Upload app.py with huggingface_hub

app.py

@@ -0,0 +1,94 @@
+import os
+import json
+import uuid
+from pathlib import Path
+import gradio as gr
+from collections import defaultdict
+from huggingface_hub import CommitScheduler
+from langchain_community.embeddings import SentenceTransformerEmbeddings
+from langchain_community.vectorstores import Chroma
+from openai import OpenAI
+
+# Set up OpenAI client (Hugging Face Inference API)
+client = OpenAI(
+    base_url="https://router.huggingface.co/featherless-ai/v1",
+    api_key="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",
+)
+
+# Vectorstore setup (embedding + ChromaDB)
+embedding_model = SentenceTransformerEmbeddings(model_name='thenlper/gte-large')
+vectorstore = Chroma(
+    collection_name='clause_index',
+    persist_directory="./clause_index",
+    embedding_function=embedding_model
+)
+
+# Log storage
+log_file = Path("logs/") / f"query_{uuid.uuid4()}.json"
+log_file.parent.mkdir(exist_ok=True)
+scheduler = CommitScheduler(
+    repo_id="legal-rag-output",
+    repo_type="dataset",
+    folder_path=log_file.parent,
+    path_in_repo="logs",
+    every=2
+)
+
+# Prompt Template
+system_message = """You are a legal AI assistant tasked with answering questions from legal contracts using only the provided context.
+Answer strictly from the context. If the answer is not found, respond: "Sorry, no relevant information found in the context."
+"""
+user_template = """
+###Context
+{context}
+
+###Question
+{question}
+"""
+
+def predict(question):
+    docs = vectorstore.similarity_search(question, k=3)
+    context = "\n\n".join([doc.page_content for doc in docs])
+    
+    prompt = [
+        {"role": "system", "content": system_message},
+        {"role": "user", "content": user_template.format(context=context, question=question)}
+    ]
+
+    try:
+        stream = client.chat.completions.create(
+            model="mistralai/Mistral-7B-Instruct-v0.2",
+            messages=prompt,
+            temperature=0.5,
+            top_p=0.7,
+            stream=True,
+        )
+        output = ""
+        for chunk in stream:
+            delta = chunk.choices[0].delta.content or ""
+            output += delta
+    except Exception as e:
+        output = f"Error: {str(e)}"
+
+    with scheduler.lock:
+        with log_file.open("a") as f:
+            f.write(json.dumps({
+                "question": question,
+                "context": context,
+                "response": output
+            }) + "\n")
+    
+    return output
+
+# Gradio UI
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Textbox(label="Enter your legal question:", lines=4),
+    outputs=gr.Textbox(label="Answer"),
+    title="⚖️ GL_LegalMind",
+    description="Ask legal contract-related questions. Answers are grounded in clause vector retrieval + Mistral LLM."
+)
+
+demo.queue()
+demo.launch()
+