added pipeline

app.py

@@ -1,6 +1,10 @@
-import numpy as np
+import tika
+tika.initVM()
+from tika import parser
+import pickle
 import gradio as gr
-from scipy.special import softmax
+from sklearn.pipeline import Pipeline
+from tempfile import _TemporaryFileWrapper
 
 doc_cls =  [
     "Договоры аренды", 
@@ -10,17 +14,34 @@ doc_cls =  [
     "Договоры поставки"
 ]
 
-def foo(doc):
-    probs = softmax(np.random.rand(len(doc_cls)))
-    return {d:p for d, p in zip(doc_cls, probs)}
+class Classifier:
+    def __init__(self, pipeline: Pipeline):
+        self.pipeline = pipeline
+    def __call__(self, doc: _TemporaryFileWrapper):
+        if not doc:
+            return
+        doc.seek(0)
+        buffer = doc.read(-1)
+        parsed = parser.from_buffer(buffer)
+        content = parsed["content"]
+        probs = self.pipeline.predict_proba([content])[0]
+        return {d:p for d, p in zip(doc_cls, probs)}
 
 def main():
+    tika.initVM()
+
+    with open("pipeline.pkl", "rb") as file:
+        pipeline: Pipeline = pickle.load(file)
+
+    classifier = Classifier(pipeline)
+
     with gr.Blocks() as demo:
         doc = gr.File(label="Документ")
         output = gr.Label(label="Результаты классификации")
         button = gr.Button(value="Классифицировать", variant="primary")
-        button.click(foo, doc, output)
+        button.click(classifier, doc, output)
     demo.launch()
+    
 
 if __name__ == "__main__":
     main()

pipeline.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e1ccfea510902c1199fe52d7621443287062ef5096be2db02102154ad3a5872
+size 4188941

requirements.txt

@@ -1 +1,3 @@
-scipy
+tika
+scikit-learn
+