feat: add Hugging Face OAuth

Chrys-assisted

Summary:
- Enable Gradio Sign in with Hugging Face for the Space.
- Use scoped visitor OAuth tokens for dataset creation and updates.
- Install Gradio OAuth support through the locked uv environment.

Validation:
- uv sync --frozen
- Python compile and OAuth behavior smoke checks
- uv lock --check
- uv build

Conversation: https://app.warp.dev/conversation/f2ab5d06-489f-47ae-abd7-e9eced4d755f

Co-Authored-By: Oz <oz-agent@warp.dev>

.gitignore

@@ -0,0 +1,6 @@
+.env
+__pycache__/
+*.pyc
+.venv/
+dist/
+*.egg-info/

README.md

@@ -7,6 +7,10 @@ sdk: gradio
 sdk_version: 5.20.1
 app_file: app.py
 pinned: false
+hf_oauth: true
+hf_oauth_scopes:
+  - contribute-repos
+  - write-repos
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -18,7 +18,7 @@ from datasets.features import Image as HFImage
 from huggingface_hub import HfApi, get_token
 import huggingface_hub
 import os
-from mistralai.client import Mistral
+from mistralai import Mistral
 import fitz  # pymupdf
 from PIL import Image
 import io
@@ -933,7 +933,7 @@ def get_hf_token(explicit_token: str | None = None) -> str | None:
 def process_files(
     file_paths: list[str],
     chunk_size: int,
-    hf_token: str,
+    hf_token: str | None,
     repo_name: str,
     append_mode: bool = False,
     min_chunk_chars: int = 20,
@@ -975,7 +975,7 @@ def process_files(
         return (
             "Error: No valid Hugging Face token found.\n"
             "Please either:\n"
-            "1. Provide a token in the input field (starts with 'hf_')\n"
+            "1. Sign in with Hugging Face when using the Space\n"
             "2. Set HF_TOKEN environment variable\n"
             "3. Run `huggingface-cli login` in your terminal"
         )
@@ -1356,9 +1356,17 @@ else:
 
 
 def _gradio_process(
-    file_objs, chunk_size, hf_token, repo_name, append_mode, min_chars, min_words
-):
-    """Bridge between Gradio file objects and core processing logic."""
+    file_objs: list[str] | str | None,
+    chunk_size: int,
+    repo_name: str,
+    append_mode: bool,
+    min_chars: int,
+    min_words: int,
+    oauth_token: gr.OAuthToken | None,
+) -> str:
+    """Bridge between Gradio inputs, the signed-in user, and core processing."""
+    if oauth_token is None:
+        return "Error: Sign in with Hugging Face before processing files."
     if not file_objs:
         return "Error: No files uploaded."
     if not isinstance(file_objs, list):
@@ -1367,7 +1375,7 @@ def _gradio_process(
     return process_files(
         file_paths,
         chunk_size,
-        hf_token,
+        oauth_token.token,
         repo_name,
         append_mode=append_mode,
         min_chunk_chars=min_chars,
@@ -1427,15 +1435,11 @@ with gr.Blocks(**GRADIO_BLOCKS_KWARGS) as demo:
 
                 with gr.Group(elem_classes=["mistral-section"]):
                     gr.Markdown("### Hugging Face output")
+                    gr.LoginButton()
                     repo_name = gr.Textbox(
                         label="Dataset repository",
                         placeholder="your-username/your-dataset-name",
                     )
-                    hf_token = gr.Textbox(
-                        label="Hugging Face token",
-                        type="password",
-                        placeholder="hf_... or set HF_TOKEN",
-                    )
                     append_mode = gr.Checkbox(
                         label="Append to existing dataset",
                         value=False,
@@ -1465,7 +1469,7 @@ with gr.Blocks(**GRADIO_BLOCKS_KWARGS) as demo:
             )
 
         gr.Markdown(
-            "*Requires `MISTRAL_API_KEY`, a Hugging Face token, or an active Hugging Face CLI login.*",
+            "*Requires `MISTRAL_API_KEY`. Sign in with Hugging Face before pushing a dataset.*",
             elem_classes=["mistral-note"],
         )
 
@@ -1480,7 +1484,6 @@ with gr.Blocks(**GRADIO_BLOCKS_KWARGS) as demo:
         inputs=[
             file_input,
             chunk_size,
-            hf_token,
             repo_name,
             append_mode,
             min_chars,
@@ -1491,14 +1494,13 @@ with gr.Blocks(**GRADIO_BLOCKS_KWARGS) as demo:
 
     gr.Examples(
         examples=[
-            [None, 512, "", "hf-username/my-first-ocr-dataset", False, 20, 3],
-            [None, 1024, "", "hf-username/large-chunk-ocr-data", True, 50, 5],
-            [None, 0, "", "hf-username/no-split-ocr-data", False, 0, 0],
+            [None, 512, "hf-username/my-first-ocr-dataset", False, 20, 3],
+            [None, 1024, "hf-username/large-chunk-ocr-data", True, 50, 5],
+            [None, 0, "hf-username/no-split-ocr-data", False, 0, 0],
         ],
         inputs=[
             file_input,
             chunk_size,
-            hf_token,
             repo_name,
             append_mode,
             min_chars,
@@ -1520,4 +1522,4 @@ def main():
 
 
 if __name__ == "__main__":
-    main()
+    main()

pyproject.toml

@@ -0,0 +1,30 @@
+[project]
+name = "pdf2dataset"
+version = "0.1.0"
+description = "Convert PDF and image files into structured Hugging Face datasets using Mistral OCR"
+readme = "README.md"
+requires-python = ">=3.10"
+license = { text = "MIT" }
+
+dependencies = [
+    "gradio[oauth]>=5.20.1",
+    "datasets>=3.0.0",
+    "huggingface_hub>=0.25.0",
+    "mistralai>=1.0.0",
+    "pillow>=10.0.0",
+    "chonkie>=1.0.0",
+    "jsonschema>=4.0.0",
+    "python-dotenv>=1.2.1",
+    "pymupdf>=1.24.0",
+]
+
+[project.scripts]
+pdf2dataset = "app:main"
+
+[tool.hatch.build.targets.wheel]
+packages = ["."]
+only-include = ["app.py"]
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"

requirements.txt

@@ -1,14 +0,0 @@
-gradio>=5.20.1
-datasets>=3.0.0
-huggingface_hub>=0.25.0
-mistralai>=1.0.0
-pillow>=10.0.0
-chonkie>=1.0.0
-jsonschema>=4.0.0
-python-dotenv>=1.2.1
-pymupdf>=1.24.0
-langchain
-langchain-community
-langchain-openai
-fastapi
-python-multipart