Create app.py

app.py

@@ -0,0 +1,186 @@
+import os
+import pandas as pd
+import  pathlib, fitz
+from langchain.vectorstores import Chroma
+# from PyPDF2 import PdfReader
+from google.colab import files
+from google.colab import userdata
+from sentence_transformers import SentenceTransformer
+from sentence_transformers.util import cos_sim
+from langchain.embeddings import HuggingFaceInstructEmbeddings
+from langchain.docstore.document import Document
+from huggingface_hub import InferenceClient
+import gradio as gr
+
+
+file_paths = ["docs/MANUU U.G. PROGRAMMES PROSPECTUS 2022-23 Eng 5 April 2022 4 PM.pdf","Prospectus 2023-24 (Eng-Version) (1)_0.pdf"]
+
+page_contents = []
+for fname in file_paths:
+  with fitz.open(fname) as doc:
+    print("Total Pages in {} are {}".format(fname,len(doc)))
+    for page in doc:
+      text = page.get_text()
+      if "............" in text:
+        continue
+      #print(text)
+      page_contents.append(text)
+      #break
+
+
+embedding_model = HuggingFaceInstructEmbeddings(
+            #model_name="hkunlp/instructor-large",
+            #model_name="jinaai/jina-embedding-b-en-v1",
+            model_name="WhereIsAI/UAE-Large-V1",
+            model_kwargs={"device": "cuda"}
+            #model_kwargs={"device": "cpu"}
+        )
+
+df_documents_chunks = pd.DataFrame({"doc_pages":page_contents})
+df_documents_chunks["index_id"] = df_documents_chunks.index
+print(df_documents_chunks)
+
+def row_to_doc(row):
+    return Document(metadata={
+            'id': row['index_id']
+        }, page_content=row['doc_pages'])
+
+
+manuuindex_df_processed_documents = df_documents_chunks.apply(lambda row:row_to_doc(row),axis=1).to_list()
+
+
+COLLECTION_NAME='Manuu_collection'
+PERSIST_DIR='MANUU_dir4'
+
+
+if os.path.exists(PERSIST_DIR):
+    print('Existing Collection : ', COLLECTION_NAME)
+    vectordb = Chroma(persist_directory=PERSIST_DIR, collection_name=COLLECTION_NAME, embedding_function=embedding_model)
+    print(f"Collection {vectordb._collection.name} has {vectordb._collection.count()} documents...")
+else:
+    print('New Collection : ', COLLECTION_NAME)
+    vectordb = Chroma.from_documents(documents=manuuindex_df_processed_documents,
+                                     embedding=embedding_model,
+                                     collection_name=COLLECTION_NAME,
+                                     persist_directory=PERSIST_DIR,
+                                     collection_metadata=None)
+    client = vectordb.persist() # Save vector database as persistent files in the output folder
+
+    print(f"Collection {vectordb._collection.name} has {vectordb._collection.count()} documents...")
+
+
+client = InferenceClient(
+    model = "mistralai/Mixtral-8x7B-Instruct-v0.1")
+
+
+def context_fn(question_text,vectordb):
+  relevant_chunks = vectordb.similarity_search_with_score(
+      query=question_text,
+    k=5,)
+  context_5 = "\n\n\n".join([i[0].page_content for i in relevant_chunks])
+
+  return context_5
+
+
+def format_prompt(message, history, context_prompt):
+  prompt = "<s>"
+  for user_prompt, bot_response in history:
+    prompt += f"[INST] {user_prompt}. Do not Give information from outside the Document Contexts and general Information[/INST]"
+    prompt += f" {bot_response}\n"
+    prompt +=  f" CONTEXT:{context_prompt}</s> "
+  prompt += f"[INST] {message} [/INST]"
+  with open('prompts.txt', 'a') as file:
+          print("user_prompt",prompt, file=file)
+          file.close()
+  return prompt
+
+
+def generate_fn(
+    prompt, history, system_prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,vectordb = vectordb
+):
+    temperature = float(temperature)
+    if temperature < 1e-2:
+        temperature = 1e-2
+    top_p = float(top_p)
+
+    generate_kwargs = dict(
+        temperature=temperature,
+        max_new_tokens=max_new_tokens,
+        top_p=top_p,
+        repetition_penalty=repetition_penalty,
+        do_sample=True,
+        seed=42,
+    )
+    context_5 = context_fn(question_text = prompt, vectordb = vectordb)
+    formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history, context_5)
+    #print("formatted_prompt",formatted_prompt)
+    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
+    output = ""
+
+    for response in stream:
+        output += response.token.text
+        yield output
+    return output
+additional_inputs=[
+    gr.Textbox(
+        label="System Prompt",
+        max_lines=1,
+        interactive=True,
+    ),
+    gr.Slider(
+        label="Temperature",
+        value=0.7,
+        minimum=0.0,
+        maximum=1.0,
+        step=0.05,
+        interactive=True,
+        info="Higher values produce more diverse outputs",
+    ),
+    gr.Slider(
+        label="Max new tokens",
+        value=256,
+        minimum=0,
+        maximum=2048,
+        step=64,
+        interactive=True,
+        info="The maximum numbers of new tokens",
+    ),
+    gr.Slider(
+        label="Top-p (nucleus sampling)",
+        value=0.90,
+        minimum=0.0,
+        maximum=1,
+        step=0.05,
+        interactive=True,
+        info="Higher values sample more low-probability tokens",
+    ),
+    gr.Slider(
+        label="Repetition penalty",
+        value=1.3,
+        minimum=1.0,
+        maximum=2.0,
+        step=0.05,
+        interactive=True,
+        info="Penalize repeated tokens",
+    )
+]
+
+examples=[["Where is Maulana Azad National Urdu University?", None, None, None, None, None,],
+          [ "When was Department of Women Education established?", None, None, None, None, None, ],
+          ["Tell me about  Department of Public Administration", None, None, None, None, None,],
+          ["What are Reservations for SCs/STs/OBCs /Women candidates/EWS Categories?", None, None, None, None, None,],
+          ["What is Upper Age Limit limit for Admissions", None, None, None, None, None,],
+          ["Fetch Details of Hostel Fee* (2022-23)?", None, None, None, None, None,],
+          ["What is Entrance Test Schedule 2023-24?", None, None, None, None, None,],
+         ]
+
+gr.ChatInterface(
+    fn=generate_fn,
+    analytics_enabled=True,
+
+    chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
+    additional_inputs=additional_inputs,
+    title="Mixtral 46.7B",
+    examples=examples,
+    concurrency_limit=20,
+).launch()