GitHub hosts HunyuanVideo, Tencent's open-source framework for large-scale video generation models, enabling AI-driven video creation.
Relatch
FreeStarting price
0Popularity
About Relatch
Relatch processes documents such as PDFs, CSV, XML, DOCX, Markdown, and JSON into structured skill files that AI models like Claude and OpenAI Codex can use for reasoning. The tool extracts tables, logic, rules, and domain-specific knowledge from raw files, including scanned PDFs, and organizes the information into typed, queryable formats. It automatically detects the type of knowledge contained in a document—such as persona patterns, codebase logic, process workflows, or domain expertise—and applies the appropriate enrichment template. The output is delivered as structured .md files for Claude or .zip packages for OpenAI Codex CLI, enabling persistent and accurate AI reasoning. Relatch also supports persona extraction, codebase intelligence mapping, process workflow extraction, and financial data schema conversion. The platform emphasizes adaptive intelligence, ensuring the right structure is built without manual selection.
Key features
- Smart OCR for scanned documents
- Structural parsing of tables, lists, and domain language
- Automatic detection of knowledge type (persona, codebase, process, domain)
- Structured output in .md or .zip formats
- Persona and voice extraction from writing samples
- Codebase intelligence mapping for architectural patterns
- Process and workflow extraction with decision trees
- Financial data schema and metric extraction
Use cases
- Converting financial reports into structured datasets for AI analysis
- Transforming technical documentation into codebase intelligence for AI coding assistants
- Building domain-specific AI agents from professional reports or manuals
Pros
- Supports multiple document formats including PDF, CSV, XML, DOCX, Markdown, and JSON
- Automatically extracts and structures tables, logic, rules, and domain knowledge
- Handles scanned PDFs and diagram-heavy files with smart OCR routing
- Adaptive templates that match document content type for accurate skill file generation
- Provides structured outputs for both Claude and OpenAI Codex
Cons
- No explicit mention of a free tier or pricing details on the homepage
- Limited to document-based knowledge extraction; does not support real-time or live data sources
- Output formats are specific to certain AI models, potentially limiting compatibility
Frequently asked questions about Relatch
What types of documents can Relatch process?
Relatch supports a variety of document formats, including PDF, CSV, XML, DOCX, Markdown, TXT, and JSON. It can handle both digital and scanned documents, ensuring no loss of critical information during processing.
How does Relatch determine the structure of the output skill files?
Relatch uses adaptive intelligence to automatically detect the type of knowledge in a document—such as persona patterns, codebase logic, process workflows, or domain expertise—and applies the appropriate enrichment template to build the correct structure.
What formats are the output skill files available in?
The output can be downloaded as structured .md files for use with Claude or as a .zip package for OpenAI Codex CLI. The files are designed to be queryable and persistent for AI reasoning.
Can Relatch process scanned PDFs or image-based documents?
Yes, Relatch includes smart OCR routing to properly read and extract information from scanned PDFs and diagram-heavy files, ensuring nothing is lost due to poor scan quality.
Who is Relatch designed for?
Relatch is designed for professionals and teams who want to leverage their existing documents—such as reports, codebases, SOPs, or financial data—to provide AI models with structured, domain-specific knowledge for accurate reasoning.
Does Relatch require manual configuration to process documents?
No, Relatch automates the entire process. Users simply upload their documents, and the tool handles extraction, enrichment, and structuring without requiring manual selection of templates or settings.
