanations of system design, data flow, and component interactions
+- **Development Docs**: Comprehensive guides for building from source, testing, and contributing
+- **Integration Testing Guide**: Detailed documentation on SDK testing approaches and limitations
+
+### 🛠️ SDK Enhancements
+- **TypeScript SDK Import Validation**: Now enforces `collect()` before OpenAI/HTTP library imports with clear error messages
+- **Python SDK Integration Tests**: Added standalone test runner to resolve pytest import conflicts
+- **Enhanced Host Allowlist**: Expanded allowlist to support more LLM providers across all SDKs
+- **Improved Error Messages**: Better guidance when import order is incorrect or SDK initialization fails
+- **Type Safety Improvements**: Added `flush_immediately` to TypeScript and Python type definitions
+
+### 🔧 Configuration & Reliability
+- **Default Flush Behavior**: All SDKs now default to `flush_immediately: true` for data safety
+- **Streamlined Setup**: Single command installs all dependencies across CLI, SDKs, and UI
+- **Example Projects**: Each SDK example includes complete evaluation setup with custom metrics and proper configuration
+- **Environment Templates**: Added comprehensive .env.example files for quick configuration
+- **Enhanced Config Validation**: Better validation and error handling for configuration files
+
+### 📦 Dependency Updates
+- **Python SDK**: Updated to latest LLM library versions
+  - OpenAI: 1.50.0 → 1.97.0
+  - Anthropic: 0.58.0 → 0.58.2  
+  - LiteLLM: 1.50.0 → 1.56.5
+  - LangChain integrations and other dependencies updated
+- **TypeScript SDK**: Enhanced import detection and validation
+- **Go SDK**: Improved configuration handling and host allowlist management
+
+### 🏗️ Infrastructure & CI/CD
+- **Automated Documentation Deployment**: GitHub Pages deployment for documentation site
+- **Enhanced Testing Pipeline**: Improved CI/CD with standalone integration testing
+- **Build System Updates**: Better monorepo management and dependency handling
+- **Development Environment**: Improved setup scripts and development workflows
+
+### 🐛 Bug Fixes & Stability
+- **TypeScript SDK**: Fixed import order validation, auto-initialization behavior, and type definitions
+- **Python SDK**: Resolved integration test failures and import conflicts with pytest
+- **Documentation Links**: Corrected GitHub organization capitalization inconsistencies  
+- **CI Authentication**: Fixed GitHub Actions token authentication for docs deployment
+- **Config Type Safety**: Added missing type definitions for flush_immediately across SDK languages
+
+### 🧪 Testing Improvements
+- **SDK Integration Testing**: New standalone test runner for Python SDK resolves pytest import conflicts
+- **Test Coverage**: Enhanced test coverage for configuration handling and SDK behavior
+- **CI Reliability**: Improved test reliability and error reporting in continuous integration
+- **Documentation Testing**: Automated testing of documentation builds and deployments

runner/package.json

@@ -1,6 +1,6 @@
 {
   "name": "trainloop-studio-runner",
-  "version": "0.8.0",
+  "version": "0.9.0",
   "private": true,
   "description": "Bundled TrainLoop Studio runner",
   "bin": {

sdk/python/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "trainloop-llm-logging"
-version = "0.8.0"
+version = "0.9.0"
 description = "TrainLoop LLM Logging SDK for data collection"
 authors = ["Mason Pierce <masonpierce@trainloop.ai>"]
 readme = "README.md"

sdk/typescript/package.json

@@ -1,6 +1,6 @@
 {
   "name": "trainloop-llm-logging",
-  "version": "0.8.0",
+  "version": "0.9.0",
   "description": "TrainLoop Evaluations - header-based request tagging and zero-touch collection",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

ui/package.json

@@ -1,7 +1,7 @@
 {
   "name": "trainloop-evals",
   "private": true,
-  "version": "0.8.0",
+  "version": "0.9.0",
   "scripts": {
     "dev": "next dev",
     "build": "next build",