Releases

Portable, backend-neutral built-in capabilities for Everruns

Embed markdown directly into the binary as egui widgets - extended fork

Shared code for egui_commonmark_extended - fork with typography and header tracking

OpenRouter provider for Everruns agents

A LangChain-inspired framework for building LLM applications in Rust. Supports OpenAI, Agents, Tools, Memory, Chains, RAG, BM25, Hybrid Retrieval, LangGraph, HyDE, Reranking, MultiQuery, and native Function Calling.

OpenAI (GPT) provider for Everruns agents

Matrix implementation for botkit

A2A (Agent-to-Agent) protocol support for langchainrust — client, server, and protocol types.

Guardrails for langchainrust — input/output safety validation for agents

Testing harness for langchainrust — record real LLM exchanges and replay them offline

Language Server Protocol — powered by reveal (spec §16.3)

Browser WebAssembly adapter for the Portable Harn Kernel

Embeddable metasearch library and CLI with browser, HTTP/RSS, and native API sources

Agent system for langchainrust — ReAct, FunctionCalling, PlanExecute, CRAG, AdaptiveRAG, DeepResearch, Handoffs, Streaming

Unified Rust bot framework with optional Discord, Telegram, and Matrix adapters

deepwiki-rs(also known as Litho) is a high-performance automatic generation engine for C4 architecture documentation, developed using Rust. It can intelligently analyze project structures, identify core components, parse dependency relationships, and leverage large language models (LLMs) to automatically generate professional architecture documentation.

Crisp language toolchain — .crp to Rust to native (bins: crisp, reveal)

Evaluation framework for langchainrust — Evaluator trait, built-in evaluators, dataset runner

Session management for langchainrust

Chain compositions for langchainrust — LLMChain, SequentialChain, RetrievalQA, etc.

reveal toolchain — surface inferred precision (spec §16)

RAG (Retrieval-Augmented Generation) module for langchainrust — BM25, Hybrid Retrieval, GraphRAG, HyDE, Reranking, MultiQuery, Document Loaders

Reference command-line prover/verifier over wickra-proof-core: fold a (spec, data) pair into a deterministic backtest report and canonical hash, and verify a claimed proof by recomputation.

Meta Model API provider for Everruns agents

Memory system for langchainrust — Buffer, Window, Summary, SummaryBuffer, ContextWindow, VectorStore, MongoDB persistence

Pretty-print Rust from CIR; invoke cargo/rustc (spec §17.1)

Top-n,k ranked code search: trigram index, best-first verify, proven match sets

Lifetime/region assignment (spec §8)

Typed, ownership-resolved intermediate representation (spec §17.1)

Telegram implementation for botkit