stable

app.py

@@ -10,11 +10,14 @@ from dotenv import load_dotenv
 
 load_dotenv()
 
+from call_openai import call_openai
+
 from langchain import PromptTemplate
 from langchain.chains import LLMChain
 from langchain.chat_models import ChatOpenAI
 import pinecone
 from langchain.embeddings import OpenAIEmbeddings
+from langchain.vectorstores import Pinecone
 
 
 OPENAI_API_KEY=os.environ["OPENAI_API_KEY"]
@@ -42,7 +45,6 @@ template = """
 You are an expert research assistant. You can access information about articles via your tool. 
 Use information ONLY from this tool. Do not invent or add any more knowladge, be strict for the articles.
 {instuction}
-
 User: {user}
 --------
 {content}
@@ -64,23 +66,24 @@ embedder = OpenAIEmbeddings()
 
 
 class PineconeSearch:
-    docsearch
-    topk
+    docsearch = None
+    topk = 2
 
     def __init__(
+        self,
         namespace,
         topk
         ):
         self.docsearch = Pinecone.from_existing_index(PINECONE_INDEX, embedder, namespace=namespace)
         self.topk=topk
 
-    def __call__(query):
-        response = self.docsearch.similarity_search(query=query, k=self.topk)
+    def __call__(self,query):
+        docs = self.docsearch.similarity_search(query=query, k=self.topk)
         context = ""
         for doc in docs:
-            context += f"Coontent:\n{doc.page_content}\n"
-            context += f"Source: {doc.metadta['url']}\n"
-            contect += "----"
+            context += f"Content:\n{doc.page_content}\n"
+            context += f"Source: {doc.metadata['url']}\n"
+            context += "----"
         return context
 
 
@@ -111,6 +114,8 @@ def print_token_and_price(response):
     print(f"Total price: {price*370:.2f} Ft")
     print("===================================")
 
+agent_prompt = """You are an expert research assistant. You can access information about articles via your tool. 
+Use information ONLY from this tool. Do not invent or add any more knowladge, be strict for the articles."""
 
 
 def stream(input_text, history, user_prompt, topic, topk) -> Generator:
@@ -120,11 +125,22 @@ def stream(input_text, history, user_prompt, topic, topk) -> Generator:
 
     # Create a funciton to call - this will run in a thread
     def task():
-        chain = LLMChain(llm=llm, prompt=prompt, callbacks=[QueueCallback(q)])
+        #chain = LLMChain(llm=llm, prompt=prompt, callbacks=[QueueCallback(q)])
         
         tool_resp = query_tool(topic, topk, input_text)
 
-        response = chain({"instuction": user_prompt, "user": input_text, "content": tool_resp})
+        response = call_openai(
+            messages=[{"role": "system", "content": agent_prompt},
+             {"role": "system", "content": user_prompt},
+             {"role": "user", "content": input_text},
+             {"role": "system", "content": tool_resp}
+             ],
+             stream="token",
+             model="gpt-4-1106-preview",
+             callback=QueueCallback(q)
+        )
+
+        print(response)
 
         #print_token_and_price(response=response)
         q.put(job_done)
@@ -170,7 +186,7 @@ namespace_drobdown = gr.Dropdown(
     )
 topk_slider = gr.Slider(
     minimum=10,
-    maximum=350,
+    maximum=100,
     value=70,
     step=10
 )
@@ -182,4 +198,4 @@ chatInterface = gr.ChatInterface(
     fn=ask_llm,
     additional_inputs=additional_inputs,
     additional_inputs_accordion_name="Agent parameters"
-    ).queue().launch()
+    ).queue().launch()