Update main.py

main.py

@@ -16,7 +16,7 @@ from langchain.schema import StrOutputParser
 from langchain.schema.runnable import Runnable
 from langchain.schema.runnable.config import RunnableConfig
 from langchain.schema.runnable import Runnable, RunnablePassthrough, RunnableLambda
-from langchain_core.prompts import PromptTemplate
+from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
 
 @cl.password_auth_callback
 def auth_callback(username: str, password: str):
@@ -47,16 +47,6 @@ async def chat_profile():
         cl.ChatProfile(name="Imagestream",markdown_description="Requêter sur un ensemble d'images",icon="./public/logo-ofipe.jpg",),
     ]
     
-@cl.step(type="llm")
-async def IA():
-    os.environ['HUGGINGFACEHUB_API_TOKEN'] = os.environ['HUGGINGFACEHUB_API_TOKEN']
-    repo_id = "mistralai/Mixtral-8x7B-Instruct-v0.1"
-
-    llm = HuggingFaceEndpoint(
-        repo_id=repo_id, max_new_tokens=8000, temperature=1.0, task="text2text-generation", streaming=True
-    )
-    return llm
-    
 @cl.on_chat_start
 async def start():
     await cl.Message(f"> REVIEWSTREAM").send()
@@ -90,18 +80,14 @@ async def start():
     {context}
     {question} [/INST] </s>
     """
-    prompt = PromptTemplate(template=template, input_variables=["question","context"])
-    #llm = await IA()
-    #chain = (
-    #    RunnablePassthrough.assign(
-    #        history=RunnableLambda(memory.load_memory_variables) | itemgetter("history")
-    #    )
-    #    | prompt | llm
-    #)
-    #cl.user_session.set("runnable", chain)
     
-@cl.on_message
-async def main(message: cl.Message):
+    os.environ['HUGGINGFACEHUB_API_TOKEN'] = os.environ['HUGGINGFACEHUB_API_TOKEN']
+    repo_id = "mistralai/Mixtral-8x7B-Instruct-v0.1"
+
+    llm = HuggingFaceEndpoint(
+        repo_id=repo_id, max_new_tokens=8000, temperature=1.0, task="text2text-generation", streaming=True
+    )
+    
     os.environ['PINECONE_API_KEY'] = os.environ['PINECONE_API_KEY']
     embeddings = HuggingFaceEmbeddings()
     index_name = "all-venus"
@@ -109,16 +95,38 @@ async def main(message: cl.Message):
         api_key=os.environ['PINECONE_API_KEY']
     )
     index = pc.Index(index_name)
-    memory = cl.user_session.get("memory")
-    runnable = cl.user_session.get("runnable")
-    
     xq = embeddings.embed_query(message.content)
     xc = index.query(vector=xq, filter={"categorie": {"$eq": "bibliographie-OPP-DGDIN"}},top_k=150, include_metadata=True)
     context_p = ""
     for result in xc['matches']:
         context_p = context_p + result['metadata']['text']
+    
+    prompt = ChatPromptTemplate.from_messages(
+        [
+            (
+                "system",
+                f"Contexte : Vous êtes un chercheur de l'enseignement supérieur et vous êtes doué pour faire des analyses d'articles de recherche sur les thématiques liées à la pédagogie. En fonction des informations suivantes et du contexte suivant seulement et strictement. Contexte : {context_p}, réponds à la question suivante de la manière la plus pertinente, la plus exhaustive et la plus détaillée possible. ",
+            ),
+            MessagesPlaceholder(variable_name="history"),
+            ("human", "{question}, dans le contexte fourni."),
+        ]
+    )
+    runnable = (
+        RunnablePassthrough.assign(
+            history=RunnableLambda(memory.load_memory_variables) | itemgetter("history")
+        )
+        | prompt
+        | model
+    )
+    cl.user_session.set("runnable", runnable)
+    
+@cl.on_message
+async def main(message: cl.Message):
+    memory = cl.user_session.get("memory")
+    runnable = cl.user_session.get("runnable")
+    
     msg = cl.Message(author="Assistant Reviewstream",content="")
-    async for chunk in runnable.astream({"question": message.content,"context":context_p},
+    async for chunk in runnable.astream({"question": message.content},
             config=RunnableConfig(callbacks=[cl.AsyncLangchainCallbackHandler(stream_final_answer=True)])):
         await msg.stream_token(chunk)