This document showcases practical implementations that demonstrate AStack's capabilities across different domains. Each example application illustrates specific architectural patterns and integration approaches, from simple text processing pipelines to complex multi-component research automation systems.
The examples demonstrate core AStack features including component composition, pipeline orchestration, agent-tool integration, and streaming data flows. Each implementation is located in the examples/ directory and represents a complete, runnable application.
AStack includes five primary example applications that demonstrate different aspects of the framework's capabilities:
| Example | Location | Primary Components | Domain |
|---|---|---|---|
| Text Processing Pipeline | examples/text-splitter/ | TextSplitter, Pipeline | Document processing |
| Research Automation Pipeline | examples/simple-deep-research/ | ContentAnalyzer, ReportEnhancer, WebDriverComponent | Information research |
| Chat Server Application | serve-astack/ | StreamingAgent, ToolInvoker | Conversational AI |
| Browser Automation Pipeline | examples/simple-browser-use/ | PlanningAgent, BrowserAgent | Web automation |
| Agent with Tools Integration | examples/agent-with-tools/ | Agent, createTool | File operations |
Sources: examples/text-splitter/index.ts1-80 examples/simple-deep-research/index.ts1-149 examples/simple-deep-research/content-analyzer.ts1-621 examples/simple-deep-research/report-enhancer.ts1-235
The text splitter example demonstrates basic pipeline usage with the TextSplitter component for document chunking and processing. This example shows how to use AStack components both in pipeline mode and standalone mode.
The example in examples/text-splitter/index.ts creates a TextSplitter component with configuration parameters and demonstrates two execution modes:
The TextSplitter is configured with chunkSize: 100, overlap: 20, and separator: /\n\s*\n/ for paragraph-based splitting. The pipeline demonstrates the addComponent and run methods for orchestration.
Sources: examples/text-splitter/index.ts47-79
| Mode | Code Pattern | Usage |
|---|---|---|
| Pipeline | pipeline.run<string[]>('textSplitter.text', sampleText) | Orchestrated execution with port routing |
| Standalone | textSplitter.run(sampleText) | Direct component execution |
The example shows both execution patterns: examples/text-splitter/index.ts66-75 demonstrating how components can operate independently or within pipelines.
Sources: examples/text-splitter/index.ts1-80
The simple-deep-research example demonstrates complex multi-component pipeline orchestration for automated research, including web scraping, content analysis, and AI-enhanced report generation.
Sources: examples/simple-deep-research/index.ts40-65
The ContentAnalyzer extends the base Component class and implements complex filtering logic for search results and content organization:
| Method | Purpose | Input/Output |
|---|---|---|
filterRelevantResults | Filter search results by relevance | SearchResultItem[] → string[] |
organizeContent | Create structured report | Map<string, string> → ResearchReport |
generateReport | Generate final research report | Multiple inputs → ResearchReport |
The component defines multiple input/output ports: searchResults, pageContents, topic (inputs) and relevantUrls, report, ready (outputs). The _transform method coordinates data flow between these ports.
Sources: examples/simple-deep-research/content-analyzer.ts83-104 examples/simple-deep-research/content-analyzer.ts465-618
The ReportEnhancer component demonstrates AI integration for content enhancement:
The component creates structured prompts for the LLM and parses JSON responses to enhance the original research report with AI-generated insights.
Sources: examples/simple-deep-research/report-enhancer.ts61-123 examples/simple-deep-research/report-enhancer.ts131-183
The serve-astack application demonstrates a full-stack implementation with Fastify backend and React frontend, showcasing streaming AI responses and intent classification.
| Component | Purpose | Integration |
|---|---|---|
StreamingMathAgent | Mathematical operations | Calculator tool + Deepseek model |
StreamingTextAgent | Text analysis | Text statistics tool + Deepseek model |
| Intent Classification | Route user requests | AI-based intent detection |
| Streaming Protocol | Real-time responses | AI SDK compatible streaming |
The server implements intent-based routing where user messages are classified and routed to appropriate specialized agents for processing.
Sources: Based on the architecture patterns from other examples and the mention of serve-astack in the overview.
The browser automation example demonstrates natural language-driven web interaction using Playwright integration with planning agents and state management.
The PlanningAgent analyzes user intent and creates step-by-step execution plans, while the BrowserAgent executes these plans through Playwright automation. The BrowserStateComponent manages DOM state and context tracking.
Sources: Based on the architecture patterns described in the overview and the component naming conventions from other examples.
This example demonstrates the zero-adaptation layer design where Agent components directly integrate with tools created via createTool without intermediate adapters.
The createTool function creates standardized tool interfaces that integrate seamlessly with Agent instances. Tools are defined with schemas and implementation functions, enabling the agent to perform complex multi-step operations.
Sources: Based on the tool integration patterns from the pipeline examples and the createTool function mentioned in the framework overview.
The examples demonstrate how AStack integrates into typical development workflows, from simple component usage to complex application architecture.
| Stage | Code Entity | File Location | Purpose |
|---|---|---|---|
| Component Creation | class CustomComponent extends Component | Custom implementations | Define reusable logic units |
| Tool Definition | createTool(name, description, implementation) | examples/*/tools/ | Create agent-callable functions |
| Model Integration | new Deepseek({ apiKey, model }) | Provider configurations | Connect to LLM services |
| Pipeline Assembly | pipeline.addComponent(), pipeline.connect() | Main application files | Orchestrate data flow |
| Execution | component.run() or pipeline.run() | Entry points | Execute standalone or pipeline mode |
Examples demonstrate consistent patterns for configuration management and environment setup across different application types.
Sources: README.md145-148 website/components/UseCases.tsx1-175
The examples showcase various integration patterns with external services, demonstrating AStack's extensibility and interoperability.
| Provider | Class | Configuration | Usage Pattern |
|---|---|---|---|
| Deepseek | Deepseek | { apiKey, model } | Direct instantiation in Agent constructor |
| OpenAI Compatible | Generic provider interface | Standard OpenAI format | Zero-adaptation integration |
The tool system demonstrates how external capabilities are integrated into the AStack ecosystem through the createTool function and standardized tool interfaces.
Examples show integration with various external services:
Sources: README.md230-260 README.md152-177 website/components/UseCases.tsx6-91
Refresh this wiki