_metadata.json` - Metadata (page info, token count, etc.)
+- Extracted images are saved directly in the output directory
+
+### Streamlit Web App
+
+Launch the interactive demo for single-page processing:
+
+```bash
+chandra_app
```
-### vLLM Server
+### vLLM Server (Optional)
-For production or batch processing:
+For production deployments or batch processing, use the vLLM server:
```bash
chandra_vllm
```
-Launches a Docker container with optimized inference. Configure via environment:
+This launches a Docker container with optimized inference settings. Configure via environment variables:
- `VLLM_API_BASE`: Server URL (default: `http://localhost:8000/v1`)
-- `VLLM_MODEL_NAME`: Model name (default: `chandra`)
+- `VLLM_MODEL_NAME`: Model name for the server (default: `chandra`)
- `VLLM_GPUS`: GPU device IDs (default: `0`)
+You can also start your own vllm server with the `datalab-to/chandra-ocr-2` model.
+
### Configuration
-Settings via environment variables or `local.env`:
+Settings can be configured via environment variables or a `local.env` file:
```bash
-MODEL_CHECKPOINT=datalab-to/chandra
-MAX_OUTPUT_TOKENS=8192
+# Model settings
+MODEL_CHECKPOINT=datalab-to/chandra-ocr-2
+MAX_OUTPUT_TOKENS=12384
+
+# vLLM settings
VLLM_API_BASE=http://localhost:8000/v1
+VLLM_MODEL_NAME=chandra
VLLM_GPUS=0
```
-## Commercial Usage
-
-Code is Apache 2.0. Model weights use a modified OpenRAIL-M license: free for research, personal use, and startups under $2M funding/revenue. Cannot be used competitively with our API. For broader commercial licensing, see [pricing](https://www.datalab.to/pricing?utm_source=gh-chandra).
-
-## Credits
+# Commercial usage
+
+This code is Apache 2.0, and our model weights use a modified OpenRAIL-M license (free for research, personal use, and startups under $2M funding/revenue, cannot be used competitively with our API). To remove the OpenRAIL license requirements, or for broader commercial licensing, visit our pricing page [here](https://www.datalab.to/pricing?utm_source=gh-chandra).
+
+# Benchmark table
+
+| **Model** | ArXiv | Old Scans Math | Tables | Old Scans | Headers and Footers | Multi column | Long tiny text | Base | Overall | Source |
+|:--------------------------|:--------:|:--------------:|:--------:|:---------:|:-------------------:|:------------:|:--------------:|:----:|:--------------:|:------:|
+| Datalab API | **90.4** | **90.2** | **90.7** | **54.6** | 91.6 | 83.7 | **92.3** | **99.9** | **86.7 ± 0.8** | Own benchmarks |
+| Chandra 2 | 90.2 | 89.3 | 89.9 | 49.8 | 92.5 | 83.5 | 92.1 | 99.6 | 85.9 ± 0.8 | Own benchmarks |
+| dots.ocr 1.5 | 85.9 | 85.5 | **90.7** | 48.2 | 94.0 | **85.3** | 81.6 | 99.7 | 83.9 | dots.ocr repo |
+| Chandra 1 | 82.2 | 80.3 | 88.0 | 50.4 | 90.8 | 81.2 | **92.3** | **99.9** | 83.1 ± 0.9 | Own benchmarks |
+| olmOCR 2 | 83.0 | 82.3 | 84.9 | 47.7 | **96.1** | 83.7 | 81.9 | 99.6 | 82.4 | olmocr repo |
+| dots.ocr | 82.1 | 64.2 | 88.3 | 40.9 | 94.1 | 82.4 | 81.2 | 99.5 | 79.1 ± 1.0 | dots.ocr repo |
+| olmOCR v0.3.0 | 78.6 | 79.9 | 72.9 | 43.9 | 95.1 | 77.3 | 81.2 | 98.9 | 78.5 ± 1.1 | olmocr repo |
+| Datalab Marker v1.10.0 | 83.8 | 69.7 | 74.8 | 32.3 | 86.6 | 79.4 | 85.7 | 99.6 | 76.5 ± 1.0 | Own benchmarks |
+| Deepseek OCR | 75.2 | 72.3 | 79.7 | 33.3 | **96.1** | 66.7 | 80.1 | 99.7 | 75.4 ± 1.0 | Own benchmarks |
+| Mistral OCR API | 77.2 | 67.5 | 60.6 | 29.3 | 93.6 | 71.3 | 77.1 | 99.4 | 72.0 ± 1.1 | olmocr repo |
+| GPT-4o (Anchored) | 53.5 | 74.5 | 70.0 | 40.7 | 93.8 | 69.3 | 60.6 | 96.8 | 69.9 ± 1.1 | olmocr repo |
+| Qwen 3 VL 8B | 70.2 | 75.1 | 45.6 | 37.5 | 89.1 | 62.1 | 43.0 | 94.3 | 64.6 ± 1.1 | Own benchmarks |
+| Gemini Flash 2 (Anchored) | 54.5 | 56.1 | 72.1 | 34.2 | 64.7 | 61.5 | 71.5 | 95.6 | 63.8 ± 1.2 | olmocr repo |
+
+
+# Multilingual benchmark table
+
+The table below covers the 43 most common languages, benchmarked across multiple models. For a comprehensive evaluation across 90 languages (Chandra 2 vs Gemini 2.5 Flash only), see the [full 90-language benchmark](#full-90-language-benchmark-table).
+
+| Language | Datalab API | Chandra 2 | Chandra 1 | Gemini 2.5 Flash | GPT-5 Mini |
+|---|:---:|:---:|:---:|:---:|:---:|
+| ar | 67.6% | 68.4% | 34.0% | 84.4% | 55.6% |
+| bn | 85.1% | 72.8% | 45.6% | 55.3% | 23.3% |
+| ca | 88.7% | 85.1% | 84.2% | 88.0% | 78.5% |
+| cs | 88.2% | 85.3% | 84.7% | 79.1% | 78.8% |
+| da | 90.1% | 91.1% | 88.4% | 86.0% | 87.7% |
+| de | 93.8% | 94.8% | 83.0% | 88.3% | 93.8% |
+| el | 89.9% | 85.6% | 85.5% | 83.5% | 82.4% |
+| es | 91.8% | 89.3% | 88.7% | 86.8% | 97.1% |
+| fa | 82.2% | 75.1% | 69.6% | 61.8% | 56.4% |
+| fi | 85.7% | 83.4% | 78.4% | 86.0% | 84.7% |
+| fr | 93.3% | 93.7% | 89.6% | 86.1% | 91.1% |
+| gu | 73.8% | 70.8% | 44.6% | 47.6% | 11.5% |
+| he | 76.4% | 70.4% | 38.9% | 50.9% | 22.3% |
+| hi | 80.5% | 78.4% | 70.2% | 82.7% | 41.0% |
+| hr | 93.4% | 90.1% | 85.9% | 88.2% | 81.3% |
+| hu | 88.1% | 82.1% | 82.5% | 84.5% | 84.8% |
+| id | 91.3% | 91.6% | 86.7% | 88.3% | 89.7% |
+| it | 94.4% | 94.1% | 89.1% | 85.7% | 91.6% |
+| ja | 87.3% | 86.9% | 85.4% | 80.0% | 76.1% |
+| jv | 87.5% | 73.2% | 85.1% | 80.4% | 69.6% |
+| kn | 70.0% | 63.2% | 20.6% | 24.5% | 10.1% |
+| ko | 89.1% | 81.5% | 82.3% | 84.8% | 78.4% |
+| la | 78.0% | 73.8% | 55.9% | 70.5% | 54.6% |
+| ml | 72.4% | 64.3% | 18.1% | 23.8% | 11.9% |
+| mr | 80.8% | 75.0% | 57.0% | 69.7% | 20.9% |
+| nl | 90.0% | 88.6% | 85.3% | 87.5% | 83.8% |
+| no | 89.2% | 90.3% | 85.5% | 87.8% | 87.4% |
+| pl | 93.8% | 91.5% | 83.9% | 89.7% | 90.4% |
+| pt | 97.0% | 95.2% | 84.3% | 89.4% | 90.8% |
+| ro | 86.2% | 84.5% | 82.1% | 76.1% | 77.3% |
+| ru | 88.8% | 85.5% | 88.7% | 82.8% | 72.2% |
+| sa | 57.5% | 51.1% | 33.6% | 44.6% | 12.5% |
+| sr | 95.3% | 90.3% | 82.3% | 89.7% | 83.0% |
+| sv | 91.9% | 92.8% | 82.1% | 91.1% | 92.1% |
+| ta | 82.9% | 77.7% | 50.8% | 53.9% | 8.1% |
+| te | 69.4% | 58.6% | 19.5% | 33.3% | 9.9% |
+| th | 71.6% | 62.6% | 47.0% | 66.7% | 53.8% |
+| tr | 88.9% | 84.1% | 68.1% | 84.1% | 78.2% |
+| uk | 93.1% | 91.0% | 88.5% | 87.9% | 81.9% |
+| ur | 54.1% | 43.2% | 28.1% | 57.6% | 16.9% |
+| vi | 85.0% | 80.4% | 81.6% | 89.5% | 83.6% |
+| zh | 87.8% | 88.7% | 88.3% | 70.0% | 70.4% |
+| **Average** | **80.4%** | **77.8%** | **69.4%** | **67.6%** | **60.5%** |
+
+# Full 90-language benchmark table
+
+We also have a more comprehensive evaluation covering 90 languages, comparing Chandra 2 against Gemini 2.5 Flash. The average scores are lower than the 43-language table above because this includes many lower-resource languages. Chandra 2 averages **72.7%** vs Gemini 2.5 Flash at **60.8%**.
+
+See the [full 90-language results](FULL_BENCHMARKS.md).
+
+## Throughput
+
+Benchmarked with vLLM on a single NVIDIA H100 80GB GPU using a diverse mix of documents (math, tables, scans, multi-column layouts) from the olmOCR benchmark set. This set is significantly slower than real-world usage - we estimate 2 pages/s in real-world usage.
+
+| Configuration | Pages/sec | Avg Latency | P95 Latency | Failure Rate |
+|---|:---:|:---:|:---:|:---:|
+| vLLM, 96 concurrent sequences | 1.44 | 60s | 156s | 0% |
+
+# Credits
+
+Thank you to the following open source projects:
- [Huggingface Transformers](https://github.com/huggingface/transformers)
-- [vLLM](https://github.com/vllm-project/vllm)
+- [VLLM](https://github.com/vllm-project/vllm)
- [olmocr](https://github.com/allenai/olmocr)
-- [Qwen3 VL](https://github.com/QwenLM/Qwen3)
-
-## Support Datalab
-If you find this repository helpful, please consider giving it a star ⭐
+- [Qwen 3.5](https://github.com/QwenLM/Qwen3)
\ No newline at end of file
diff --git a/assets/benchmarks/bench.png b/assets/benchmarks/bench.png
index a3dd954..8df0778 100644
Binary files a/assets/benchmarks/bench.png and b/assets/benchmarks/bench.png differ
diff --git a/assets/benchmarks/bench_top.png b/assets/benchmarks/bench_top.png
new file mode 100644
index 0000000..5cf65b4
Binary files /dev/null and b/assets/benchmarks/bench_top.png differ
diff --git a/assets/benchmarks/multilingual.png b/assets/benchmarks/multilingual.png
new file mode 100644
index 0000000..91ed4a0
Binary files /dev/null and b/assets/benchmarks/multilingual.png differ
diff --git a/assets/examples/books/exercises.png b/assets/examples/books/exercises.png
deleted file mode 100644
index 32adc11..0000000
Binary files a/assets/examples/books/exercises.png and /dev/null differ
diff --git a/assets/examples/books/geo_textbook_page.png b/assets/examples/books/geo_textbook_page.png
deleted file mode 100644
index eacc107..0000000
Binary files a/assets/examples/books/geo_textbook_page.png and /dev/null differ
diff --git a/assets/examples/forms/handwritten_form.png b/assets/examples/forms/handwritten_form.png
index 0f5b709..9caf252 100644
Binary files a/assets/examples/forms/handwritten_form.png and b/assets/examples/forms/handwritten_form.png differ
diff --git a/assets/examples/forms/lease.png b/assets/examples/forms/lease.png
deleted file mode 100644
index 1761d55..0000000
Binary files a/assets/examples/forms/lease.png and /dev/null differ
diff --git a/assets/examples/forms/lease_filled.png b/assets/examples/forms/lease_filled.png
new file mode 100644
index 0000000..24b3569
Binary files /dev/null and b/assets/examples/forms/lease_filled.png differ
diff --git a/assets/examples/handwriting/cursive_writing.png b/assets/examples/handwriting/cursive_writing.png
new file mode 100644
index 0000000..e358c96
Binary files /dev/null and b/assets/examples/handwriting/cursive_writing.png differ
diff --git a/assets/examples/handwriting/doctor_note.png b/assets/examples/handwriting/doctor_note.png
deleted file mode 100644
index 5ce581a..0000000
Binary files a/assets/examples/handwriting/doctor_note.png and /dev/null differ
diff --git a/assets/examples/handwriting/handwritten_notes.png b/assets/examples/handwriting/handwritten_notes.png
new file mode 100644
index 0000000..08c3a56
Binary files /dev/null and b/assets/examples/handwriting/handwritten_notes.png differ
diff --git a/assets/examples/handwriting/math_hw.png b/assets/examples/handwriting/math_hw.png
deleted file mode 100644
index e48e953..0000000
Binary files a/assets/examples/handwriting/math_hw.png and /dev/null differ
diff --git a/assets/examples/languages/arabic.png b/assets/examples/languages/arabic.png
new file mode 100644
index 0000000..bde3c65
Binary files /dev/null and b/assets/examples/languages/arabic.png differ
diff --git a/assets/examples/languages/hindi.png b/assets/examples/languages/hindi.png
new file mode 100644
index 0000000..2e2ae94
Binary files /dev/null and b/assets/examples/languages/hindi.png differ
diff --git a/assets/examples/languages/japanese.png b/assets/examples/languages/japanese.png
new file mode 100644
index 0000000..0461423
Binary files /dev/null and b/assets/examples/languages/japanese.png differ
diff --git a/assets/examples/languages/russian.png b/assets/examples/languages/russian.png
new file mode 100644
index 0000000..ad817d2
Binary files /dev/null and b/assets/examples/languages/russian.png differ
diff --git a/assets/examples/math/attn_all.png b/assets/examples/math/attn_all.png
deleted file mode 100644
index 91dee65..0000000
Binary files a/assets/examples/math/attn_all.png and /dev/null differ
diff --git a/assets/examples/math/chinese_math.png b/assets/examples/math/chinese_math.png
new file mode 100644
index 0000000..e6cb2bb
Binary files /dev/null and b/assets/examples/math/chinese_math.png differ
diff --git a/assets/examples/math/cs229.png b/assets/examples/math/cs229.png
new file mode 100644
index 0000000..d505532
Binary files /dev/null and b/assets/examples/math/cs229.png differ
diff --git a/assets/examples/math/ega.png b/assets/examples/math/ega.png
deleted file mode 100644
index de02a3a..0000000
Binary files a/assets/examples/math/ega.png and /dev/null differ
diff --git a/assets/examples/math/handwritten_math.png b/assets/examples/math/handwritten_math.png
new file mode 100644
index 0000000..baafa06
Binary files /dev/null and b/assets/examples/math/handwritten_math.png differ
diff --git a/assets/examples/math/worksheet.png b/assets/examples/math/worksheet.png
deleted file mode 100644
index ccc1f5f..0000000
Binary files a/assets/examples/math/worksheet.png and /dev/null differ
diff --git a/assets/examples/newspapers/la_times.png b/assets/examples/newspapers/la_times.png
deleted file mode 100644
index 0c91d2d..0000000
Binary files a/assets/examples/newspapers/la_times.png and /dev/null differ
diff --git a/assets/examples/newspapers/nyt.png b/assets/examples/newspapers/nyt.png
deleted file mode 100644
index e83149d..0000000
Binary files a/assets/examples/newspapers/nyt.png and /dev/null differ
diff --git a/assets/examples/other/charts.png b/assets/examples/other/charts.png
new file mode 100644
index 0000000..bb2a0d3
Binary files /dev/null and b/assets/examples/other/charts.png differ
diff --git a/assets/examples/other/chemistry.png b/assets/examples/other/chemistry.png
new file mode 100644
index 0000000..36d3041
Binary files /dev/null and b/assets/examples/other/chemistry.png differ
diff --git a/assets/examples/other/flowchart.png b/assets/examples/other/flowchart.png
deleted file mode 100644
index 4417937..0000000
Binary files a/assets/examples/other/flowchart.png and /dev/null differ
diff --git a/assets/examples/other/transcript.png b/assets/examples/other/transcript.png
deleted file mode 100644
index e1a6ee8..0000000
Binary files a/assets/examples/other/transcript.png and /dev/null differ
diff --git a/assets/examples/tables/10k.png b/assets/examples/tables/10k.png
deleted file mode 100644
index 69a7bea..0000000
Binary files a/assets/examples/tables/10k.png and /dev/null differ
diff --git a/assets/examples/tables/complex_tables.png b/assets/examples/tables/complex_tables.png
new file mode 100644
index 0000000..48cce15
Binary files /dev/null and b/assets/examples/tables/complex_tables.png differ
diff --git a/assets/examples/tables/financial_table.png b/assets/examples/tables/financial_table.png
new file mode 100644
index 0000000..4dbed07
Binary files /dev/null and b/assets/examples/tables/financial_table.png differ
diff --git a/assets/examples/tables/water_damage.png b/assets/examples/tables/water_damage.png
deleted file mode 100644
index a2cb26f..0000000
Binary files a/assets/examples/tables/water_damage.png and /dev/null differ
diff --git a/chandra/model/hf.py b/chandra/model/hf.py
index 6be4e47..3f4ebf0 100644
--- a/chandra/model/hf.py
+++ b/chandra/model/hf.py
@@ -13,29 +13,21 @@ def generate_hf(
bbox_scale: int = settings.BBOX_SCALE,
**kwargs,
) -> List[GenerationResult]:
- from qwen_vl_utils import process_vision_info
-
if max_output_tokens is None:
max_output_tokens = settings.MAX_OUTPUT_TOKENS
- messages = [
- process_batch_element(item, model.processor, bbox_scale) for item in batch
- ]
- text = model.processor.apply_chat_template(
- messages, tokenize=False, add_generation_prompt=True
- )
+ conversations = [[process_batch_element(item)] for item in batch]
- image_inputs, _ = process_vision_info(messages)
- inputs = model.processor(
- text=text,
- images=image_inputs,
- padding=True,
+ inputs = model.processor.apply_chat_template(
+ conversations,
+ tokenize=True,
+ add_generation_prompt=True,
+ return_dict=True,
return_tensors="pt",
- padding_side="left",
+ padding=True,
)
- inputs = inputs.to("cuda")
+ inputs = inputs.to(model.device)
- # Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=max_output_tokens)
generated_ids_trimmed = [
out_ids[len(in_ids) :]
@@ -53,25 +45,29 @@ def generate_hf(
return results
-def process_batch_element(item: BatchInputItem, processor, bbox_scale: int):
+def process_batch_element(item: BatchInputItem):
prompt = item.prompt
prompt_type = item.prompt_type
if not prompt:
- prompt = PROMPT_MAPPING[prompt_type].replace("{bbox_scale}", str(bbox_scale))
+ prompt = PROMPT_MAPPING[prompt_type]
content = []
image = scale_to_fit(item.image) # Guarantee max size
content.append({"type": "image", "image": image})
-
content.append({"type": "text", "text": prompt})
- message = {"role": "user", "content": content}
- return message
+ return {"role": "user", "content": content}
def load_model():
- import torch
- from transformers import Qwen3VLForConditionalGeneration, Qwen3VLProcessor
+ try:
+ import torch
+ from transformers import AutoModelForImageTextToText, AutoProcessor
+ except ImportError:
+ raise ImportError(
+ "HuggingFace backend requires additional dependencies. "
+ "Install with: pip install chandra-ocr[hf]"
+ )
device_map = "auto"
if settings.TORCH_DEVICE:
@@ -84,10 +80,11 @@ def load_model():
if settings.TORCH_ATTN:
kwargs["attn_implementation"] = settings.TORCH_ATTN
- model = Qwen3VLForConditionalGeneration.from_pretrained(
+ model = AutoModelForImageTextToText.from_pretrained(
settings.MODEL_CHECKPOINT, **kwargs
)
model = model.eval()
- processor = Qwen3VLProcessor.from_pretrained(settings.MODEL_CHECKPOINT)
+ processor = AutoProcessor.from_pretrained(settings.MODEL_CHECKPOINT)
+ processor.tokenizer.padding_side = "left"
model.processor = processor
return model
diff --git a/chandra/model/util.py b/chandra/model/util.py
index 7cef96f..f7e97a3 100644
--- a/chandra/model/util.py
+++ b/chandra/model/util.py
@@ -1,42 +1,65 @@
-import math
from typing import Tuple
from PIL import Image
-from chandra.output import parse_markdown
-
def scale_to_fit(
img: Image.Image,
max_size: Tuple[int, int] = (3072, 2048),
- min_size: Tuple[int, int] = (28, 28),
+ min_size: Tuple[int, int] = (1792, 28),
+ grid_size: int = 28,
):
resample_method = Image.Resampling.LANCZOS
width, height = img.size
# Check for empty or invalid image
- if width == 0 or height == 0:
+ if width <= 0 or height <= 0:
return img
- max_width, max_height = max_size
- min_width, min_height = min_size
-
+ original_ar = width / height
current_pixels = width * height
- max_pixels = max_width * max_height
- min_pixels = min_width * min_height
+ max_pixels = max_size[0] * max_size[1]
+ min_pixels = min_size[0] * min_size[1]
+ # 1. Determine ideal float scale based on pixel bounds
+ scale = 1.0
if current_pixels > max_pixels:
- scale_factor = (max_pixels / current_pixels) ** 0.5
-
- new_width = math.floor(width * scale_factor)
- new_height = math.floor(height * scale_factor)
+ scale = (max_pixels / current_pixels) ** 0.5
elif current_pixels < min_pixels:
- scale_factor = (min_pixels / current_pixels) ** 0.5
+ scale = (min_pixels / current_pixels) ** 0.5
+
+ # 2. Convert dimensions to integer "grid blocks"
+ w_blocks = max(1, round((width * scale) / grid_size))
+ h_blocks = max(1, round((height * scale) / grid_size))
+
+ # 3. Refinement Loop: Ensure we are under the max limit
+ while (w_blocks * h_blocks * grid_size * grid_size) > max_pixels:
+ if w_blocks == 1 and h_blocks == 1:
+ break
+
+ if w_blocks == 1:
+ h_blocks -= 1
+ continue
+ if h_blocks == 1:
+ w_blocks -= 1
+ continue
+
+ # Compare distortion: Which move preserves Aspect Ratio better?
+ ar_w_loss = abs(((w_blocks - 1) / h_blocks) - original_ar)
+ ar_h_loss = abs((w_blocks / (h_blocks - 1)) - original_ar)
- new_width = math.ceil(width * scale_factor)
- new_height = math.ceil(height * scale_factor)
- else:
+ if ar_w_loss < ar_h_loss:
+ w_blocks -= 1
+ else:
+ h_blocks -= 1
+
+ # 4. Calculate final pixel dimensions
+ new_width = w_blocks * grid_size
+ new_height = h_blocks * grid_size
+
+ # Return original if no changes were needed
+ if (new_width, new_height) == (width, height):
return img
return img.resize((new_width, new_height), resample=resample_method)
@@ -49,12 +72,6 @@ def detect_repeat_token(
cut_from_end: int = 0,
scaling_factor: float = 3.0,
):
- try:
- predicted_tokens = parse_markdown(predicted_tokens)
- except Exception as e:
- print(f"Error parsing markdown: {e}")
- return True
-
if cut_from_end > 0:
predicted_tokens = predicted_tokens[:-cut_from_end]
diff --git a/chandra/model/vllm.py b/chandra/model/vllm.py
index fce2ed6..94ad975 100644
--- a/chandra/model/vllm.py
+++ b/chandra/model/vllm.py
@@ -30,6 +30,8 @@ def generate_vllm(
max_failure_retries: int | None = None,
bbox_scale: int = settings.BBOX_SCALE,
vllm_api_base: str = settings.VLLM_API_BASE,
+ temperature: float = 0.0,
+ top_p: float = 0.1,
) -> List[GenerationResult]:
client = OpenAI(
api_key=settings.VLLM_API_KEY,
@@ -51,14 +53,10 @@ def generate_vllm(
models = client.models.list()
model_name = models.data[0].id
- def _generate(
- item: BatchInputItem, temperature: float = 0, top_p: float = 0.1
- ) -> GenerationResult:
+ def _generate(item: BatchInputItem, temperature, top_p) -> GenerationResult:
prompt = item.prompt
if not prompt:
- prompt = PROMPT_MAPPING[item.prompt_type].replace(
- "{bbox_scale}", str(bbox_scale)
- )
+ prompt = PROMPT_MAPPING[item.prompt_type]
content = []
image = scale_to_fit(item.image)
@@ -93,11 +91,12 @@ def _generate(
return result
def process_item(item, max_retries, max_failure_retries=None):
- result = _generate(item)
+ result = _generate(item, temperature=temperature, top_p=top_p)
retries = 0
while _should_retry(result, retries, max_retries, max_failure_retries):
- result = _generate(item, temperature=0.3, top_p=0.95)
+ retry_temperature = min(temperature + 0.2 * (retries + 1), 0.8)
+ result = _generate(item, temperature=retry_temperature, top_p=0.95)
retries += 1
return result
diff --git a/chandra/output.py b/chandra/output.py
index a8a8b18..030a4fd 100644
--- a/chandra/output.py
+++ b/chandra/output.py
@@ -1,5 +1,4 @@
import hashlib
-import json
import re
from dataclasses import dataclass, asdict
from functools import lru_cache
@@ -28,7 +27,7 @@ def extract_images(html: str, chunks: dict, image: Image.Image):
for idx, chunk in enumerate(chunks):
div_idx += 1
if chunk["label"] in ["Image", "Figure"]:
- img = chunk["content"].find("img")
+ img = BeautifulSoup(chunk["content"], "html.parser").find("img")
if not img:
continue
bbox = chunk["bbox"]
@@ -54,6 +53,9 @@ def parse_html(
div_idx += 1
label = div.get("data-label")
+ if label == "Blank-Page":
+ continue
+
# Skip headers and footers if not included
if label and not include_headers_footers:
if label in ["Page-Header", "Page-Footer"]:
@@ -74,6 +76,12 @@ def parse_html(
img = BeautifulSoup(f"
", "html.parser")
div.append(img)
+ # Strip img tags without src in non-image blocks (model hallucinations)
+ if label not in ["Image", "Figure"]:
+ for img_tag in div.find_all("img"):
+ if not img_tag.get("src"):
+ img_tag.decompose()
+
# Wrap text content in tags if no inner HTML tags exist
if label in ["Text"] and not re.search(
"<.+>", str(div.decode_contents()).strip()
@@ -196,19 +204,20 @@ def parse_layout(html: str, image: Image.Image, bbox_scale=settings.BBOX_SCALE):
height_scaler = height / bbox_scale
layout_blocks = []
for div in top_level_divs:
+ label = div.get("data-label")
+ if label == "Blank-Page":
+ continue
+
bbox = div.get("data-bbox")
try:
- bbox = json.loads(bbox)
+ bbox = bbox.split(" ")
+ bbox = list(map(int, bbox))
assert len(bbox) == 4, "Invalid bbox length"
except Exception:
- try:
- bbox = bbox.split(" ")
- assert len(bbox) == 4, "Invalid bbox length"
- except Exception:
- bbox = [0, 0, 1, 1]
+ print(f"Invalid bbox format: {bbox}, defaulting to full image")
+ bbox = [0, 0, 1, 1]
- bbox = list(map(int, bbox))
# Normalize bbox
bbox = [
max(0, int(bbox[0] * width_scaler)),
@@ -216,8 +225,16 @@ def parse_layout(html: str, image: Image.Image, bbox_scale=settings.BBOX_SCALE):
min(int(bbox[2] * width_scaler), width),
min(int(bbox[3] * height_scaler), height),
]
- label = div.get("data-label", "block")
+ if not label:
+ label = "block"
content = str(div.decode_contents())
+
+ # Strip nested data-bbox attributes (not needed in open source)
+ content_soup = BeautifulSoup(content, "html.parser")
+ for tag in content_soup.find_all(attrs={"data-bbox": True}):
+ del tag["data-bbox"]
+ content = str(content_soup)
+
layout_blocks.append(LayoutBlock(bbox=bbox, label=label, content=content))
return layout_blocks
diff --git a/chandra/prompts.py b/chandra/prompts.py
index 954897e..5b19946 100644
--- a/chandra/prompts.py
+++ b/chandra/prompts.py
@@ -34,6 +34,7 @@
"thead",
"big",
"code",
+ "chem",
]
ALLOWED_ATTRIBUTES = [
"class",
@@ -48,6 +49,8 @@
"href",
"alt",
"align",
+ "data-bbox",
+ "data-label",
]
PROMPT_ENDING = f"""
@@ -57,15 +60,17 @@
* Inline math: Surround math with tags. Math expressions should be rendered in KaTeX-compatible LaTeX. Use display for block math.
* Tables: Use colspan and rowspan attributes to match table structure.
* Formatting: Maintain consistent formatting with the image, including spacing, indentation, subscripts/superscripts, and special characters.
-* Images: Include a description of any images in the alt attribute of an ![]()
tag. Do not fill out the src property.
+* Images: Include a description of any images in the alt attribute of an ![]()
tag. Do not fill out the src property. Describe in detail inside the div tag. Also convert charts to high fidelity data, and convert diagrams to mermaid.
* Forms: Mark checkboxes and radio buttons properly.
* Text: join lines together properly into paragraphs using
...
tags. Use
tags for line breaks within paragraphs, but only when absolutely necessary to maintain meaning.
+* Chemistry: Use ... tags for chemical formulas with reactive SMILES.
+* Lists: Preserve indents and proper list markers.
* Use the simplest possible HTML structure that accurately represents the content of the block.
* Make sure the text is accurate and easy for a human to read and interpret. Reading order should be correct and natural.
""".strip()
OCR_LAYOUT_PROMPT = f"""
-OCR this image to HTML, arranged as layout blocks. Each layout block should be a div with the data-bbox attribute representing the bounding box of the block in [x0, y0, x1, y1] format. Bboxes are normalized 0-{{bbox_scale}}. The data-label attribute is the label for the block.
+OCR this image to HTML, arranged as layout blocks. Each layout block should be a div with the data-bbox attribute representing the bounding box of the block in x0 y0 x1 y1 format. Bboxes are normalized 0-1000. The data-label attribute is the label for the block.
Use the following labels:
- Caption
@@ -83,6 +88,10 @@
- Form
- Table-Of-Contents
- Figure
+- Chemical-Block
+- Diagram
+- Bibliography
+- Blank-Page
{PROMPT_ENDING}
""".strip()
@@ -97,3 +106,9 @@
"ocr_layout": OCR_LAYOUT_PROMPT,
"ocr": OCR_PROMPT,
}
+
+if __name__ == "__main__":
+ print("OCR Layout Prompt:")
+ print(OCR_LAYOUT_PROMPT)
+ print("\nOCR Prompt:")
+ print(OCR_PROMPT)
diff --git a/chandra/scripts/vllm.py b/chandra/scripts/vllm.py
index 26ea073..5c602c1 100644
--- a/chandra/scripts/vllm.py
+++ b/chandra/scripts/vllm.py
@@ -1,11 +1,66 @@
+import argparse
+import json
+import math
import os
import subprocess
import sys
from chandra.settings import settings
+# H100 80GB is the baseline for scaling
+BASELINE_VRAM_GB = 80
+BASELINE_MAX_BATCHED_TOKENS = 8192
+BASELINE_MAX_NUM_SEQS = 64
+
+GPU_VRAM_GB = {
+ "h100": 80,
+ "a100-80": 80,
+ "a100": 40,
+ "a100-40": 40,
+ "l40s": 48,
+ "a10": 24,
+ "l4": 24,
+ "4090": 24,
+ "3090": 24,
+ "t4": 16,
+}
+
+
+def get_gpu_settings(gpu: str):
+ vram = GPU_VRAM_GB.get(gpu)
+ if vram is None:
+ available = ", ".join(sorted(GPU_VRAM_GB.keys()))
+ print(f"Unknown GPU '{gpu}'. Available: {available}")
+ sys.exit(1)
+
+ ratio = vram / BASELINE_VRAM_GB
+ # Scale and round down to nearest power of 2 for batched tokens
+ raw_tokens = BASELINE_MAX_BATCHED_TOKENS * ratio
+ max_batched_tokens = max(1024, 2 ** math.floor(math.log2(raw_tokens)))
+ # Scale and round down to nearest multiple of 8 for seqs
+ max_num_seqs = max(8, (int(BASELINE_MAX_NUM_SEQS * ratio) // 8) * 8)
+
+ return max_batched_tokens, max_num_seqs
+
def main():
+ parser = argparse.ArgumentParser(description="Launch vLLM server for Chandra")
+ parser.add_argument(
+ "--gpu",
+ default="h100",
+ choices=sorted(GPU_VRAM_GB.keys()),
+ help="GPU type for optimal settings (default: h100)",
+ )
+ parser.add_argument(
+ "--mtp",
+ action="store_true",
+ default=False,
+ help="Enable MTP speculative decoding (disabled by default, unstable with vLLM)",
+ )
+ args = parser.parse_args()
+
+ max_batched_tokens, max_num_seqs = get_gpu_settings(args.gpu)
+
cmd = [
"sudo",
"docker",
@@ -16,33 +71,41 @@ def main():
f"device={settings.VLLM_GPUS}",
"-v",
f"{os.path.expanduser('~')}/.cache/huggingface:/root/.cache/huggingface",
- "--env",
- "VLLM_ATTENTION_BACKEND=TORCH_SDPA",
"-p",
"8000:8000",
"--ipc=host",
- "vllm/vllm-openai:latest",
+ "vllm/vllm-openai:v0.17.0",
"--model",
settings.MODEL_CHECKPOINT,
"--no-enforce-eager",
"--max-num-seqs",
- "32",
+ str(max_num_seqs),
"--dtype",
"bfloat16",
"--max-model-len",
- "32768",
+ "18000",
"--max_num_batched_tokens",
- "65536",
+ str(max_batched_tokens),
"--gpu-memory-utilization",
- ".9",
+ ".85",
+ "--enable-prefix-caching",
+ "--mm-processor-kwargs",
+ json.dumps({"min_pixels": 3136, "max_pixels": 6291456}),
"--served-model-name",
settings.VLLM_MODEL_NAME,
]
- print(f"Starting vLLM server with command: {' '.join(cmd)}")
+ if args.mtp:
+ spec_config = json.dumps({"method": "mtp", "num_speculative_tokens": 1})
+ cmd.extend(["--speculative-config", spec_config])
+
+ vram = GPU_VRAM_GB[args.gpu]
+ print(f"GPU: {args.gpu} ({vram}GB VRAM)")
+ print(f"max-num-batched-tokens: {max_batched_tokens}, max-num-seqs: {max_num_seqs}")
+ print(f"MTP: {'enabled' if args.mtp else 'disabled'}")
+ print(f"Command: {' '.join(cmd)}")
try:
- # Use subprocess.run() which blocks and streams output automatically
subprocess.run(cmd, check=True)
except KeyboardInterrupt:
print("\nShutting down vLLM server...")
diff --git a/chandra/settings.py b/chandra/settings.py
index ceb06ba..6d0f8a0 100644
--- a/chandra/settings.py
+++ b/chandra/settings.py
@@ -9,11 +9,11 @@ class Settings(BaseSettings):
IMAGE_DPI: int = 192
MIN_PDF_IMAGE_DIM: int = 1024
MIN_IMAGE_DIM: int = 1536
- MODEL_CHECKPOINT: str = "datalab-to/chandra"
+ MODEL_CHECKPOINT: str = "datalab-to/chandra-ocr-2"
TORCH_DEVICE: str | None = None
MAX_OUTPUT_TOKENS: int = 12384
TORCH_ATTN: str | None = None
- BBOX_SCALE: int = 1024
+ BBOX_SCALE: int = 1000
# vLLM server settings
VLLM_API_KEY: str = "EMPTY"
diff --git a/pyproject.toml b/pyproject.toml
index 6be56bf..9ad2c0f 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
[project]
name = "chandra-ocr"
-version = "0.1.9"
+version = "0.2.0"
description = "OCR model that converts documents to markdown, HTML, or JSON."
readme = "README.md"
requires-python = ">=3.10"
@@ -13,7 +13,6 @@ dependencies = [
"beautifulsoup4>=4.14.2",
"click>=8.0.0",
"filetype>=1.2.0",
- "flask>=3.0.0",
"markdownify==1.1.0",
"openai>=2.2.0",
"pillow>=10.2.0",
@@ -21,13 +20,22 @@ dependencies = [
"pydantic-settings>=2.11.0",
"pypdfium2>=4.30.0",
"python-dotenv>=1.1.1",
- "qwen-vl-utils>=0.0.14",
+ "six>=1.17.0",
+]
+
+[project.optional-dependencies]
+hf = [
"torch>=2.8.0",
"torchvision>=0.23.0",
- "transformers>=4.57.1",
- "streamlit>=1.50.0",
+ "transformers>=5.2.0",
"accelerate>=1.11.0",
]
+app = [
+ "streamlit>=1.50.0",
+]
+all = [
+ "chandra-ocr[hf,app]",
+]
[build-system]
requires = ["setuptools>=61"] # or "setuptools>=61", "flit-core", etc.
@@ -44,6 +52,8 @@ include = ["chandra*"]
[dependency-groups]
dev = [
+ "chandra-ocr[all]",
+ "flask>=3.0.0",
"pre-commit>=4.3.0",
"pytest>=8.4.2",
]
diff --git a/uv.lock b/uv.lock
index e68162f..3fbe1fd 100644
--- a/uv.lock
+++ b/uv.lock
@@ -47,6 +47,15 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/aa/f3/0b6ced594e51cc95d8c1fc1640d3623770d01e4969d29c0bd09945fafefa/altair-5.5.0-py3-none-any.whl", hash = "sha256:91a310b926508d560fe0148d02a194f38b824122641ef528113d029fcd129f8c", size = 731200 },
]
+[[package]]
+name = "annotated-doc"
+version = "0.0.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/57/ba/046ceea27344560984e26a590f90bc7f4a75b06701f653222458922b558c/annotated_doc-0.0.4.tar.gz", hash = "sha256:fbcda96e87e9c92ad167c2e53839e57503ecfda18804ea28102353485033faa4", size = 7288 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/1e/d3/26bf1008eb3d2daa8ef4cacc7f3bfdc11818d111f7e2d0201bc6e3b49d45/annotated_doc-0.0.4-py3-none-any.whl", hash = "sha256:571ac1dc6991c450b25a9c2d84a3705e2ae7a53467b5d111c24fa8baabbed320", size = 5303 },
+]
+
[[package]]
name = "annotated-types"
version = "0.7.0"
@@ -80,56 +89,6 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/3a/2a/7cc015f5b9f5db42b7d48157e23356022889fc354a2813c15934b7cb5c0e/attrs-25.4.0-py3-none-any.whl", hash = "sha256:adcf7e2a1fb3b36ac48d97835bb6d8ade15b8dcce26aba8bf1d14847b57a3373", size = 67615 },
]
-[[package]]
-name = "av"
-version = "15.1.0"
-source = { registry = "https://pypi.org/simple" }
-sdist = { url = "https://files.pythonhosted.org/packages/e9/c3/83e6e73d1592bc54436eae0bc61704ae0cff0c3cfbde7b58af9ed67ebb49/av-15.1.0.tar.gz", hash = "sha256:39cda2dc810e11c1938f8cb5759c41d6b630550236b3365790e67a313660ec85", size = 3774192 }
-wheels = [
- { url = "https://files.pythonhosted.org/packages/3a/6a/91e3e68ae0d1b53b480ec69a96f2ae820fb007bc60e6b821741f31c7ba4e/av-15.1.0-cp310-cp310-macosx_13_0_arm64.whl", hash = "sha256:cf067b66cee2248220b29df33b60eb4840d9e7b9b75545d6b922f9c41d88c4ee", size = 21781685 },
- { url = "https://files.pythonhosted.org/packages/bc/6d/afa951b9cb615c3bc6d95c4eed280c6cefb52c006f4e15e79043626fab39/av-15.1.0-cp310-cp310-macosx_13_0_x86_64.whl", hash = "sha256:26426163d96fc3bde9a015ba4d60da09ef848d9284fe79b4ca5e60965a008fc5", size = 26962481 },
- { url = "https://files.pythonhosted.org/packages/3c/42/0c384884235c42c439cef28cbd129e4624ad60229119bf3c6c6020805119/av-15.1.0-cp310-cp310-manylinux_2_28_aarch64.whl", hash = "sha256:92f524541ce74b8a12491d8934164a5c57e983da24826547c212f60123de400b", size = 37571839 },
- { url = "https://files.pythonhosted.org/packages/25/c0/5c967b0872fce1add80a8f50fa7ce11e3e3e5257c2b079263570bc854699/av-15.1.0-cp310-cp310-manylinux_2_28_x86_64.whl", hash = "sha256:659f9d6145fb2c58e8b31907283b6ba876570f5dd6e7e890d74c09614c436c8e", size = 39070227 },
- { url = "https://files.pythonhosted.org/packages/e2/81/e333056d49363c35a74b828ed5f87c96dfbcc1a506b49d79a31ac773b94d/av-15.1.0-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:07a8ae30c0cfc3132eff320a6b27d18a5e0dda36effd0ae28892888f4ee14729", size = 39619362 },
- { url = "https://files.pythonhosted.org/packages/d5/ae/50cc2af1bf68452cbfec8d1b2554c18f6d167c8ba6d7ad7707797dfd1541/av-15.1.0-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:e33a76e38f03bb5de026b9f66ccf23dc01ddd2223221096992cb52ac22e62538", size = 40371627 },
- { url = "https://files.pythonhosted.org/packages/50/e6/381edf1779106dd31c9ef1ac9842f643af4465b8a87cbc278d3eaa76229a/av-15.1.0-cp310-cp310-win_amd64.whl", hash = "sha256:aa4bf12bdce20edc2a3b13a2776c474c5ab63e1817d53793714504476eeba82e", size = 31340369 },
- { url = "https://files.pythonhosted.org/packages/47/58/4e44cf6939be7aba96a4abce024e1be11ba7539ecac74d09369b8c03aa05/av-15.1.0-cp311-cp311-macosx_13_0_arm64.whl", hash = "sha256:b785948762a8d45fc58fc24a20251496829ace1817e9a7a508a348d6de2182c3", size = 21767323 },
- { url = "https://files.pythonhosted.org/packages/9b/f6/a946544cdb49f6d892d2761b1d61a8bc6ce912fe57ba06769bdc640c0a7f/av-15.1.0-cp311-cp311-macosx_13_0_x86_64.whl", hash = "sha256:9c7131494a3a318612b4ee4db98fe5bc50eb705f6b6536127c7ab776c524fd8b", size = 26946268 },
- { url = "https://files.pythonhosted.org/packages/70/7c/b33513c0af73d0033af59a98f035b521c5b93445a6af7e9efbf41a6e8383/av-15.1.0-cp311-cp311-manylinux_2_28_aarch64.whl", hash = "sha256:2b9623ae848625c59213b610c8665817924f913580c7c5c91e0dc18936deb00d", size = 38062118 },
- { url = "https://files.pythonhosted.org/packages/5e/95/31b7fb34f9fea7c7389240364194f4f56ad2d460095038cc720f50a90bb3/av-15.1.0-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:c8ef597087db560514617143532b1fafc4825ebb2dda9a22418f548b113a0cc7", size = 39571086 },
- { url = "https://files.pythonhosted.org/packages/e7/b0/7b0b45474a4e90c35c11d0032947d8b3c7386872957ce29c6f12add69a74/av-15.1.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:08eac47a90ebae1e2bd5935f400dd515166019bab4ff5b03c4625fa6ac3a0a5e", size = 40112634 },
- { url = "https://files.pythonhosted.org/packages/aa/04/038b94bc9a1ee10a451c867d4a2fc91e845f83bfc2dae9df25893abcb57f/av-15.1.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:d3f66ff200ea166e606cb3c5cb1bd2fc714effbec2e262a5d67ce60450c8234a", size = 40878695 },
- { url = "https://files.pythonhosted.org/packages/1d/3d/9f8f96c0deeaaf648485a3dbd1699b2f0580f2ce8a36cb616c0138ba7615/av-15.1.0-cp311-cp311-win_amd64.whl", hash = "sha256:57b99544d91121b8bea570e4ddf61700f679a6b677c1f37966bc1a22e1d4cd5c", size = 31335683 },
- { url = "https://files.pythonhosted.org/packages/d1/58/de78b276d20db6ffcd4371283df771721a833ba525a3d57e753d00a9fe79/av-15.1.0-cp312-cp312-macosx_13_0_arm64.whl", hash = "sha256:40c5df37f4c354ab8190c6fd68dab7881d112f527906f64ca73da4c252a58cee", size = 21760991 },
- { url = "https://files.pythonhosted.org/packages/56/cc/45f85775304ae60b66976360d82ba5b152ad3fd91f9267d5020a51e9a828/av-15.1.0-cp312-cp312-macosx_13_0_x86_64.whl", hash = "sha256:af455ce65ada3d361f80c90c810d9bced4db5655ab9aa513024d6c71c5c476d5", size = 26953097 },
- { url = "https://files.pythonhosted.org/packages/f3/f8/2d781e5e71d02fc829487e775ccb1185e72f95340d05f2e84eb57a11e093/av-15.1.0-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:86226d2474c80c3393fa07a9c366106029ae500716098b72b3ec3f67205524c3", size = 38319710 },
- { url = "https://files.pythonhosted.org/packages/ac/13/37737ef2193e83862ccacff23580c39de251da456a1bf0459e762cca273c/av-15.1.0-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:11326f197e7001c4ca53a83b2dbc67fd39ddff8cdf62ce6be3b22d9f3f9338bd", size = 39915519 },
- { url = "https://files.pythonhosted.org/packages/26/e9/e8032c7b8f2a4129a03f63f896544f8b7cf068e2db2950326fa2400d5c47/av-15.1.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:a631ea879cc553080ee62874f4284765c42ba08ee0279851a98a85e2ceb3cc8d", size = 40286166 },
- { url = "https://files.pythonhosted.org/packages/e2/23/612c0fd809444d04b8387a2dfd942ccc77829507bd78a387ff65a9d98c24/av-15.1.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:8f383949b010c3e731c245f80351d19dc0c08f345e194fc46becb1cb279be3ff", size = 41150592 },
- { url = "https://files.pythonhosted.org/packages/15/74/6f8e38a3b0aea5f28e72813672ff45b64615f2c69e6a4a558718c95edb9f/av-15.1.0-cp312-cp312-win_amd64.whl", hash = "sha256:d5921aa45f4c1f8c1a8d8185eb347e02aa4c3071278a2e2dd56368d54433d643", size = 31336093 },
- { url = "https://files.pythonhosted.org/packages/2e/bc/78b2ffa8235eeffc29aa4a8cc47b02e660cfec32f601f39a00975fb06d0e/av-15.1.0-cp313-cp313-macosx_13_0_arm64.whl", hash = "sha256:2f77853c3119c59d1bff4214ccbe46e3133eccff85ed96adee51c68684443f4e", size = 21726244 },
- { url = "https://files.pythonhosted.org/packages/1a/99/66d69453a2dce028e6e8ebea085d90e880aac03d3a3ab7d8ec16755ffd75/av-15.1.0-cp313-cp313-macosx_13_0_x86_64.whl", hash = "sha256:c0bc4471c156a0a1c70a607502434f477bc8dfe085eef905e55b4b0d66bcd3a5", size = 26918663 },
- { url = "https://files.pythonhosted.org/packages/fa/51/1a7dfbeda71f2772bc46d758af0e7fab1cc8388ce4bc7f24aecbc4bfd764/av-15.1.0-cp313-cp313-manylinux_2_28_aarch64.whl", hash = "sha256:37839d4fa1407f047af82560dfc0f94d8d6266071eff49e1cbe16c4483054621", size = 38041408 },
- { url = "https://files.pythonhosted.org/packages/d7/97/2c4e0288ad4359b6064cb06ae79c2ff3a84ac73d27e91f2161b75fcd86fa/av-15.1.0-cp313-cp313-manylinux_2_28_x86_64.whl", hash = "sha256:729179cd8622815e8b6f6854d13a806fe710576e08895c77e5e4ad254609de9a", size = 39642563 },
- { url = "https://files.pythonhosted.org/packages/ea/94/2362502149e276d00957edabcc201a5f4d5109a8a7b4fd30793714a532f3/av-15.1.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:4abdf085bfa4eec318efccff567831b361ea56c045cc38366811552e3127c665", size = 40022119 },
- { url = "https://files.pythonhosted.org/packages/df/58/1a0ce1b3835d9728da0a7a54aeffaa0a2b1a88405eaed9322efd55212a54/av-15.1.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:f985661644879e4520d28a995fcb2afeb951bc15a1d51412eb8e5f36da85b6fe", size = 40885158 },
- { url = "https://files.pythonhosted.org/packages/30/e6/054bb64e424d90b77ed5fc6a7358e4013fb436154c998fc90a89a186313f/av-15.1.0-cp313-cp313-win_amd64.whl", hash = "sha256:7d7804a44c8048bb4b014a99353dd124663a12cd1d4613ba2bd3b457c3b1d539", size = 31312256 },
- { url = "https://files.pythonhosted.org/packages/6f/8b/89eae6dca10d7d2b83c131025a31ccc750be78699ac0304439faa1d1df99/av-15.1.0-cp314-cp314-macosx_13_0_arm64.whl", hash = "sha256:5dd73c6447947edcb82e5fecf96e1f146aeda0f169c7ad4c54df4d9f66f63fde", size = 21730645 },
- { url = "https://files.pythonhosted.org/packages/a3/f0/abffaf69405ed68041524be12a1e294faf396971d6a0e70eb00e93687df7/av-15.1.0-cp314-cp314-macosx_13_0_x86_64.whl", hash = "sha256:a81cd515934a5d51290aa66b059b7ed29c4a212e704f3c5e99e32877ff1c312c", size = 26913753 },
- { url = "https://files.pythonhosted.org/packages/37/9e/7af078bcfc3cd340c981ac5d613c090ab007023d2ac13b05acd52f22f069/av-15.1.0-cp314-cp314-manylinux_2_28_aarch64.whl", hash = "sha256:57cc7a733a7e7d7a153682f35c9cf5d01e8269367b049c954779de36fc3d0b10", size = 38027048 },
- { url = "https://files.pythonhosted.org/packages/02/76/1f9dac11ad713e3619288993ea04e9c9cf4ec0f04e5ee81e83b8129dd8f3/av-15.1.0-cp314-cp314-manylinux_2_28_x86_64.whl", hash = "sha256:a77b75bdb6899a64302ff923a5246e0747b3f0a3ecee7d61118db407a22c3f53", size = 39565396 },
- { url = "https://files.pythonhosted.org/packages/8b/32/2188c46e2747247458ffc26b230c57dd28e61f65ff7b9e6223a411af5e98/av-15.1.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:d0a1154ce081f1720082a133cfe12356c59f62dad2b93a7a1844bf1dcd010d85", size = 40015050 },
- { url = "https://files.pythonhosted.org/packages/1e/41/b57fbce9994580619d7574817ece0fe0e7b822cde2af57904549d0150b8d/av-15.1.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:8a7bf5a34dee15c86790414fa86a144e6d0dcc788bc83b565fdcbc080b4fbc90", size = 40821225 },
- { url = "https://files.pythonhosted.org/packages/b1/36/e85cd1f0d3369c6764ad422882895d082f7ececb66d3df8aeae3234ef7a6/av-15.1.0-cp314-cp314-win_amd64.whl", hash = "sha256:e30c9a6fd9734784941384a2e25fad3c22881a7682f378914676aa7e795acdb7", size = 31311750 },
- { url = "https://files.pythonhosted.org/packages/80/d8/08a681758a4e49adfda409a6a35eff533f42654c6a6cfa102bc5cae1a728/av-15.1.0-cp314-cp314t-macosx_13_0_arm64.whl", hash = "sha256:60666833d7e65ebcfc48034a072de74349edbb62c9aaa3e6722fef31ca028eb6", size = 21828343 },
- { url = "https://files.pythonhosted.org/packages/4a/52/29bec3fe68669b21f7d1ab5d94e21f597b8dfd37f50a3e3c9af6a8da925c/av-15.1.0-cp314-cp314t-macosx_13_0_x86_64.whl", hash = "sha256:53fbdae45aa2a49a22e864ff4f4017416ef62c060a172085d3247ba0a101104e", size = 27001666 },
- { url = "https://files.pythonhosted.org/packages/9d/54/2c1d1faced66d708f5df328e800997cb47f90b500a214130c3a0f2ad601e/av-15.1.0-cp314-cp314t-manylinux_2_28_aarch64.whl", hash = "sha256:e6c51061667983dc801502aff9140bbc4f0e0d97f879586f17fb2f9a7e49c381", size = 39496753 },
- { url = "https://files.pythonhosted.org/packages/c3/76/06ded5e52c4dcc2d9b5184c6da8de5ea77bd7ecb79a59a2b9700f1984949/av-15.1.0-cp314-cp314t-manylinux_2_28_x86_64.whl", hash = "sha256:2f80ec387f04aa34868662b11018b5f09654ae1530a61e24e92a142a24b10b62", size = 40784729 },
- { url = "https://files.pythonhosted.org/packages/52/ef/797b76f3b39c99a96e387f501bbc07dca340b27d3dda12862fe694066b63/av-15.1.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:4975e03177d37d8165c99c8d494175675ba8acb72458fb5d7e43f746a53e0374", size = 41284953 },
- { url = "https://files.pythonhosted.org/packages/31/47/e4656f00e62fd059ea5a40b492dea784f5aecfe1dfac10c0d7a0664ce200/av-15.1.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:8f78f3dad11780b4cdd024cdb92ce43cb170929297c00f2f4555c2b103f51e55", size = 41985340 },
- { url = "https://files.pythonhosted.org/packages/b1/c9/15bb4fd7a1f39d70db35af2b9c20a0ae19e4220eb58a8b8446e903b98d72/av-15.1.0-cp314-cp314t-win_amd64.whl", hash = "sha256:9a20c5eba3ec49c2f4b281797021923fc68a86aeb66c5cda4fd0252fa8004951", size = 31487337 },
-]
-
[[package]]
name = "beautifulsoup4"
version = "4.14.2"
@@ -181,14 +140,12 @@ wheels = [
[[package]]
name = "chandra-ocr"
-version = "0.1.0"
+version = "0.2.0"
source = { editable = "." }
dependencies = [
- { name = "accelerate" },
{ name = "beautifulsoup4" },
{ name = "click" },
{ name = "filetype" },
- { name = "flask" },
{ name = "markdownify" },
{ name = "openai" },
{ name = "pillow" },
@@ -196,26 +153,42 @@ dependencies = [
{ name = "pydantic-settings" },
{ name = "pypdfium2" },
{ name = "python-dotenv" },
- { name = "qwen-vl-utils" },
+ { name = "six" },
+]
+
+[package.optional-dependencies]
+all = [
+ { name = "accelerate" },
{ name = "streamlit" },
{ name = "torch" },
{ name = "torchvision" },
{ name = "transformers" },
]
+app = [
+ { name = "streamlit" },
+]
+hf = [
+ { name = "accelerate" },
+ { name = "torch" },
+ { name = "torchvision" },
+ { name = "transformers" },
+]
[package.dev-dependencies]
dev = [
+ { name = "chandra-ocr", extra = ["all"] },
+ { name = "flask" },
{ name = "pre-commit" },
{ name = "pytest" },
]
[package.metadata]
requires-dist = [
- { name = "accelerate", specifier = ">=1.11.0" },
+ { name = "accelerate", marker = "extra == 'hf'", specifier = ">=1.11.0" },
{ name = "beautifulsoup4", specifier = ">=4.14.2" },
+ { name = "chandra-ocr", extras = ["hf", "app"], marker = "extra == 'all'" },
{ name = "click", specifier = ">=8.0.0" },
{ name = "filetype", specifier = ">=1.2.0" },
- { name = "flask", specifier = ">=3.0.0" },
{ name = "markdownify", specifier = "==1.1.0" },
{ name = "openai", specifier = ">=2.2.0" },
{ name = "pillow", specifier = ">=10.2.0" },
@@ -223,15 +196,17 @@ requires-dist = [
{ name = "pydantic-settings", specifier = ">=2.11.0" },
{ name = "pypdfium2", specifier = ">=4.30.0" },
{ name = "python-dotenv", specifier = ">=1.1.1" },
- { name = "qwen-vl-utils", specifier = ">=0.0.14" },
- { name = "streamlit", specifier = ">=1.50.0" },
- { name = "torch", specifier = ">=2.8.0" },
- { name = "torchvision", specifier = ">=0.23.0" },
- { name = "transformers", specifier = ">=4.57.1" },
+ { name = "six", specifier = ">=1.17.0" },
+ { name = "streamlit", marker = "extra == 'app'", specifier = ">=1.50.0" },
+ { name = "torch", marker = "extra == 'hf'", specifier = ">=2.8.0" },
+ { name = "torchvision", marker = "extra == 'hf'", specifier = ">=0.23.0" },
+ { name = "transformers", marker = "extra == 'hf'", specifier = ">=5.2.0" },
]
[package.metadata.requires-dev]
dev = [
+ { name = "chandra-ocr", extras = ["all"] },
+ { name = "flask", specifier = ">=3.0.0" },
{ name = "pre-commit", specifier = ">=4.3.0" },
{ name = "pytest", specifier = ">=8.4.2" },
]
@@ -430,17 +405,34 @@ wheels = [
[[package]]
name = "hf-xet"
-version = "1.1.10"
-source = { registry = "https://pypi.org/simple" }
-sdist = { url = "https://files.pythonhosted.org/packages/74/31/feeddfce1748c4a233ec1aa5b7396161c07ae1aa9b7bdbc9a72c3c7dd768/hf_xet-1.1.10.tar.gz", hash = "sha256:408aef343800a2102374a883f283ff29068055c111f003ff840733d3b715bb97", size = 487910 }
-wheels = [
- { url = "https://files.pythonhosted.org/packages/f7/a2/343e6d05de96908366bdc0081f2d8607d61200be2ac802769c4284cc65bd/hf_xet-1.1.10-cp37-abi3-macosx_10_12_x86_64.whl", hash = "sha256:686083aca1a6669bc85c21c0563551cbcdaa5cf7876a91f3d074a030b577231d", size = 2761466 },
- { url = "https://files.pythonhosted.org/packages/31/f9/6215f948ac8f17566ee27af6430ea72045e0418ce757260248b483f4183b/hf_xet-1.1.10-cp37-abi3-macosx_11_0_arm64.whl", hash = "sha256:71081925383b66b24eedff3013f8e6bbd41215c3338be4b94ba75fd75b21513b", size = 2623807 },
- { url = "https://files.pythonhosted.org/packages/15/07/86397573efefff941e100367bbda0b21496ffcdb34db7ab51912994c32a2/hf_xet-1.1.10-cp37-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6b6bceb6361c80c1cc42b5a7b4e3efd90e64630bcf11224dcac50ef30a47e435", size = 3186960 },
- { url = "https://files.pythonhosted.org/packages/01/a7/0b2e242b918cc30e1f91980f3c4b026ff2eedaf1e2ad96933bca164b2869/hf_xet-1.1.10-cp37-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:eae7c1fc8a664e54753ffc235e11427ca61f4b0477d757cc4eb9ae374b69f09c", size = 3087167 },
- { url = "https://files.pythonhosted.org/packages/4a/25/3e32ab61cc7145b11eee9d745988e2f0f4fafda81b25980eebf97d8cff15/hf_xet-1.1.10-cp37-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:0a0005fd08f002180f7a12d4e13b22be277725bc23ed0529f8add5c7a6309c06", size = 3248612 },
- { url = "https://files.pythonhosted.org/packages/2c/3d/ab7109e607ed321afaa690f557a9ada6d6d164ec852fd6bf9979665dc3d6/hf_xet-1.1.10-cp37-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:f900481cf6e362a6c549c61ff77468bd59d6dd082f3170a36acfef2eb6a6793f", size = 3353360 },
- { url = "https://files.pythonhosted.org/packages/ee/0e/471f0a21db36e71a2f1752767ad77e92d8cde24e974e03d662931b1305ec/hf_xet-1.1.10-cp37-abi3-win_amd64.whl", hash = "sha256:5f54b19cc347c13235ae7ee98b330c26dd65ef1df47e5316ffb1e87713ca7045", size = 2804691 },
+version = "1.4.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/09/08/23c84a26716382c89151b5b447b4beb19e3345f3a93d3b73009a71a57ad3/hf_xet-1.4.2.tar.gz", hash = "sha256:b7457b6b482d9e0743bd116363239b1fa904a5e65deede350fbc0c4ea67c71ea", size = 672357 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/18/06/e8cf74c3c48e5485c7acc5a990d0d8516cdfb5fdf80f799174f1287cc1b5/hf_xet-1.4.2-cp313-cp313t-macosx_10_12_x86_64.whl", hash = "sha256:ac8202ae1e664b2c15cdfc7298cbb25e80301ae596d602ef7870099a126fcad4", size = 3796125 },
+ { url = "https://files.pythonhosted.org/packages/66/d4/b73ebab01cbf60777323b7de9ef05550790451eb5172a220d6b9845385ec/hf_xet-1.4.2-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:6d2f8ee39fa9fba9af929f8c0d0482f8ee6e209179ad14a909b6ad78ffcb7c81", size = 3555985 },
+ { url = "https://files.pythonhosted.org/packages/ff/e7/ded6d1bd041c3f2bca9e913a0091adfe32371988e047dd3a68a2463c15a2/hf_xet-1.4.2-cp313-cp313t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:4642a6cf249c09da8c1f87fe50b24b2a3450b235bf8adb55700b52f0ea6e2eb6", size = 4212085 },
+ { url = "https://files.pythonhosted.org/packages/97/c1/a0a44d1f98934f7bdf17f7a915b934f9fca44bb826628c553589900f6df8/hf_xet-1.4.2-cp313-cp313t-manylinux_2_28_aarch64.whl", hash = "sha256:769431385e746c92dc05492dde6f687d304584b89c33d79def8367ace06cb555", size = 3988266 },
+ { url = "https://files.pythonhosted.org/packages/7a/82/be713b439060e7d1f1d93543c8053d4ef2fe7e6922c5b31642eaa26f3c4b/hf_xet-1.4.2-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:c9dd1c1bc4cc56168f81939b0e05b4c36dd2d28c13dc1364b17af89aa0082496", size = 4188513 },
+ { url = "https://files.pythonhosted.org/packages/21/a6/cbd4188b22abd80ebd0edbb2b3e87f2633e958983519980815fb8314eae5/hf_xet-1.4.2-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:fca58a2ae4e6f6755cc971ac6fcdf777ea9284d7e540e350bb000813b9a3008d", size = 4428287 },
+ { url = "https://files.pythonhosted.org/packages/b2/4e/84e45b25e2e3e903ed3db68d7eafa96dae9a1d1f6d0e7fc85120347a852f/hf_xet-1.4.2-cp313-cp313t-win_amd64.whl", hash = "sha256:163aab46854ccae0ab6a786f8edecbbfbaa38fcaa0184db6feceebf7000c93c0", size = 3665574 },
+ { url = "https://files.pythonhosted.org/packages/ee/71/c5ac2b9a7ae39c14e91973035286e73911c31980fe44e7b1d03730c00adc/hf_xet-1.4.2-cp313-cp313t-win_arm64.whl", hash = "sha256:09b138422ecbe50fd0c84d4da5ff537d27d487d3607183cd10e3e53f05188e82", size = 3528760 },
+ { url = "https://files.pythonhosted.org/packages/1e/0f/fcd2504015eab26358d8f0f232a1aed6b8d363a011adef83fe130bff88f7/hf_xet-1.4.2-cp314-cp314t-macosx_10_12_x86_64.whl", hash = "sha256:949dcf88b484bb9d9276ca83f6599e4aa03d493c08fc168c124ad10b2e6f75d7", size = 3796493 },
+ { url = "https://files.pythonhosted.org/packages/82/56/19c25105ff81731ca6d55a188b5de2aa99d7a2644c7aa9de1810d5d3b726/hf_xet-1.4.2-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:41659966020d59eb9559c57de2cde8128b706a26a64c60f0531fa2318f409418", size = 3555797 },
+ { url = "https://files.pythonhosted.org/packages/bf/e3/8933c073186849b5e06762aa89847991d913d10a95d1603eb7f2c3834086/hf_xet-1.4.2-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:5c588e21d80010119458dd5d02a69093f0d115d84e3467efe71ffb2c67c19146", size = 4212127 },
+ { url = "https://files.pythonhosted.org/packages/eb/01/f89ebba4e369b4ed699dcb60d3152753870996f41c6d22d3d7cac01310e1/hf_xet-1.4.2-cp314-cp314t-manylinux_2_28_aarch64.whl", hash = "sha256:a296744d771a8621ad1d50c098d7ab975d599800dae6d48528ba3944e5001ba0", size = 3987788 },
+ { url = "https://files.pythonhosted.org/packages/84/4d/8a53e5ffbc2cc33bbf755382ac1552c6d9af13f623ed125fe67cc3e6772f/hf_xet-1.4.2-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:f563f7efe49588b7d0629d18d36f46d1658fe7e08dce3fa3d6526e1c98315e2d", size = 4188315 },
+ { url = "https://files.pythonhosted.org/packages/d1/b8/b7a1c1b5592254bd67050632ebbc1b42cc48588bf4757cb03c2ef87e704a/hf_xet-1.4.2-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:5b2e0132c56d7ee1bf55bdb638c4b62e7106f6ac74f0b786fed499d5548c5570", size = 4428306 },
+ { url = "https://files.pythonhosted.org/packages/a0/0c/40779e45b20e11c7c5821a94135e0207080d6b3d76e7b78ccb413c6f839b/hf_xet-1.4.2-cp314-cp314t-win_amd64.whl", hash = "sha256:2f45c712c2fa1215713db10df6ac84b49d0e1c393465440e9cb1de73ecf7bbf6", size = 3665826 },
+ { url = "https://files.pythonhosted.org/packages/51/4c/e2688c8ad1760d7c30f7c429c79f35f825932581bc7c9ec811436d2f21a0/hf_xet-1.4.2-cp314-cp314t-win_arm64.whl", hash = "sha256:6d53df40616f7168abfccff100d232e9d460583b9d86fa4912c24845f192f2b8", size = 3529113 },
+ { url = "https://files.pythonhosted.org/packages/b4/86/b40b83a2ff03ef05c4478d2672b1fc2b9683ff870e2b25f4f3af240f2e7b/hf_xet-1.4.2-cp37-abi3-macosx_10_12_x86_64.whl", hash = "sha256:71f02d6e4cdd07f344f6844845d78518cc7186bd2bc52d37c3b73dc26a3b0bc5", size = 3800339 },
+ { url = "https://files.pythonhosted.org/packages/64/2e/af4475c32b4378b0e92a587adb1aa3ec53e3450fd3e5fe0372a874531c00/hf_xet-1.4.2-cp37-abi3-macosx_11_0_arm64.whl", hash = "sha256:e9b38d876e94d4bdcf650778d6ebbaa791dd28de08db9736c43faff06ede1b5a", size = 3559664 },
+ { url = "https://files.pythonhosted.org/packages/3c/4c/781267da3188db679e601de18112021a5cb16506fe86b246e22c5401a9c4/hf_xet-1.4.2-cp37-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:77e8c180b7ef12d8a96739a4e1e558847002afe9ea63b6f6358b2271a8bdda1c", size = 4217422 },
+ { url = "https://files.pythonhosted.org/packages/68/47/d6cf4a39ecf6c7705f887a46f6ef5c8455b44ad9eb0d391aa7e8a2ff7fea/hf_xet-1.4.2-cp37-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:c3b3c6a882016b94b6c210957502ff7877802d0dbda8ad142c8595db8b944271", size = 3992847 },
+ { url = "https://files.pythonhosted.org/packages/2d/ef/e80815061abff54697239803948abc665c6b1d237102c174f4f7a9a5ffc5/hf_xet-1.4.2-cp37-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:9d9a634cc929cfbaf2e1a50c0e532ae8c78fa98618426769480c58501e8c8ac2", size = 4193843 },
+ { url = "https://files.pythonhosted.org/packages/54/75/07f6aa680575d9646c4167db6407c41340cbe2357f5654c4e72a1b01ca14/hf_xet-1.4.2-cp37-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:6b0932eb8b10317ea78b7da6bab172b17be03bbcd7809383d8d5abd6a2233e04", size = 4432751 },
+ { url = "https://files.pythonhosted.org/packages/cd/71/193eabd7e7d4b903c4aa983a215509c6114915a5a237525ec562baddb868/hf_xet-1.4.2-cp37-abi3-win_amd64.whl", hash = "sha256:ad185719fb2e8ac26f88c8100562dbf9dbdcc3d9d2add00faa94b5f106aea53f", size = 3671149 },
+ { url = "https://files.pythonhosted.org/packages/b4/7e/ccf239da366b37ba7f0b36095450efae4a64980bdc7ec2f51354205fdf39/hf_xet-1.4.2-cp37-abi3-win_arm64.whl", hash = "sha256:32c012286b581f783653e718c1862aea5b9eb140631685bb0c5e7012c8719a87", size = 3533426 },
]
[[package]]
@@ -473,21 +465,22 @@ wheels = [
[[package]]
name = "huggingface-hub"
-version = "0.35.3"
+version = "1.7.1"
source = { registry = "https://pypi.org/simple" }
dependencies = [
{ name = "filelock" },
{ name = "fsspec" },
- { name = "hf-xet", marker = "platform_machine == 'aarch64' or platform_machine == 'amd64' or platform_machine == 'arm64' or platform_machine == 'x86_64'" },
+ { name = "hf-xet", marker = "platform_machine == 'AMD64' or platform_machine == 'aarch64' or platform_machine == 'amd64' or platform_machine == 'arm64' or platform_machine == 'x86_64'" },
+ { name = "httpx" },
{ name = "packaging" },
{ name = "pyyaml" },
- { name = "requests" },
{ name = "tqdm" },
+ { name = "typer" },
{ name = "typing-extensions" },
]
-sdist = { url = "https://files.pythonhosted.org/packages/10/7e/a0a97de7c73671863ca6b3f61fa12518caf35db37825e43d63a70956738c/huggingface_hub-0.35.3.tar.gz", hash = "sha256:350932eaa5cc6a4747efae85126ee220e4ef1b54e29d31c3b45c5612ddf0b32a", size = 461798 }
+sdist = { url = "https://files.pythonhosted.org/packages/b4/a8/94ccc0aec97b996a3a68f3e1fa06a4bd7185dd02bf22bfba794a0ade8440/huggingface_hub-1.7.1.tar.gz", hash = "sha256:be38fe66e9b03c027ad755cb9e4b87ff0303c98acf515b5d579690beb0bf3048", size = 722097 }
wheels = [
- { url = "https://files.pythonhosted.org/packages/31/a0/651f93d154cb72323358bf2bbae3e642bdb5d2f1bfc874d096f7cb159fa0/huggingface_hub-0.35.3-py3-none-any.whl", hash = "sha256:0e3a01829c19d86d03793e4577816fe3bdfc1602ac62c7fb220d593d351224ba", size = 564262 },
+ { url = "https://files.pythonhosted.org/packages/6f/75/ca21955d6117a394a482c7862ce96216239d0e3a53133ae8510727a8bcfa/huggingface_hub-1.7.1-py3-none-any.whl", hash = "sha256:38c6cce7419bbde8caac26a45ed22b0cea24152a8961565d70ec21f88752bfaa", size = 616308 },
]
[[package]]
@@ -637,6 +630,18 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/41/45/1a4ed80516f02155c51f51e8cedb3c1902296743db0bbc66608a0db2814f/jsonschema_specifications-2025.9.1-py3-none-any.whl", hash = "sha256:98802fee3a11ee76ecaca44429fda8a41bff98b00a0f2838151b113f210cc6fe", size = 18437 },
]
+[[package]]
+name = "markdown-it-py"
+version = "4.0.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+ { name = "mdurl" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/5b/f5/4ec618ed16cc4f8fb3b701563655a69816155e79e24a17b651541804721d/markdown_it_py-4.0.0.tar.gz", hash = "sha256:cb0a2b4aa34f932c007117b194e945bd74e0ec24133ceb5bac59009cda1cb9f3", size = 73070 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/94/54/e7d793b573f298e1c9013b8c4dade17d481164aa517d1d7148619c2cedbf/markdown_it_py-4.0.0-py3-none-any.whl", hash = "sha256:87327c59b172c5011896038353a81343b6754500a08cd7a4973bb48c6d578147", size = 87321 },
+]
+
[[package]]
name = "markdownify"
version = "1.1.0"
@@ -735,6 +740,15 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/70/bc/6f1c2f612465f5fa89b95bead1f44dcb607670fd42891d8fdcd5d039f4f4/markupsafe-3.0.3-cp314-cp314t-win_arm64.whl", hash = "sha256:32001d6a8fc98c8cb5c947787c5d08b0a50663d139f1305bac5885d98d9b40fa", size = 14146 },
]
+[[package]]
+name = "mdurl"
+version = "0.1.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d6/54/cfe61301667036ec958cb99bd3efefba235e65cdeb9c84d24a8293ba1d90/mdurl-0.1.2.tar.gz", hash = "sha256:bb413d29f5eea38f31dd4754dd7377d4465116fb207585f97bf925588687c1ba", size = 8729 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/b3/38/89ba8ad64ae25be8de66a6d463314cf1eb366222074cfda9ee839c56a4b4/mdurl-0.1.2-py3-none-any.whl", hash = "sha256:84008a41e51615a49fc9966191ff91509e3c40b939176e643fd50a5c2196b8f8", size = 9979 },
+]
+
[[package]]
name = "mpmath"
version = "1.3.0"
@@ -1617,21 +1631,6 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/f1/12/de94a39c2ef588c7e6455cfbe7343d3b2dc9d6b6b2f40c4c6565744c873d/pyyaml-6.0.3-cp314-cp314t-win_arm64.whl", hash = "sha256:ebc55a14a21cb14062aa4162f906cd962b28e2e9ea38f9b4391244cd8de4ae0b", size = 149341 },
]
-[[package]]
-name = "qwen-vl-utils"
-version = "0.0.14"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
- { name = "av" },
- { name = "packaging" },
- { name = "pillow" },
- { name = "requests" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/b6/b1/ad4fc2260a3badd278b38d642f3b987412f1f6682f0ef2b31b0572d5caa8/qwen_vl_utils-0.0.14.tar.gz", hash = "sha256:9c7cad5ae803b3a10f8bb7194deb12aeacdd032f92f4224e880c73587a7346ad", size = 8453 }
-wheels = [
- { url = "https://files.pythonhosted.org/packages/c4/43/80f67e0336cb2fc725f8e06f7fe35c1d0fe946f4d2b8b2175e797e07349e/qwen_vl_utils-0.0.14-py3-none-any.whl", hash = "sha256:5e28657bfd031e56bd447c5901b58ddfc3835285ed100f4c56580e0ade054e96", size = 8120 },
-]
-
[[package]]
name = "referencing"
version = "0.36.2"
@@ -1768,6 +1767,19 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/1e/db/4254e3eabe8020b458f1a747140d32277ec7a271daf1d235b70dc0b4e6e3/requests-2.32.5-py3-none-any.whl", hash = "sha256:2462f94637a34fd532264295e186976db0f5d453d1cdd31473c85a6a161affb6", size = 64738 },
]
+[[package]]
+name = "rich"
+version = "14.3.3"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+ { name = "markdown-it-py" },
+ { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b3/c6/f3b320c27991c46f43ee9d856302c70dc2d0fb2dba4842ff739d5f46b393/rich-14.3.3.tar.gz", hash = "sha256:b8daa0b9e4eef54dd8cf7c86c03713f53241884e814f4e2f5fb342fe520f639b", size = 230582 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/14/25/b208c5683343959b670dc001595f2f3737e051da617f66c31f7c4fa93abc/rich-14.3.3-py3-none-any.whl", hash = "sha256:793431c1f8619afa7d3b52b2cdec859562b950ea0d4b6b505397612db8d5362d", size = 310458 },
+]
+
[[package]]
name = "rpds-py"
version = "0.27.1"
@@ -1934,6 +1946,15 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/a3/dc/17031897dae0efacfea57dfd3a82fdd2a2aeb58e0ff71b77b87e44edc772/setuptools-80.9.0-py3-none-any.whl", hash = "sha256:062d34222ad13e0cc312a4c02d73f059e86a4acbfbdea8f8f76b28c99f306922", size = 1201486 },
]
+[[package]]
+name = "shellingham"
+version = "1.5.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/58/15/8b3609fd3830ef7b27b655beb4b4e9c62313a4e8da8c676e142cc210d58e/shellingham-1.5.4.tar.gz", hash = "sha256:8dbca0739d487e5bd35ab3ca4b36e11c4078f3a234bfce294b0a0291363404de", size = 10310 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/e0/f9/0595336914c5619e5f28a1fb793285925a8cd4b432c9da0a987836c7f822/shellingham-1.5.4-py2.py3-none-any.whl", hash = "sha256:7ecfff8f2fd72616f7481040475a65b2bf8af90a56c89140852d1120324e8686", size = 9755 },
+]
+
[[package]]
name = "six"
version = "1.17.0"
@@ -2222,24 +2243,23 @@ wheels = [
[[package]]
name = "transformers"
-version = "4.57.1"
+version = "5.3.0"
source = { registry = "https://pypi.org/simple" }
dependencies = [
- { name = "filelock" },
{ name = "huggingface-hub" },
{ name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
{ name = "numpy", version = "2.3.3", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
{ name = "packaging" },
{ name = "pyyaml" },
{ name = "regex" },
- { name = "requests" },
{ name = "safetensors" },
{ name = "tokenizers" },
{ name = "tqdm" },
+ { name = "typer" },
]
-sdist = { url = "https://files.pythonhosted.org/packages/d6/68/a39307bcc4116a30b2106f2e689130a48de8bd8a1e635b5e1030e46fcd9e/transformers-4.57.1.tar.gz", hash = "sha256:f06c837959196c75039809636cd964b959f6604b75b8eeec6fdfc0440b89cc55", size = 10142511 }
+sdist = { url = "https://files.pythonhosted.org/packages/fc/1a/70e830d53ecc96ce69cfa8de38f163712d2b43ac52fbd743f39f56025c31/transformers-5.3.0.tar.gz", hash = "sha256:009555b364029da9e2946d41f1c5de9f15e6b1df46b189b7293f33a161b9c557", size = 8830831 }
wheels = [
- { url = "https://files.pythonhosted.org/packages/71/d3/c16c3b3cf7655a67db1144da94b021c200ac1303f82428f2beef6c2e72bb/transformers-4.57.1-py3-none-any.whl", hash = "sha256:b10d05da8fa67dc41644dbbf9bc45a44cb86ae33da6f9295f5fbf5b7890bd267", size = 11990925 },
+ { url = "https://files.pythonhosted.org/packages/b8/88/ae8320064e32679a5429a2c9ebbc05c2bf32cefb6e076f9b07f6d685a9b4/transformers-5.3.0-py3-none-any.whl", hash = "sha256:50ac8c89c3c7033444fb3f9f53138096b997ebb70d4b5e50a2e810bf12d3d29a", size = 10661827 },
]
[[package]]
@@ -2257,6 +2277,21 @@ wheels = [
{ url = "https://files.pythonhosted.org/packages/20/63/8cb444ad5cdb25d999b7d647abac25af0ee37d292afc009940c05b82dda0/triton-3.4.0-cp313-cp313t-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:7936b18a3499ed62059414d7df563e6c163c5e16c3773678a3ee3d417865035d", size = 155659780 },
]
+[[package]]
+name = "typer"
+version = "0.24.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+ { name = "annotated-doc" },
+ { name = "click" },
+ { name = "rich" },
+ { name = "shellingham" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/f5/24/cb09efec5cc954f7f9b930bf8279447d24618bb6758d4f6adf2574c41780/typer-0.24.1.tar.gz", hash = "sha256:e39b4732d65fbdcde189ae76cf7cd48aeae72919dea1fdfc16593be016256b45", size = 118613 }
+wheels = [
+ { url = "https://files.pythonhosted.org/packages/4a/91/48db081e7a63bb37284f9fbcefda7c44c277b18b0e13fbc36ea2335b71e6/typer-0.24.1-py3-none-any.whl", hash = "sha256:112c1f0ce578bfb4cab9ffdabc68f031416ebcc216536611ba21f04e9aa84c9e", size = 56085 },
+]
+
[[package]]
name = "typing-extensions"
version = "4.15.0"
+Chandra OCR 2 is a state of the art OCR model that converts images and PDFs into structured HTML/Markdown/JSON while preserving layout information.
-## Benchmarks
+## News
-Overall scores on the [olmocr bench](https://github.com/allenai/olmocr):
+- 3/2026 - Chandra 2 is here with significant improvements to math, tables, layout, and multilingual OCR
+- 10/2025 - Chandra 1 launched
-
+## Features
-## Hosted API
+- Tops external olmocr benchmark and significant improvement in internal multilingual benchmarks
+- Convert documents to markdown, html, or json with detailed layout information
+- Support for 90+ languages ([benchmark below](#multilingual-benchmark))
+- Excellent handwriting support
+- Reconstructs forms accurately, including checkboxes
+- Strong performance with tables, math, and complex layouts
+- Extracts images and diagrams, and adds captions and structured data
+- Two inference modes: local (HuggingFace) and remote (vLLM server)
-A hosted API with additional accuracy improvements is available at [datalab.to](https://www.datalab.to/). Try the [free playground](https://www.datalab.to/playground) without installing.
+
-## Community
+## Hosted API
-Join [Discord](https://discord.gg//KuZwXNGnfH) to discuss development and get help.
+- We have a hosted API for Chandra [here](https://www.datalab.to/), which is more accurate and faster.
+- There is a free playground [here](https://www.datalab.to/playground) if you want to try Chandra without installing.
-## Quick Start
+## Quickstart
+
+The easiest way to start is with the CLI tools:
```shell
pip install chandra-ocr
-# Start vLLM server, then run OCR
+# With vLLM (recommended, lightweight install)
chandra_vllm
chandra input.pdf ./output
-# Or use HuggingFace locally
+# With HuggingFace (requires torch)
+pip install chandra-ocr[hf]
chandra input.pdf ./output --method hf
-# Interactive web app
+# Interactive streamlit app
+pip install chandra-ocr[app]
chandra_app
```
-**Python:**
-
-```python
-from chandra.model import InferenceManager
-from chandra.input import load_pdf_images
-
-manager = InferenceManager(method="hf")
-images = load_pdf_images("document.pdf")
-results = manager.generate(images)
-print(results[0].markdown)
-```
-
-## How it Works.
-
-- **Two inference modes**: Run locally via HuggingFace Transformers, or deploy a vLLM server for production throughput
-- **Layout-aware output**: Every text block, table, and image comes with bounding box coordinates
-- **Structured formats**: Output as Markdown, HTML, or JSON with full layout metadata
-- **40+ languages** supported
+## Benchmarks
-## What It Handles
+Multilingual performance was a focus for us with Chandra 2. There isn't a good public multilingual OCR benchmark, so we made our own. This tests tables, math, ordering, layout, and text accuracy.
-**Handwriting** — Doctor notes, filled forms, homework. Chandra reads cursive and messy print that trips up traditional OCR.
+
-**Tables** — Preserves structure including merged cells (colspan/rowspan). Works on financial filings, invoices, and data tables.
+See full scores [below](#multilingual-benchmark-table). We also have a [full 90-language benchmark](FULL_BENCHMARKS.md).
-**Math** — Inline and block equations rendered as LaTeX. Handles textbooks, worksheets, and research papers.
+We also benchmarked Chandra 2 with the widely accepted olmocr benchmark:
-**Forms** — Reconstructs checkboxes, radio buttons, and form fields with their values.
+
