{"full_name":"VikParuchuri/surya","owner":"VikParuchuri","name":"surya","shard":"1","url":"https://github.com/VikParuchuri/surya","description":"OCR, layout analysis, reading order, table recognition in 90+ languages","homepage":"https://www.datalab.to","topics":[],"tag_pages":[],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/176631703?v=4","created_at":"2024-01-10T05:17:42Z","pushed_at":"2026-11-13T15:24:54Z","metrics":{"stars":21592,"forks":1572,"watchers":120,"issues":309,"prs":261,"releases":89,"issues_open":2,"issues_closed":307,"prs_open":0,"prs_closed":65,"prs_merged":196,"discussions":0},"growth":{"day":{"stars":2,"forks":-1,"base_day":"2026-11-15"},"week":{"stars":15,"forks":2,"base_day":"2026-11-04"},"month":{"stars":91,"forks":10,"base_day":"2026-10-17"}},"history":{"dates":["2026-09-06","2026-09-07","2026-09-08","2026-09-09","2026-09-10","2026-09-11","2026-09-12","2026-09-13","2026-09-14","2026-09-15","2026-09-16","2026-09-17","2026-09-18","2026-09-19","2026-09-20","2026-09-21","2026-09-22","2026-09-23","2026-09-24","2026-09-25","2026-09-26","2026-09-27","2026-09-28","2026-09-29","2026-09-30","2026-10-01","2026-10-02","2026-10-03","2026-10-04","2026-10-05","2026-10-06","2026-10-07","2026-10-08","2026-10-09","2026-10-10","2026-10-11","2026-10-12","2026-10-13","2026-10-14","2026-10-15","2026-10-16","2026-10-17","2026-10-18","2026-10-19","2026-10-20","2026-10-21","2026-10-22","2026-10-23","2026-10-24","2026-10-25","2026-10-26","2026-10-27","2026-10-28","2026-10-29","2026-10-30","2026-10-31","2026-11-01","2026-11-02","2026-11-03","2026-11-04","2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[21354,21358,21359,21363,21371,21375,21380,21382,21384,21387,21392,21394,21395,21396,21400,21403,21409,21414,21414,21415,21416,21418,21420,21425,21430,21430,21437,21447,21450,21451,21451,21454,21473,21476,21480,21484,21483,21484,21490,21496,21497,21501,21506,21506,21507,21512,21516,21519,21523,21531,21534,21538,21542,21547,21548,21559,21564,21565,21573,21577,21587,21589,21593,21593,21592,21590,21592],"forks":[1537,1537,1537,1536,1537,1540,1542,1542,1542,1543,1543,1544,1543,1544,1545,1546,1547,1547,1547,1547,1547,1546,1547,1548,1548,1548,1549,1550,1550,1552,1552,1555,1559,1560,1560,1561,1563,1561,1561,1562,1563,1562,1562,1562,1562,1561,1563,1564,1568,1568,1568,1569,1569,1567,1567,1568,1569,1569,1568,1570,1571,1572,1574,1574,1572,1573,1572],"watchers":[127,128,128,128,128,128,128,128,128,128,128,128,128,127,127,127,127,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,126,126,126,126,126,126,126,126,126,126,126,null,126,126,126,121,120,120],"issues_open":[null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,164,164,164,2,2,2],"prs_open":[null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,null,40,40,40,0,0,0]},"boards":{},"same_owner":[],"issues":[{"number":574,"title":"[FEAT] Apple throughput","comments":0,"created_at":"2026-11-13T16:20:25Z"},{"number":573,"title":"Verify evals on Papers with Code","comments":0,"created_at":"2026-11-13T06:53:06Z"}],"discussions":[],"latest_release":{"name":"Surya v0.24.0 — Zero-PyTorch Lightweight Architecture, Interactive Web Viewer (surya_gui) & Agent-Ready CLI","tag":"v0.24.0","published_at":"2026-11-13T15:24:54Z","body":"# Surya v0.24.0 — Zero-PyTorch Lightweight Architecture, Interactive Web Viewer (`surya_gui`) & Agent-Ready CLI\n\nSurya `0.24.0` completes the transition to **Surya 2.5** (`datalab-to/surya-ocr-2.5`) — our 689M-parameter vision-language model for **full-page & per-block OCR**, **layout analysis & reading order**, **inline & block LaTeX math (`<math>`)**, and **HTML table recognition (`<table>` with `colspan` / `rowspan`)** across 90+ languages:\n\n- **85.9 on [OmniParseBench](https://huggingface.co/datasets/datalab-to/omni_parse_bench)** (highest of any open model under 1B parameters) and **83.7% on [olmOCR-bench](https://huggingface.co/datasets/allenai/olmOCR-bench)**.\n- **4.5 pages/s on a single H100** at 192 DPI (`6.2 pages/s` at 96 DPI) — **1.7× faster than Surya 2** and **3.2× faster on Apple Silicon (`llama.cpp` Metal)** because compact block headers (`<|Label|><|x0|><|y0|><|x1|><|y1|>`) and compact table tokens cut output tokens per page by **62%**.\n- **Zero Heavy ML Dependencies (`pip install surya-ocr` installs in seconds)** — removed `torch`, `torchvision`, `transformers`, `opencv-python-headless`, and `streamlit` from runtime dependencies (`>10,000` lines of legacy v1/RF-DETR code removed). Surya now runs as a lightweight client/orchestrator over **`vllm`** (Docker or Podman), **`llama.cpp`** (`llama-server` / `llama-cli`), or **`mlx-vlm`** (Apple Silicon).\n- **New Built-In Interactive Web Viewer (`surya_gui`)** — zero-extra-dependency browser UI to inspect layout boxes, reading order, KaTeX-typeset math, HTML tables, and JSON page by page across all Surya modes.\n\n> Need handwriting, complex forms, or managed VLM parsing at scale with zero infrastructure? Try **[Chandra on the Datalab API](https://www.datalab.to/?utm_source=gh-surya-release)** — get **5,000 free pages every month** when you sign up with a work email.\n\n---\n\n## 1. Built-In Interactive Web Viewer (`surya_gui`)\n\n`surya_gui` now launches a fast, self-contained web viewer directly from the standard library (no `streamlit` installation required):\n\n```bash\nsurya_gui --port 8000\n```\n\n- Drag & drop any **PDF** or **image** (`PNG`, `JPG`, `WEBP`, `GIF`, `TIFF`, `BMP`).\n- Switch live between **Full-Page OCR**, **Block OCR** (96 DPI layout + 192 DPI block crops), **Layout**, and **Table Recognition**.\n- Interactive SVG overlays for **layout bounding boxes**, **canonical block labels**, and **reading order** with bidirectional hover highlighting between the page canvas and the `Blocks` tab.\n- Inspect **Preview** (with live KaTeX math rendering and styled HTML tables), **Blocks**, **HTML**, and **JSON**, with one-click `Copy HTML` and `.json` download.\n\n---\n\n## 2. Clean Top-Level Python API & New Predictor Capabilities\n\n```python\nfrom PIL import Image\nfrom surya import (\n    LayoutPredictor,\n    RecognitionPredictor,\n    SuryaInferenceManager,\n    TableRecPredictor,\n)\n\nmanager = SuryaInferenceManager()  # auto-spawns vllm (GPU) or llama-server (CPU / Apple Silicon)\nimage = Image.open(\"page.png\").convert(\"RGB\")\n\n# 1. Full-page OCR (default: one VLM call per page -> blocks, bboxes, reading order, HTML + <math>)\nrec = RecognitionPredictor(manager)\npages = rec([image])\n\n# Optional: pass skip_labels=[] to also OCR text inside Figure / Picture / Diagram regions\npages_with_figures = rec([image], skip_labels=[])\n\n# 2. Two-stage Layout + Block-mode OCR (auto-runs LayoutPredictor if layouts not passed)\nblock_pages = rec([image], full_page=False)\n\n# 3. HTML Table Recognition (<table> with colspan / rowspan)\ntables = TableRecPredictor(manager)([image])\nprint(tables[0].html)\n```\n\n### What's New in the Predictors & Backends\n- **Top-level package exports**: Import `SuryaInferenceManager`, `get_default_manager`, `LayoutPredictor`, `RecognitionPredictor`, `TableRecPredictor`, and `__version__` directly from `surya`.\n- **Configurable `skip_labels` (`#544`)**: Pass `skip_labels=[]` (or a custom set of labels) to `RecognitionPredictor` to OCR text inside `Figure`, `Picture`, or `Diagram","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}