converge-provider

Multi-provider LLM abstraction layer for the Converge runtime.

Website: converge.zone | Docs: docs.rs

Supported Providers

Features

Model Selection

• Cost-aware selection (CostClass: VeryLow → VeryHigh)
• Latency requirements (tokens/second thresholds)
• Quality requirements (capability levels)
• AgentRequirements builder for declarative selection

Prompt Engineering

• EDN format for token-efficient prompts (~40% reduction)
• XML format for Claude models
• JSON format for OpenAI models
• Structured response parsing

Installation

[dependencies]
converge-provider = "0.2"

Example

use converge_provider::{AnthropicProvider, LlmProvider, LlmRequest};

// Create provider from environment
let provider = AnthropicProvider::from_env("claude-3-5-sonnet-20241022")?;

// Make a request
let request = LlmRequest {
    prompt: "Analyze market trends for Q4".to_string(),
    max_tokens: Some(1000),
    temperature: Some(0.7),
};

let response = provider.complete(&request)?;
println!("Response: {}", response.content);

Model Selection

use converge_provider::{ModelSelector, AgentRequirements, CostClass};

let selector = ModelSelector::default();

let requirements = AgentRequirements::fast_and_cheap()
    .with_max_cost(CostClass::Low);

let (provider, model) = selector.select(&requirements)?;

License

MIT