Open-Source

🎉 Initial Release

The first public release of Documind. Here’s what’s included:

🚀 Features:

• Data Extraction: Extract structured data from PDF documents based on custom schema definitions.
• Schema Support: Easily define schemas to tailor your extraction process for specific document types.
• Accurate Parsing: Handles nested data structures and complex formatting.
• Simple Setup: Start processing documents immediately.

Removed Supabase dependency

Supabase dependency has been removed on the npm package. No need for external storage.

Schema Templates

Schema templates are now available for you to use.

You can get started quickly with some common schemas and contribute new ones.

Read more about them here

Local LLM Models

You can now set a custom base URL and use to use local models

• Local model options are Llava and Llama3.2-vision
• Simply add BASE_URL in environment variables to use

More information here

Autoschema and Ingestion formats

• No need to specify schemas, just set autoSchema to true and Documind will automatically generate a suitable schema and apply it.
• We’ve included functions to simple convert your documents to formats suitable for LLMs. Currently only text and markdown formats are supported.

New file and schema field types

• Support for DOC, TXT, PNG, JPG and HTML file types
• Schemas fields now include boolean and enum

Release of v1.1.0

• Flexible Arrays: If you’re extracting a list of single-type data, results are now returned as a simple array instead of unnecessary object wrapping.
• Markdown in Results: Now you also get the document’s markdown alongside extracted data.
• Full Ollama integration: No more needing an OpenAI key when using the local Ollama model—just provide the base URL, and you’re good to go.
• Google Gemini Models: More options! You can now use Documind with Google’s Gemini models.
• Autoschema Upgrade: Just tell Documind what to extract in plain English, and it will generate the schema for you automatically.