Upload 7 files

main.py

@@ -0,0 +1,212 @@
+import chromadb
+from llama_index.core.base.embeddings.base import similarity
+#from llama_index.llms.ollama import Ollama
+from llama_index.llms.groq import Groq
+from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings, DocumentSummaryIndex
+from llama_index.core import StorageContext, get_response_synthesizer
+from llama_index.core.retrievers import VectorIndexRetriever
+from llama_index.core.query_engine import RetrieverQueryEngine
+from llama_index.vector_stores.chroma import ChromaVectorStore
+from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+from llama_index.core import load_index_from_storage
+import os
+from dotenv import load_dotenv
+from llama_index.core.callbacks import CallbackManager, LlamaDebugHandler, CBEventType
+from llama_index.core.node_parser import SentenceSplitter
+from llama_index.core.postprocessor import SimilarityPostprocessor
+import time
+import gradio as gr
+from llama_index.core.memory import ChatMemoryBuffer
+from llama_parse import LlamaParse
+from llama_index.core import PromptTemplate
+from llama_index.core.llms import ChatMessage, MessageRole
+from llama_index.core.chat_engine import CondenseQuestionChatEngine
+
+
+load_dotenv()
+GROQ_API_KEY = os.getenv('GROQ_API_KEY')
+LLAMAINDEX_API_KEY = os.getenv('LLAMAINDEX_API_KEY')
+
+# set up callback manager
+llama_debug = LlamaDebugHandler(print_trace_on_end=True)
+callback_manager = CallbackManager([llama_debug])
+Settings.callback_manager = callback_manager
+
+# set up LLM
+llm = Groq(model="llama3-70b-8192")#"llama3-8b-8192")
+Settings.llm = llm
+
+# set up embedding model
+embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")
+Settings.embed_model = embed_model
+
+# create splitter
+splitter = SentenceSplitter(chunk_size=2048, chunk_overlap=50)
+Settings.transformations = [splitter]
+
+# create parser
+parser = LlamaParse(
+    api_key=LLAMAINDEX_API_KEY, 
+    result_type="markdown",  # "markdown" and "text" are available
+    verbose=True,
+)
+
+#create index
+if os.path.exists("./vectordb"):
+    print("Index Exists!")
+    storage_context = StorageContext.from_defaults(persist_dir="./vectordb")
+    index = load_index_from_storage(storage_context)
+else:
+    filename_fn = lambda filename: {"file_name": filename}
+    required_exts = [".pdf",".docx"]
+    file_extractor = {".pdf": parser}
+    reader = SimpleDirectoryReader(
+        input_dir="./data",
+        file_extractor=file_extractor,
+        required_exts=required_exts,
+        recursive=True,
+        file_metadata=filename_fn
+    )
+    documents = reader.load_data()
+    print("index creating with `%d` documents", len(documents))
+    index = VectorStoreIndex.from_documents(documents, embed_model=embed_model, transformations=[splitter])
+    index.storage_context.persist(persist_dir="./vectordb")
+
+"""
+#create document summary index
+if os.path.exists("./docsummarydb"):
+    print("Index Exists!")
+    storage_context = StorageContext.from_defaults(persist_dir="./docsummarydb")
+    doc_index = load_index_from_storage(storage_context)
+else:
+    filename_fn = lambda filename: {"file_name": filename}
+    required_exts = [".pdf",".docx"]
+    reader = SimpleDirectoryReader(
+        input_dir="./data",
+        required_exts=required_exts,
+        recursive=True,
+        file_metadata=filename_fn
+    )
+    documents = reader.load_data()
+    print("index creating with `%d` documents", len(documents))
+    
+    response_synthesizer = get_response_synthesizer(
+        response_mode="tree_summarize", use_async=True
+    )
+    doc_index = DocumentSummaryIndex.from_documents(
+        documents,
+        llm = llm,
+        transformations = [splitter],
+        response_synthesizer = response_synthesizer,
+        show_progress = True
+    )
+    doc_index.storage_context.persist(persist_dir="./docsummarydb")
+"""
+"""
+retriever = DocumentSummaryIndexEmbeddingRetriever(
+    doc_index,
+    similarity_top_k=5,
+)
+"""
+
+# set up retriever
+retriever = VectorIndexRetriever(
+    index = index,
+    similarity_top_k = 10,
+    #vector_store_query_mode="mmr",
+    #vector_store_kwargs={"mmr_threshold": 0.4}
+)
+
+# set up response synthesizer
+response_synthesizer = get_response_synthesizer()
+
+### customising prompts worsened the result###
+"""
+# set up prompt template
+qa_prompt_tmpl = (
+    "Context information from multiple sources is below.\n"
+    "---------------------\n"
+    "{context_str}\n"
+    "---------------------\n"
+    "Given the information from multiple sources and not prior knowledge, "
+    "answer the query.\n"
+    "Query: {query_str}\n"
+    "Answer: "
+)
+qa_prompt = PromptTemplate(qa_prompt_tmpl)
+"""
+# setting up query engine
+query_engine = RetrieverQueryEngine(
+    retriever = retriever,
+    node_postprocessors=[SimilarityPostprocessor(similarity_cutoff=0.53)],
+    response_synthesizer=get_response_synthesizer(response_mode="tree_summarize",verbose=True) 
+)
+print(query_engine.get_prompts())
+
+#response = query_engine.query("What happens if the distributor wants its own warehouse for pizzahood?")
+#print(response)
+
+
+memory = ChatMemoryBuffer.from_defaults(token_limit=10000)
+
+custom_prompt = PromptTemplate(
+    """\
+    Given a conversation (between Human and Assistant) and a follow up message from Human, \
+    rewrite the message to be a standalone question that captures all relevant context \
+    from the conversation. If you are unsure, ask for more information.
+
+    <Chat History>
+    {chat_history}
+
+    <Follow Up Message>
+    {question}
+
+    <Standalone question>
+    """
+)
+
+# list of `ChatMessage` objects
+custom_chat_history = [
+    ChatMessage(
+        role=MessageRole.USER,
+        content="Hello assistant.",
+    ),
+    ChatMessage(role=MessageRole.ASSISTANT, content="Hello user."),
+]
+
+chat_engine = CondenseQuestionChatEngine.from_defaults(
+    query_engine=query_engine,
+    condense_question_prompt=custom_prompt,
+    chat_history=custom_chat_history,
+    verbose=True,
+    memory=memory
+)
+
+
+# gradio with streaming support
+with gr.Blocks() as demo:
+    chat_engine = chat_engine
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox(label="⏎ for sending",
+            placeholder="Ask me something",)
+    clear = gr.Button("Delete")
+
+    def user(user_message, history):
+        return "", history + [[user_message, None]]
+
+    def bot(history):
+        user_message = history[-1][0]
+        #bot_message = chat_engine.chat(user_message)
+        bot_message = query_engine.query(user_message + "Let's think step by step to get the correct answer. If you cannot provide an answer, say you don't know.")
+        history[-1][1] = ""
+        for character in bot_message.response:
+            history[-1][1] += character
+            time.sleep(0.01)
+            yield history
+
+    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
+        bot, chatbot, chatbot
+    )
+    clear.click(lambda: None, None, chatbot, queue=False)
+# demo.queue()
+demo.launch(share=False)

requirements.txt

@@ -0,0 +1,13 @@
+docx2txt
+python-pptx
+torch
+pillow
+llama-index
+llama-index-llms-ollama
+llama-index-llms-groq
+llama-index-embeddings-huggingface
+llama-index-vector-stores-chroma
+llama-parse
+streamlit
+gradio
+groq

vectordb/graph_store.json

@@ -0,0 +1 @@
+{"graph_dict": {}}

vectordb/image__vector_store.json

@@ -0,0 +1 @@
+{"embedding_dict": {}, "text_id_to_ref_doc_id": {}, "metadata_dict": {}}

vectordb/index_store.json

@@ -0,0 +1 @@
+{"index_store/data": {"48f6023d-a0ad-42af-9a1b-61eb7b03baae": {"__type__": "vector_store", "__data__": "{\"index_id\": \"48f6023d-a0ad-42af-9a1b-61eb7b03baae\", \"summary\": null, \"nodes_dict\": {\"a3c26284-67c9-40a2-aca1-195c66f5ed3b\": \"a3c26284-67c9-40a2-aca1-195c66f5ed3b\", \"4b6e6913-3a74-46e5-93c0-fc960b1c029e\": \"4b6e6913-3a74-46e5-93c0-fc960b1c029e\", \"8438e420-d1ae-486d-aa3e-1b5845afe94b\": \"8438e420-d1ae-486d-aa3e-1b5845afe94b\", \"203991ea-aba5-4fba-b64c-3ce587ad56fe\": \"203991ea-aba5-4fba-b64c-3ce587ad56fe\", \"656a90d5-7ee7-4208-a427-e93827afc069\": \"656a90d5-7ee7-4208-a427-e93827afc069\", \"8906902c-cbc4-454c-b099-9d7d6c35377e\": \"8906902c-cbc4-454c-b099-9d7d6c35377e\", \"4b10db88-e5a4-490a-9b71-86bc3af4ece0\": \"4b10db88-e5a4-490a-9b71-86bc3af4ece0\", \"efc003b2-8c76-4b94-aeb5-e14df213138c\": \"efc003b2-8c76-4b94-aeb5-e14df213138c\", \"b77919e0-eec5-482e-b687-bbce4ae98a3a\": \"b77919e0-eec5-482e-b687-bbce4ae98a3a\", \"2b04b3b9-bbd3-4ac6-93a6-547d91e7303c\": \"2b04b3b9-bbd3-4ac6-93a6-547d91e7303c\", \"f0603697-7239-485d-a787-4484abe3a8ef\": \"f0603697-7239-485d-a787-4484abe3a8ef\", \"4566c459-0c09-4c13-bb8f-4e6b45faa2fe\": \"4566c459-0c09-4c13-bb8f-4e6b45faa2fe\", \"421bdc8b-6dba-4115-91ef-52d3ce3004a3\": \"421bdc8b-6dba-4115-91ef-52d3ce3004a3\", \"c21ed19c-da94-474d-a81d-5c3dca1003aa\": \"c21ed19c-da94-474d-a81d-5c3dca1003aa\", \"2af0f75f-e5f3-4fae-9f29-41bae9949bff\": \"2af0f75f-e5f3-4fae-9f29-41bae9949bff\", \"3c28d204-6963-4733-9647-dc97a05689ed\": \"3c28d204-6963-4733-9647-dc97a05689ed\", \"f9810544-3c78-423b-89ca-2cfc88beeace\": \"f9810544-3c78-423b-89ca-2cfc88beeace\", \"17da159c-3ce3-4910-a1f5-44282bfa8c16\": \"17da159c-3ce3-4910-a1f5-44282bfa8c16\", \"9eb20c0d-b7ec-472c-bf72-ce2daef7c864\": \"9eb20c0d-b7ec-472c-bf72-ce2daef7c864\", \"e87f76d4-d8c9-43fa-9d77-6b65dd7e29ce\": \"e87f76d4-d8c9-43fa-9d77-6b65dd7e29ce\", \"a399470b-eb05-44af-99fe-f7e38c3caecd\": \"a399470b-eb05-44af-99fe-f7e38c3caecd\", \"a3f0211f-ac45-4eb6-8f16-bb41dcd12eb8\": \"a3f0211f-ac45-4eb6-8f16-bb41dcd12eb8\", \"44cee3ec-3421-4a4e-bf33-cef03b55a6da\": \"44cee3ec-3421-4a4e-bf33-cef03b55a6da\", \"947f049c-76c2-4eb8-a548-6bd17ffc49a6\": \"947f049c-76c2-4eb8-a548-6bd17ffc49a6\", \"29e5a730-7783-4e13-98c8-b9b82ad9bf54\": \"29e5a730-7783-4e13-98c8-b9b82ad9bf54\", \"dd1a7215-2f76-43df-9e7e-6e891fb91964\": \"dd1a7215-2f76-43df-9e7e-6e891fb91964\", \"9c56f5d7-f81e-4c6b-b4a1-20e4bce63cc4\": \"9c56f5d7-f81e-4c6b-b4a1-20e4bce63cc4\", \"5b812df0-f254-444f-aa9f-4009970d677c\": \"5b812df0-f254-444f-aa9f-4009970d677c\", \"6bd7ff33-94a7-46b3-85fc-54c227301aea\": \"6bd7ff33-94a7-46b3-85fc-54c227301aea\", \"4ca9ea4b-6f18-42a7-b7b3-73d7caca9cc8\": \"4ca9ea4b-6f18-42a7-b7b3-73d7caca9cc8\", \"f1e12447-88d8-42c4-b3c3-b886f644d8bd\": \"f1e12447-88d8-42c4-b3c3-b886f644d8bd\", \"51127064-c4a7-4a8a-8505-f5033d6defe0\": \"51127064-c4a7-4a8a-8505-f5033d6defe0\", \"c0672079-149c-40df-9490-f6c7e679e93c\": \"c0672079-149c-40df-9490-f6c7e679e93c\", \"b44a4582-fcfa-4486-aa1d-ef598cbae369\": \"b44a4582-fcfa-4486-aa1d-ef598cbae369\", \"298f25b5-df2f-4950-ae37-6b1fc128c027\": \"298f25b5-df2f-4950-ae37-6b1fc128c027\", \"9677b70e-2e4e-4981-9c3a-82a48365ebe8\": \"9677b70e-2e4e-4981-9c3a-82a48365ebe8\", \"5efc76b5-9446-427a-b103-14715bc40dcc\": \"5efc76b5-9446-427a-b103-14715bc40dcc\", \"c853f933-d581-4852-b240-3b765a56dc00\": \"c853f933-d581-4852-b240-3b765a56dc00\", \"d5d042ec-42fd-4605-b20c-b4ad96cc8d89\": \"d5d042ec-42fd-4605-b20c-b4ad96cc8d89\", \"fd5173ae-7079-4408-851c-e0d8858eea8d\": \"fd5173ae-7079-4408-851c-e0d8858eea8d\", \"726d7911-8760-45db-b1f0-61969a95b408\": \"726d7911-8760-45db-b1f0-61969a95b408\", \"3e05e100-dee0-42fa-972f-c7b4c90b7d65\": \"3e05e100-dee0-42fa-972f-c7b4c90b7d65\", \"a0bb1b82-a7b6-4094-9f19-adf858b863da\": \"a0bb1b82-a7b6-4094-9f19-adf858b863da\", \"f608f2a3-c509-47c7-993f-7af25d9a95be\": \"f608f2a3-c509-47c7-993f-7af25d9a95be\", \"1341bd03-98d2-4c65-855c-47eec6c7bf4f\": \"1341bd03-98d2-4c65-855c-47eec6c7bf4f\", \"e3f7bd08-56a2-4bd9-9d14-9444fb4476b6\": \"e3f7bd08-56a2-4bd9-9d14-9444fb4476b6\", \"68838477-935f-466b-9c87-3e909edcd94d\": \"68838477-935f-466b-9c87-3e909edcd94d\", \"d6e642eb-dcf9-4bbf-a35c-c8d1ab7e0ce4\": \"d6e642eb-dcf9-4bbf-a35c-c8d1ab7e0ce4\", \"ece2bdc4-45a5-4dde-a857-79ce459094c4\": \"ece2bdc4-45a5-4dde-a857-79ce459094c4\", \"23011792-6904-433f-befb-362d6d557eb0\": \"23011792-6904-433f-befb-362d6d557eb0\", \"1a66fb1b-a184-463f-8d77-0d5751e50e3a\": \"1a66fb1b-a184-463f-8d77-0d5751e50e3a\", \"be50a3c4-1832-43ef-a5cc-f73d5aca8080\": \"be50a3c4-1832-43ef-a5cc-f73d5aca8080\", \"ddc6346e-fcf0-492f-8ca8-55497527dd60\": \"ddc6346e-fcf0-492f-8ca8-55497527dd60\", \"9bbac132-9124-45a8-b52a-d946db8a5d06\": \"9bbac132-9124-45a8-b52a-d946db8a5d06\", \"54b0313d-a70d-4b56-bd0b-6a92c81041c5\": \"54b0313d-a70d-4b56-bd0b-6a92c81041c5\", \"cd563902-254b-46a0-b0e7-fcb1e56a19c2\": \"cd563902-254b-46a0-b0e7-fcb1e56a19c2\", \"2a818b39-1082-41a4-9346-9362c611f298\": \"2a818b39-1082-41a4-9346-9362c611f298\", \"74e8ede2-5286-4dd6-b3ac-1a16f452e5a0\": \"74e8ede2-5286-4dd6-b3ac-1a16f452e5a0\", \"037c672c-a214-4f8f-ac4d-57e0bcb008da\": \"037c672c-a214-4f8f-ac4d-57e0bcb008da\", \"be15fd6d-49b3-48d1-a514-8bf1e2a371ff\": \"be15fd6d-49b3-48d1-a514-8bf1e2a371ff\", \"2712582a-92f1-4000-a95a-abd2f5bbd97e\": \"2712582a-92f1-4000-a95a-abd2f5bbd97e\", \"a97efe1b-4f5e-4c2f-b460-e71a2236e8d4\": \"a97efe1b-4f5e-4c2f-b460-e71a2236e8d4\", \"57881336-f839-4e44-9fdd-95cbe84e367f\": \"57881336-f839-4e44-9fdd-95cbe84e367f\", \"a33461ae-75ff-4532-ab83-b37457d63c18\": \"a33461ae-75ff-4532-ab83-b37457d63c18\", \"b312cb3a-b226-43d6-9a00-987b4917f87a\": \"b312cb3a-b226-43d6-9a00-987b4917f87a\", \"ea95dcf1-784d-4893-b424-6bca00bd8d9f\": \"ea95dcf1-784d-4893-b424-6bca00bd8d9f\", \"6b42afc7-4181-4785-ab43-f4bb09fc9bb9\": \"6b42afc7-4181-4785-ab43-f4bb09fc9bb9\", \"801568cd-3b2b-4108-922b-caea0ed2d363\": \"801568cd-3b2b-4108-922b-caea0ed2d363\", \"ee453ad7-30b5-464d-ad07-535141acfab7\": \"ee453ad7-30b5-464d-ad07-535141acfab7\", \"efce897c-19f3-48e3-a3ef-297ad13637df\": \"efce897c-19f3-48e3-a3ef-297ad13637df\", \"04b694ab-5732-4a0f-8720-2936bf2007b0\": \"04b694ab-5732-4a0f-8720-2936bf2007b0\", \"b9d65d34-dba7-403f-b8dc-31251745e9e1\": \"b9d65d34-dba7-403f-b8dc-31251745e9e1\", \"84f1380e-6ee2-434a-b5d4-ec5d5b0c8b2c\": \"84f1380e-6ee2-434a-b5d4-ec5d5b0c8b2c\", \"45048ae2-7b7c-4bbe-8d2c-0e03d62560ed\": \"45048ae2-7b7c-4bbe-8d2c-0e03d62560ed\", \"63ca5478-034b-437f-939f-4c04fb174cb7\": \"63ca5478-034b-437f-939f-4c04fb174cb7\", \"dc30fdd2-978c-492a-b9c0-ffbbedb755bc\": \"dc30fdd2-978c-492a-b9c0-ffbbedb755bc\", \"46ceae09-a108-4415-970a-58eabe1c6e65\": \"46ceae09-a108-4415-970a-58eabe1c6e65\", \"bb07f394-2a9c-44fb-8be8-724c1d91e8a6\": \"bb07f394-2a9c-44fb-8be8-724c1d91e8a6\", \"948f9f83-3ff1-4b79-b5f2-e9b533478138\": \"948f9f83-3ff1-4b79-b5f2-e9b533478138\", \"865c9a9c-cc9f-4c1a-a7ee-d4794292ef50\": \"865c9a9c-cc9f-4c1a-a7ee-d4794292ef50\", \"663055d9-883c-49a1-8278-74cbab4cd280\": \"663055d9-883c-49a1-8278-74cbab4cd280\", \"c9ddfdcf-fffc-493a-bfeb-1259131aaa11\": \"c9ddfdcf-fffc-493a-bfeb-1259131aaa11\", \"fda16941-587f-4bf8-b025-e87fefd41b70\": \"fda16941-587f-4bf8-b025-e87fefd41b70\", \"6cb8a292-8c4e-4883-8dcd-d4e041d21e4c\": \"6cb8a292-8c4e-4883-8dcd-d4e041d21e4c\", \"d2c8c914-9885-4dcf-ae68-5e93bde98184\": \"d2c8c914-9885-4dcf-ae68-5e93bde98184\", \"a1dcdaac-9783-47dc-a124-fbfd58ba5fd4\": \"a1dcdaac-9783-47dc-a124-fbfd58ba5fd4\", \"e6ab407f-56ae-42d1-bb13-73d34a44cc73\": \"e6ab407f-56ae-42d1-bb13-73d34a44cc73\"}, \"doc_id_dict\": {}, \"embeddings_dict\": {}}"}}}