Core LLM task framework for Qualve.
LLM task plugin for Qualve, with adapters for Gemini, OpenAI, and Claude.
Pokemon-style agent management for OpenComposer, following open-swe patterns for model configuration and provider systems.
Core orchestration package following Open SWE patterns with LLM task management and state handling. Provides master coordination using OpenRouter API with proper token tracking and caching.
Engine-agnostic Task abstraction for LLM structured-output: Task trait + Grammar (JSON Schema, Lark, Regex) + ImageAnalysis
Rust ONNX inference for LiquidAI LFM2.5-VL (vision-language) models — implements the engine-agnostic llmtask::Task contract via llguidance for schema-constrained sampling
Qwen3-VL vision-language structured-output engine over mistralrs, implementing the engine-agnostic llmtask::Task contract