See release notes.
See release notes.
Improved Sparrow backend with upgraded MLX libs, new Mistral 3.2 model and improved UI.
Enhanced document processing with bounding box annotations and visual coordinate extraction for precise field location tracking
Microservice approach for LLM request processing with instruction calling
Added dashboard to Sparrow UI to visualize system performance
Sparrow is updated with new free tier functionality for https://sparrow.katanaml.io/. New vision backend models are integrated, such as Mistral Small 3.1 and Qwen 2.5 72B
Added Sparrow Agent functionality to orchestrate complex document processing tasks. This new component allows you to combine multiple data extraction operations into a single workflow with visual monitoring and tracking through Prefect integration. Users can now process documents through an intuitive API that handles classification, extraction, and validation in one seamless pipeline.
Implemented image cropping, this is useful for form pages, to reduce overall image size. Add various UI improvements to Sparrow UI Shell, enabled better logging
This release brings Sparrow deployment on local Mac Mini M4 Pro 64GB machine - https://sparrow.katanaml.io
Added support auto detect tables and send cropped table images for inference
Added support for multi-page PDF document through CMD and API
Removed dependencies to LlamaIndex, Haystack, Unstructured and other libraries, as main Sparrow focus is on Sparrow Parse
This release starts new phase in Sparrow development - Vision LLM support for document data processing.
- Sparrow Parse library supports Vision LLM
- Sparrow Parse provides factory class implementation to run inference locally or on cloud GPU
- Sparrow supports JSON as input query
- JSON query validation and LLM response JSON validation is performed
- Sparrow Parse integration
- Sparrow Parse is integrated into Instructor agent. README updated with example for Instructor agent
- New Instructor agent
- Added instructor agent for better JSON response generation
- New agents with Unstructured
- Added unstructured-light and unstructured agents for better data pre-processing
- Virtual Environments support
- Fixes in LlamaIndex agent to run with latest LlamaIndex versions
- LLM function calling agent
- OCR + LLM support, new vprocessor agent
- Improved FastAPI endpoints
- Added Haystack agent for structured data
- Changed plugins to agents
- Added support for plugin architecture. This allows to use within Sparrow various toolkits, such as LlamaIndex or Haystack
- Significant code refactoring
- Minor improvements related to data ingestion
- Fixed bug to clean Vector DB, when new document is inserted
- Tested with Notus and Openhermes LLMs
- Tested with longer and more realistic documents
- Upgraded LlamaIndex and LangChain
- Lemming LLM RAG