workflow-multi-provider

README

Multi-Provider Workflow Example

This example demonstrates how to implement multi-provider strategies using workflow agents. It shows three common patterns for working with multiple LLM providers:

Patterns Demonstrated

1. Fastest Response Pattern

Uses a parallel workflow with MergeFirst strategy to return the first response from any provider. This is useful when:

• You want the lowest latency possible
• All providers give similar quality responses
• You have redundancy requirements

2. Consensus Pattern

Uses a parallel workflow with a custom merge function to compare responses from multiple providers. This is useful when:

• You need high confidence in the answer
• Working with factual questions
• Want to detect and handle disagreements between models

3. Primary with Fallback Pattern

Uses a sequential workflow with error handling to try providers in order. This is useful when:

• You have a preferred provider but need reliability
• Want to control costs (try cheaper providers first)
• Need graceful degradation

Running the Example

The example works with real providers if API keys are available, or falls back to mock agents for demonstration:

# With real providers (set any combination of these)
export OPENAI_API_KEY=your-key
export ANTHROPIC_API_KEY=your-key  
export GEMINI_API_KEY=your-key

# Run the example
go run main.go

Key Concepts

1. Workflow Agents: The example uses ParallelAgent and SequentialAgent to orchestrate multiple LLM agents.

2. Merge Strategies:

   • MergeFirst: Returns the first result
   • MergeAll: Combines all results
   • Custom merge functions for consensus logic

3. Error Handling: The sequential workflow demonstrates fallback behavior with ErrorActionContinue.

4. State Management: All agents use the state-based execution model with proper state passing between steps.

Comparison with Provider-Level Multi

This example shows multi-provider patterns at the agent level, which provides:

• Better integration with the agent framework
• Access to hooks, events, and state management
• Ability to mix different types of agents (not just LLM providers)
• More flexibility in orchestration patterns

For provider-level multi-provider usage, see the multi and consensus examples.