add endpoint

handler.py

@@ -0,0 +1,56 @@
+# handler.py
+import os, json, time, tempfile, base64
+import torch
+from docling.document_converter import DocumentConverter
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from supabase import create_client
+from huggingface_hub import EndpointHandler
+
+class Handler(EndpointHandler):
+    def __init__(self):
+        from dotenv import load_dotenv
+        load_dotenv()
+        # supabase
+        self.supabase = create_client(
+          os.getenv("DATABASE_URL"),
+          os.getenv("SUPABASE_SERVICE_ROLE_KEY")
+        )
+        # model
+        device = "mps" if torch.backends.mps.is_available() else "cpu"
+        dtype  = torch.float16 if device=="mps" else torch.float32
+        self.model = AutoModelForCausalLM.from_pretrained(
+          "numind/NuExtract-1.5-tiny",
+          torch_dtype=dtype, trust_remote_code=True
+        ).to(device).eval()
+        self.tokenizer = AutoTokenizer.from_pretrained(
+          "numind/NuExtract-1.5-tiny", trust_remote_code=True
+        )
+
+    def __call__(self, payload):
+        """
+        Expects JSON: {
+          "inputs": "<raw text or base64-pdf>",
+          "is_pdf": false|true
+        }
+        """
+        text = payload["inputs"]
+        if payload.get("is_pdf"):
+            b = base64.b64decode(text)
+            with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as tmp:
+                tmp.write(b)
+                path = tmp.name
+            conv = DocumentConverter().convert(path)
+            text = conv.document.export_to_text()
+            os.remove(path)
+
+        # build your prompt + generate (same as predict_NuExtract)
+        prompt = (
+          "<|input|>\n### Instruction:\nRemplis la template JSON…\n"
+          "### Text:\n" + text + "\n<|output|>"
+        )
+        enc = self.tokenizer(prompt, return_tensors="pt",
+                              truncation=True).to(self.model.device)
+        out_ids = self.model.generate(**enc, max_new_tokens=1024)
+        out = self.tokenizer.decode(out_ids[0], skip_special_tokens=True)
+        result = out.split("<|output|>")[1]
+        return {"result": json.loads(result)}

requirements.txt

@@ -0,0 +1,86 @@
+annotated-types==0.7.0
+attrs==25.3.0
+beautifulsoup4==4.13.4
+certifi==2025.4.26
+charset-normalizer==3.4.2
+click==8.1.8
+dill==0.4.0
+docling==2.32.0
+docling-core==2.31.0
+docling-ibm-models==3.4.3
+docling-parse==4.0.1
+easyocr==1.7.2
+et_xmlfile==2.0.0
+filelock==3.18.0
+filetype==1.2.0
+fsspec==2025.3.2
+huggingface-hub==0.31.4
+idna==3.10
+imageio==2.37.0
+Jinja2==3.1.6
+jsonlines==3.1.0
+jsonref==1.1.0
+jsonschema==4.23.0
+jsonschema-specifications==2025.4.1
+latex2mathml==3.78.0
+lazy_loader==0.4
+lxml==5.4.0
+markdown-it-py==3.0.0
+marko==2.1.3
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mpire==2.10.2
+mpmath==1.3.0
+multiprocess==0.70.18
+networkx==3.4.2
+ninja==1.11.1.4
+numpy==2.2.6
+opencv-python-headless==4.11.0.86
+openpyxl==3.1.5
+packaging==25.0
+pandas==2.2.3
+pillow==11.2.1
+pluggy==1.6.0
+pyclipper==1.3.0.post6
+pydantic==2.11.4
+pydantic-settings==2.9.1
+pydantic_core==2.33.2
+Pygments==2.19.1
+pylatexenc==2.10
+pypdfium2==4.30.1
+python-bidi==0.6.6
+python-dateutil==2.9.0.post0
+python-docx==1.1.2
+python-dotenv==1.1.0
+python-pptx==1.0.2
+pytz==2025.2
+PyYAML==6.0.2
+referencing==0.36.2
+regex==2024.11.6
+requests==2.32.3
+rich==14.0.0
+rpds-py==0.25.0
+rtree==1.4.0
+safetensors==0.5.3
+scikit-image==0.25.2
+scipy==1.15.3
+semchunk==2.2.2
+setuptools==80.8.0
+shapely==2.1.1
+shellingham==1.5.4
+six==1.17.0
+soupsieve==2.7
+sympy==1.14.0
+tabulate==0.9.0
+tifffile==2025.5.10
+tokenizers==0.21.1
+torch==2.7.0
+torchvision==0.22.0
+tqdm==4.67.1
+transformers==4.51.3
+typer==0.15.4
+typing-inspection==0.4.0
+typing_extensions==4.13.2
+tzdata==2025.2
+urllib3==2.4.0
+XlsxWriter==3.2.3