FastADK v0.2.0: Enterprise-Ready Agent Development Kit

FastADK v0.2.0: Enterprise-Ready Agent Development Kit

Release Highlights

FastADK 0.2.0 is a major release that transforms the framework into an enterprise-ready agent development toolkit. This version introduces comprehensive token & cost tracking, advanced memory management, performance optimizations, extended integrations, improved developer experience, and robust observability features.

🚀 Major Features

💰 Token and Cost Management

• Complete token usage tracking across all providers
• Cost estimation based on provider-specific pricing
• Configurable budgets with alerts and automatic cutoffs
• Detailed metrics for token usage optimization

🧠 Advanced Context and Memory

• Sliding window memory with automatic summarization
• Vector store integration for semantic memory capabilities
• Persistent storage backends (Redis, SQL) for production use
• Customizable context policies for fine-grained control

⚡ Performance Optimizations

• True async execution for all model and tool calls
• Parallel workflow execution for complex operations
• Intelligent caching to reduce redundant LLM calls
• Lazy tool execution for improved efficiency
• Batch processing utilities for high-throughput scenarios

🔌 Extensibility and Integrations

• Pluggable model provider interface for custom backends
• Dynamic plugin system for third-party extensions
• Slack and Discord adapters for easy deployment
• Multi-agent orchestration for complex scenarios
• Fine-tuning helpers for specialized models

🛠️ Developer Experience

• Interactive REPL mode for rapid prototyping
• Enhanced CLI tools for project scaffolding and management
• Comprehensive testing utilities with mocks and scenarios
• VSCode integration with snippets and settings
• Configuration validation and hot reload capabilities

📊 Observability and Monitoring

• Structured logging for all agent events
• OpenTelemetry integration for distributed tracing
• Prometheus metrics for real-time monitoring
• Sensitive data redaction for privacy compliance

📚 Documentation and Examples

• Expanded cookbook with practical recipes
• New example agents for real-world use cases
• Streamlit UI example for web interfaces
• Comprehensive API reference and conceptual guides

📋 Detailed Changelog

Added

Token and Cost Tracking

• Added TokenUsage data class to record prompt, completion, and total tokens
• Extended each provider client to capture usage from responses
• Created CostCalculator utility for estimating costs based on token usage
• Added TokenBudget component with configurable budgets and alerts
• Added structured logging for token usage and costs
• Added Prometheus metrics for token usage and cost estimation

Context Management and Memory Improvements

• Added ContextPolicy abstract class with built-in policies (MostRecent, SummarizeOlder, HybridVectorRetrieval)
• Implemented SummarizationService for LLM-based conversation summarization
• Added vector store backend integration with VectorMemoryBackend
• Implemented persistent memory backends for Redis and SQL
• Added configuration options for memory policies

Scalability and Performance Optimizations

• Improved async execution for all model and tool calls
• Enhanced Workflow class to support concurrent sub-tasks
• Implemented CacheManager with in-memory LRU and Redis options
• Added lazy tool execution logic to improve efficiency
• Created BatchUtils for common bulk operations

Extensibility and Integration

• Defined ModelProviderABC interface for custom providers
• Implemented PluginManager for discovering and loading custom modules
• Added Slack and Discord adapters for integration
• Added multi-agent orchestration API
• Included fine-tuning helper module

Developer Experience and Tooling

• Extended CLI with repl, init, and config validate commands
• Added verbose debugging with chain-of-thought capture
• Created MockLLM and MockTool classes for testing
• Added test scenario decorators for structured testing
• Added VSCode snippets and configuration support
• Implemented config reload endpoint

Observability and Monitoring

• Added structured JSON logging for all agent events
• Integrated OpenTelemetry for traces and spans
• Added Prometheus metrics endpoint
• Implemented redaction filter for sensitive data

UI/UX and Documentation

• Added cookbook, plugin guide, and performance tuning documentation
• Expanded examples folder with new agent templates
• Implemented project scaffolding with fastadk init
• Added Streamlit chat UI example
• Added community feedback links

Changed

• Improved async performance and error handling
• Enhanced memory management for long conversations
• Optimized token usage and cost calculations
• Upgraded provider integrations with the latest APIs

Fixed

• Various bug fixes and stability improvements
• Fixed memory leaks in long-running conversations
• Addressed race conditions in async workflows
• Improved error reporting and exception handling

🔧 Installation

pip install fastadk==0.2.0

Or using uv:

uv add fastadk==0.2.0

📖 Documentation

For comprehensive documentation, visit:
https://fastadk.dev

🎯 Example Usage

from fastadk.core import Agent, BaseAgent, tool

@Agent(
    model="gemini-1.5-pro",
    memory_backend="redis",
    cache_enabled=True
)
class WeatherAgent(BaseAgent):
    @tool
    def get_weather(self, location: str) -> str:
        """Get weather for a location."""
        # Implementation...
        return f"Weather for {location}: Sunny, 75°F"
    
    @tool
    def get_forecast(self, location: str, days: int = 5) -> list:
        """Get forecast for a location."""
        # Implementation...
        return [{"day": i, "forecast": "Sunny"} for i in range(1, days+1)]

# Create and run the agent
agent = WeatherAgent()
result = await agent.run("What's the weather in New York?")

❤️ Acknowledgements

Thank you to all contributors who helped make this release possible! Special thanks to our early adopters who provided valuable feedback and use cases.

🔜 What's Next

We're already planning the next release with:

• Enhanced RAG capabilities
• More third-party integrations
• Function calling improvements
• Multi-modal support

Stay tuned and join our community!