chandra
OCR model that handles complex tables, forms, handwriting with full layout.
Learn how Chandra's paginate_output option formats multi-page documents. Insert visual page separators into Markdown and HTML outputs for clearer OCR results.
What is MAX_VLLM_RETRIES in Chandra? Understanding vLLM Generation Retry LogicDiscover MAX_VLLM_RETRIES in Chandra. Learn how this setting controls vLLM generation retries for repetitive output or service errors and optimize your LLM performance.
How the scale_to_fit Function in Chandra Prepares Images for Model InputUnderstand how Chandra's scale_to_fit function prepares images for model input by normalizing size constraints, aligning to a 28x28 grid, and preserving aspect ratios for vision-language models.
Understanding the Relationship Between Chandra's BatchInputItem and BatchOutputItem SchemasExplore the one-to-one relationship between Chandra's BatchInputItem and BatchOutputItem schemas. Understand how input images transform into detailed model outputs in the inference pipeline.
How to Use the Chandra Streamlit App for Interactive Single-Page Document ProcessingEffortlessly process single-page documents with the Chandra Streamlit app. Run layout-aware OCR locally or remotely, visualizing text, layout, and markdown output instantly.
How detect_repeat_token Identifies Generation Failures in ChandraLearn how Chandra inspects model outputs for repeating token sequences using detect_repeat_token to identify and prevent generation failures.
Chandra load_file Function: Supported File Formats and Processing LogicExplore Chandra's load_file function. Discover supported formats like PDF and Pillow-readable images, and understand how it converts them to RGB Pillow Image objects.
MAX_OUTPUT_TOKENS in Chandra: Purpose, Default Value, and When to Adjust ItUnderstand MAX_OUTPUT_TOKENS in Chandra. Learn its purpose, default value, and when to adjust this config constant to prevent context window overflows and resource exhaustion.
How ThreadPoolExecutor in Chandra's vLLM Module Manages Concurrent Page ProcessingDiscover how Chandra's vLLM module uses ThreadPoolExecutor to boost concurrent page processing Up to 64 worker threads optimize I/O bound HTTP requests while maintaining batch order
How Chandra Header and Footer Filtering Works: A Complete Technical GuideLearn how Chandra's header and footer filtering works. Discover how to omit page headers and footers using the include_headers_footers parameter in this technical guide.
Supported Prompt Types in Chandra: How PROMPT_MAPPING Configures OCR ModesExplore supported prompt types in Chandra: ocr_layout and ocr. Learn how PROMPT_MAPPING in prompts.py configures LLM inference for OCR modes.
How BBOX_SCALE Controls Layout Detection and Chunk Parsing in ChandraDiscover how the BBOX_SCALE setting in Chandra controls layout detection and chunk parsing by defining the OCR model's virtual coordinate grid and impacting pixel coordinate conversion.
Have a question about this repo?
These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:
curl -s "https://instagit.com/install.md" 
Maintain an open-source project? Get it listed too →