optillm-core Introduction

optillm-core is the foundational library that provides the shared traits, types, and interfaces used by all OptimLLM implementations.

What is optillm-core?

It provides:

• Traits: Abstract interfaces for extending functionality
• Types: Common data structures used across implementations
• Error Handling: Unified error types
• Utilities: Helper functions and macros

Key Components

Traits

• Optimizer: Interface for optimization techniques
• ModelClient: Interface for LLM provider communication

Types

• Solution: Result from optimization containing reasoning and answer
• Prompt: Request structure for LLM calls
• ResponseEvent: Events streamed from LLM responses
• TokenUsage: Token consumption tracking

Error Types

• OptillmError: Comprehensive error enum
• Result<T>: Type alias for Result<T, OptillmError>

Dependencies

• tokio: Async runtime
• async-trait: Async trait support
• serde: Serialization/deserialization
• thiserror: Error handling

When to Use optillm-core

Use optillm-core when:

• Creating a new optimization technique
• Implementing a new LLM provider
• Building on top of OptimLLM functionality
• Integrating OptimLLM into your system

Integration Pattern

Rust

// Add dependency
[dependencies]
optillm-core = { path = "../core" }

// Import types
use optillm_core::{
    Optimizer, ModelClient, Solution, Prompt,
    ResponseEvent, Result,
};

// Implement traits
#[async_trait]
impl Optimizer for MyOptimizer { }

// Use in your code
let result = my_optimizer.optimize(query, client).await?;

Next Steps

• Read ModelClient for LLM integration
• Read Optimizer Trait for creating optimizers
• Check Types for data structure details
• Review Error Handling for error patterns

See the API Reference for complete API documentation.