Source Available

Products

Infrastructure for building autonomous AI systems. All packages available on npm under the @luciformresearch scope.

RagForge Logo

RagForge

@luciformresearch/ragforge
CORE FRAMEWORK

AI agent framework with a persistent local knowledge base for code, documentation, and web content. Enables AI assistants like Claude to maintain searchable context across multiple projects.

Features

Persistent Brain

Neo4j-powered knowledge graph stored locally at ~/.ragforge

Daemon Architecture

Activates on demand with file watching and incremental ingestion

Universal Ingestion

Code, documents (PDF, DOCX), media (glTF, images), web pages

Semantic Search

Vector embeddings via Gemini for intelligent retrieval

MCP Integration

Works with Claude Desktop and MCP-compatible clients

ResearchAgent

Autonomous codebase exploration with structured queries

Packages

@luciformresearch/ragforge

Core library

@luciformresearch/ragforge-cli

CLI & MCP server

@luciformresearch/ragforge-studio

Desktop app (Electron)

CodeParsers Logo

CodeParsers

@luciformresearch/codeparsers
AST EXTRACTION

Unified code parsing library using tree-sitter WASM bindings. Works in both Node.js and browser environments with a consistent API across all supported languages.

Supported Languages

TypeScript/TSXPythonCC++C#GoRustHTMLCSSSCSSVue SFCSvelteMarkdown

Features

  • Tree-sitter Based — Robust, production-ready parsing with WASM bindings
  • Consistent API — Same interface across all parsers
  • Scope Extraction — Functions, classes, imports, decorators, references
  • Browser Compatible — ESM-only modules with bundled WASM grammars
  • Dependency Graph — Cross-file dependency extraction for codebase analysis
</>

XMLParser

@luciformresearch/xmlparser
LLM OUTPUTS

TypeScript XML parser designed for AI pipelines. Excels at parsing LLM-generated XML with permissive mode error recovery while maintaining strict validation for production.

Features

  • Permissive Mode — Robust error recovery for malformed LLM outputs with configurable recovery caps
  • Streaming SAX API — Parse large inputs incrementally as they arrive
  • Namespace Support — xmlns mapping with namespace-aware queries (findByNS, findAllByNS)
  • Security Focused — Depth limits, text-length limits, entity expansion guards
  • Dual Build — ESM/CJS with TypeScript definitions included
L

Lucie Defraiteur

AI Developer & Creator of RagForge

Demo mode — Gemini Flash for cost efficiency. Production uses Gemini Pro.

Bonjour! Je suis Lucie.

Ask me anything about RagForge, CodeParsers, or my work in AI development.