Skip to content

05-reference-apps.md

Latest commit

 

History

History
70 lines (53 loc) · 3.85 KB

05-reference-apps.md

File metadata and controls

70 lines (53 loc) · 3.85 KB

Reference Application Field Guide

This guide links the refreshed documentation set to the live applications in the repository. Use it to decide which codebase to inspect for a given scenario.

Deep Research App

  • Goal: Structured research workflows for engineers and analysts
  • Key Patterns: YAML-defined sequential + concurrent phases, optional plan review
  • Agents: Planner, Researcher (web/search), Synthesizer, Reviewer
  • Data Sources: Tavily or other MCP-enabled search services, Azure OpenAI for synthesis
  • Docs: deep_research_app/README.md, deep_research_app/docs/QUICKSTART.md, decision/pattern guides in docs/hackathon
  • Why reference it: Best choice for showing YAML workflows, concurrent fan-out, and resume-able research sessions

Patterns Sandbox

  • Goal: Demonstrate every MAF builder with real-time visualization
  • Key Patterns: Sequential, Concurrent, Group Chat, Handoff, Magentic, Deep Research
  • Agents: Generic helpers that echo patterns, plus sample MCP integrations
  • Docs: patterns/README.md
  • Why reference it: Fastest playground for experimenting with orchestration settings before copying them into other apps

Advisor Productivity App

  • Goal: Transform meeting transcripts into insights, recommendations, and follow-up packages
  • Key Patterns: Sequential flow with optional branching for sentiment and action items
  • Agents: Transcription orchestrator, summarizer, recommendation writer, sentiment analyzer
  • Services: Azure Speech, Azure OpenAI, Cosmos DB, Application Insights, Content Safety
  • Docs: advisor_productivity_app/README.md
  • Why reference it: Strong template for combining speech services with human-in-the-loop validation and telemetry

Multimodal Insights App

  • Goal: Analyze audio, video, and documents in a single workflow with plan approvals
  • Key Patterns: Planner + Sequential execution, optional parallel ingestion per file type
  • Agents: Planner, Multimodal Processor, Sentiment, Summarizer, Analytics
  • Services: Azure Speech, Document Intelligence, Azure OpenAI, Cosmos DB
  • Docs: multimodal_insights_app/README.md, multimodal_insights_app/docs/ARCHITECTURE.md, docs/MAF_PATTERN_INTEGRATION.md
  • Why reference it: Exemplifies multimodal preprocessing and agent coordination with Cosmos persistence

FinAgent App (Static)

  • Goal: Equity research copilot with predefined execution modes
  • Key Patterns: Sequential, Concurrent, Group Chat
  • Agents: Company, SEC, Earnings, Fundamentals, Technicals, Report
  • Services: FMP API, SEC EDGAR, Azure OpenAI, optional PDF export
  • Docs: finagent_app/README.md, finagent_app/docs/QUICKSTART.md
  • Why reference it: Shows how to blend financial data sources with multiple coordination strategies

FinAgent Dynamic App

  • Goal: Human-approved dynamic planning over the same financial domain
  • Key Patterns: ReAct planner with approval workflow, Sequential execution, Synthesis dual-context pattern
  • Agents: Planner, Company, SEC, Earnings, Fundamentals, Technicals, Summarizer
  • Services: Azure OpenAI, Cosmos DB, Yahoo Finance MCP, FMP API
  • Docs: finagent_dynamic_app/README.md, finagent_dynamic_app/docs/QUICKSTART.md, finagent_dynamic_app/docs/SYNTHESIS_AGENT_PATTERN.md
  • Why reference it: Teaches how to add approval gates, persist plan history, and manage session-wide context

Choosing the Right Starting Point

Objective Start Here
Learn orchestration patterns quickly Patterns Sandbox
Build a research assistant Deep Research App
Showcase multimodal analytics Multimodal Insights App
Deliver financial intelligence FinAgent or FinAgent Dynamic
Handle meeting intelligence Advisor Productivity App

Next: use 06-development-guide.md to review the development guide.