Create app.py

app.py

@@ -0,0 +1,212 @@
+import os
+os.environ["OPENAI_API_KEY"]
+
+from llama_index import (
+    VectorStoreIndex,
+    SummaryIndex,
+    SimpleKeywordTableIndex,
+    SimpleDirectoryReader,
+    ServiceContext,
+    StorageContext,
+    load_index_from_storage
+)
+from llama_index.schema import IndexNode
+from llama_index.tools import QueryEngineTool, ToolMetadata
+from llama_index.llms import OpenAI
+
+llm = OpenAI(temperature=0, model="gpt-3.5-turbo")
+service_context = ServiceContext.from_defaults(llm=llm)
+
+arv_index = load_index_from_storage(StorageContext.from_defaults(persist_dir = "./arv/"))
+# arv_summary_index = load_index_from_storage(StorageContext.from_defaults(persist_dir = "./arv_summary/"))
+arv_vector_query_engine = arv_index.as_query_engine(similarity_top_k = 3)
+# arv_summary_query_engine = arv_summary_index.as_query_engine()
+
+nishauri_index = load_index_from_storage(StorageContext.from_defaults(persist_dir = "./nishauri/"))
+# nishauri_summary_index = load_index_from_storage(StorageContext.from_defaults(persist_dir = "./nishauri_summary/"))
+nishauri_vector_query_engine = nishauri_index.as_query_engine(similarity_top_k = 3)
+# nishauri_summary_query_engine = nishauri_summary_index.as_query_engine()
+
+from llama_index.agent import OpenAIAgent
+
+agents = {}
+
+# define tools
+query_engine_tools = [
+    QueryEngineTool(
+        query_engine=arv_vector_query_engine,
+        metadata=ToolMetadata(
+            name="arv_vector_tool",
+            description=(
+                "Useful for retrieving specific context about HIV care and treatment."
+            ),
+        ),
+    ),
+    # QueryEngineTool(
+    #     query_engine=arv_summary_query_engine,
+    #     metadata=ToolMetadata(
+    #         name="arv_summary_tool",
+    #         description=(
+    #             "Useful for summarization questions related to HIV care and treatment."
+    #         ),
+    #     ),
+    # ),
+]
+
+# build agent
+function_llm = OpenAI(model="gpt-3.5-turbo-0613", temperature = 0)
+agent = OpenAIAgent.from_tools(
+    query_engine_tools,
+    llm=function_llm,
+    verbose=True,
+)
+
+agents["arv"] = agent
+
+# define tools
+query_engine_tools = [
+    QueryEngineTool(
+        query_engine=nishauri_vector_query_engine,
+        metadata=ToolMetadata(
+            name="nishauri_vector_tool",
+            description=(
+                "Useful for retrieving specific context about the Nishauri mobile application through which users are asking questions"
+            ),
+        ),
+    ),
+    # QueryEngineTool(
+    #     query_engine=nishauri_summary_query_engine,
+    #     metadata=ToolMetadata(
+    #         name="nishauri_summary_tool",
+    #         description=(
+    #             "Useful for summarization questions related to the Nishauri mobile application through which users are asking questions"
+    #         ),
+    #     ),
+    # ),
+]
+
+# build agent
+function_llm = OpenAI(model="gpt-3.5-turbo-0613", temperature = 0)
+agent = OpenAIAgent.from_tools(
+    query_engine_tools,
+    llm=function_llm,
+    verbose=True,
+)
+
+agents["nishauri"] = agent
+
+# define top-level nodes
+nodes = []
+
+arv_summary = (
+    "This content contains care and treatment guidance for people living with HIV."
+    " Use this source to answer questions about ARV medications, side effects from medication,"
+    " understanding viral loads, and any question about HIV care and treatment."
+    " This is the default source to use for answering any question that isn't about how to find"
+    " information in the Nishauri app."
+)
+node = IndexNode(text=arv_summary, index_id="arv")
+nodes.append(node)
+
+nishauri_summary = (
+    "This content contains guidance on the Nishauri mobile application through which users are asking questions."
+    " Reference this document when users ask questions such as how to find their viral load"
+    " or lab histories, how to find their appointment histories,"
+    " and want to know how to change their upcoming appointments."
+    " Do not use this to answer any other questions."
+)
+node = IndexNode(text=nishauri_summary, index_id="nishauri")
+nodes.append(node)
+
+# define top-level retriever
+vector_index = VectorStoreIndex(nodes)
+vector_retriever = vector_index.as_retriever(similarity_top_k=1)
+
+# define recursive retriever
+from llama_index.retrievers import RecursiveRetriever
+from llama_index.query_engine import RetrieverQueryEngine
+from llama_index.response_synthesizers import get_response_synthesizer
+
+# note: can pass `agents` dict as `query_engine_dict` since every agent can be used as a query engine
+recursive_retriever = RecursiveRetriever(
+    "vector",
+    retriever_dict={"vector": vector_retriever},
+    query_engine_dict=agents,
+    verbose=True,
+)
+
+response_synthesizer = get_response_synthesizer(
+    # service_context=service_context,
+    response_mode="compact",
+)
+query_engine = RetrieverQueryEngine.from_args(
+    recursive_retriever,
+    response_synthesizer=response_synthesizer,
+    service_context=service_context,
+)
+
+
+preamble = (" The person asking the following prompt is a person living with HIV in Kenya."
+            " For every response, recognize that they already have HIV and do not suggest that they have to get tested"
+            " for HIV or take post-exposure prophylaxis, as that is not relevant, though their partners perhaps should."  
+            " Do not suggest anything that is not relevant to someone who already has HIV."
+           " They are asking questions through a mobile application called Nishauri"
+           " through which they can see their lab results, appointment histories, and upcoming appointments."
+           " Here is some information that is authoritative and should guide responses, when relevant."
+           " For questions about viral load, be sure to provide specific information"
+           " about cutoffs for viral load categories. Under 50 copies/ml is low detectable level,"
+           " 50 - 199 copies/ml is low level viremia, 200 - 999 is high level viremia, and "
+           " 1000 and above is suspected treatment failure." 
+           " A high viral load or non-suppressed viral load is any viral load above 200 copies/ml."
+           " A suppressed viral load is one below 200 copies / ml."
+           " An established client is one who is on their current ART regimen for a period greater"
+           " than 6 months, had no active OI or in the previous 6 months, has adhered to scheduled"
+           " clinic visits for the previous 6 months and Viral load results has been less than 200 copies/ml"
+           " within the last 6 months."
+           " For questions about when patients should get their viral loads taken," 
+           " if they are newly initiated on ART, the first viral load sample should be taken after 3 months of"
+           " taking ART. Otherwise, if they are not new on ART, then if their previous result was below 50 to 199 cp/ml,"
+           " their viral load should be taken after every 12 months. If  their previous result was above 200cp/ml,"
+           " then viral load sample should be taken after three months."
+           " Please answer the prompt using the information retrieved" 
+           " and do not rely at all on your prior knowledge."
+           " Please keep your reply to no longer than three sentences, and please use simple language. ")
+
+prompt_intro = (" Here is the prompt: ")
+
+import gradio as gr
+
+# num_queries = 0
+# conversation_history = []
+# context = ""
+
+def nishauri(question: str, conversation_history: list[str]):
+    
+    # global num_queries, context
+    
+    # if num_queries == 0:
+        
+        # response = query_engine.query(preamble + prompt_intro + question)
+    
+    # if num_queries > 0:  
+    
+    context = " ".join([item["user"] + " " + item["chatbot"] for item in conversation_history])
+    response = query_engine.query(preamble +
+                                  "the user previously asked and received the following: " +
+                                  context +
+                                  prompt_intro +
+                                  question)
+
+    conversation_history.append({"user": question, "chatbot": response.response})   
+
+    # num_queries += 1
+    return response, conversation_history
+
+demo = gr.Interface(
+    title = "Nishauri Chatbot Demo",
+    fn=nishauri,
+    inputs=["text", gr.State(value=[])],
+    outputs=["text", gr.State()],
+)
+
+demo.launch()