feat: update knowledge

README.md

@@ -0,0 +1,14 @@
+# Depip agent
+## 1. Prequisite
+- Miniconda
+- Python3
+## 2. Install
+```sh
+conda env create --file environment.yml          
+conda activate depip-agent
+pip install -r requirements.txt  
+```
+## 3. Start app
+```sh
+fastapi run app/main.py
+```

app/agent/agent.py

@@ -14,9 +14,9 @@ def __init__(self, embedding_model: BedrockEmbedding, llm_model: BedrockLLM) ->
             "programmable_ip_license",
             "Searches and returns knowledge from the programmable ip license or story protocol",
         )
-
-        self.agent_executor = create_react_agent(llm_model.llm, tools=[tool], checkpointer=memory,
-                                                 state_modifier=SystemMessage("Your name always is Depip. I provided knowledge for you about story protocol and programmable ip license. Help human answer their questions"))
+        instruction = "You are Depip, an friendly agent has knowledge about story protocol and programmable ip license. Help human answer their questions about this context only."
+        self.agent_executor = create_react_agent(llm_model.llm, tools=[tool], 
+                                                 checkpointer=memory, state_modifier=instruction)
 
     def invoke(self, query: str, session_id: str):
         config = {

app/embedding_model/bedrock_embedding.py

@@ -4,6 +4,7 @@
 from depip_agent.app.utils.boto3_client import Boto3Client
 from langchain_text_splitters import RecursiveCharacterTextSplitter
 from langchain_core.vectorstores import InMemoryVectorStore, VectorStoreRetriever
+from langchain_community.vectorstores import FAISS
 class BedrockEmbedding:
     def __init__(self) -> None:
         boto3Client = Boto3Client()
@@ -23,3 +24,11 @@ def createEmbeddingPDF(self):
 
         self.retriever = vectorstore.as_retriever()
         print("--- %s seconds to create embedding ---" % (time.time() - start_time))
+
+    def loadVectorStore(self):
+        print('load vector store')
+        start_time = time.time()
+        faiss_index_path = "app/knowledge/faiss_index"
+        loaded_vectorstore = FAISS.load_local(folder_path=faiss_index_path, embeddings=self.embedding_model, allow_dangerous_deserialization=True)
+        self.retriever = loaded_vectorstore.as_retriever()
+        print("--- %s seconds to load vector store ---" % (time.time() - start_time))

app/llm_model/bedrock_llm.py

@@ -7,42 +7,42 @@
 from langchain_core.vectorstores import VectorStoreRetriever
 class BedrockLLM:
 
-    system_prompt = (
-        "You are an assistant for question-answering tasks. "
-        "Use the following pieces of retrieved context to answer "
-        "the question. If you don't know the answer, say that you "
-        "don't know. Use three sentences maximum and keep the "
-        "answer concise."
-        "\n\n"
-        "{context}"
-    )
-    contextualize_q_system_prompt = (
-        "Given a chat history and the latest user question "
-        "which might reference context in the chat history, "
-        "formulate a standalone question which can be understood "
-        "without the chat history. Do NOT answer the question, "
-        "just reformulate it if needed and otherwise return it as is."
-    )
-    prompt = ChatPromptTemplate.from_messages(
-        [
-            ("system", system_prompt),
-            ("human", "{input}"),
-        ]
-    )
+    # system_prompt = (
+    #     "You are an assistant for question-answering tasks. "
+    #     "Use the following pieces of retrieved context to answer "
+    #     "the question. If you don't know the answer, say that you "
+    #     "don't know. Use three sentences maximum and keep the "
+    #     "answer concise."
+    #     "\n\n"
+    #     "{context}"
+    # )
+    # contextualize_q_system_prompt = (
+    #     "Given a chat history and the latest user question "
+    #     "which might reference context in the chat history, "
+    #     "formulate a standalone question which can be understood "
+    #     "without the chat history. Do NOT answer the question, "
+    #     "just reformulate it if needed and otherwise return it as is."
+    # )
+    # prompt = ChatPromptTemplate.from_messages(
+    #     [
+    #         ("system", system_prompt),
+    #         ("human", "{input}"),
+    #     ]
+    # )
 
 
     def __init__(self, retriever: VectorStoreRetriever) -> None:
         boto3Client = Boto3Client()
 
         self.llm = ChatBedrock(client=boto3Client.client, model_id='anthropic.claude-3-sonnet-20240229-v1:0')
-        question_answer_chain = create_stuff_documents_chain(self.llm, self.prompt)
-        contextualize_q_prompt = ChatPromptTemplate.from_messages(
-            [
-                ("system", self.contextualize_q_system_prompt),
-                MessagesPlaceholder("chat_history"),
-                ("human", "{input}"),
-            ]
-        )
-        history_aware_retriever = create_history_aware_retriever(self.llm, retriever, contextualize_q_prompt)
-        self.rag_chain = create_retrieval_chain(history_aware_retriever, question_answer_chain)
+        # question_answer_chain = create_stuff_documents_chain(self.llm, self.prompt)
+        # contextualize_q_prompt = ChatPromptTemplate.from_messages(
+        #     [
+        #         ("system", self.contextualize_q_system_prompt),
+        #         MessagesPlaceholder("chat_history"),
+        #         ("human", "{input}"),
+        #     ]
+        # )
+        # history_aware_retriever = create_history_aware_retriever(self.llm, retriever, contextualize_q_prompt)
+        # self.rag_chain = create_retrieval_chain(history_aware_retriever, question_answer_chain)
 

app/main.py

@@ -8,7 +8,8 @@
 from depip_agent.app.dto import invoke_input
 
 embedding_model = BedrockEmbedding()
-embedding_model.createEmbeddingPDF()
+# embedding_model.createEmbeddingPDF()
+embedding_model.loadVectorStore()
 llm_model = BedrockLLM(embedding_model.retriever)
 agent = Agent(embedding_model=embedding_model, llm_model=llm_model)
 

environment.yml

@@ -1,7 +1,7 @@
-name: depip_agent
+name: depip-agent
 channels:
   - defaults
 dependencies:
-  - python
+  - python==3.12.1
   - pip:
     - -r file:requirements.txt

requirements.txt

@@ -13,6 +13,7 @@ click==8.1.7
 dataclasses-json==0.6.7
 dnspython==2.7.0
 email_validator==2.2.0
+faiss-cpu==1.9.0
 fastapi==0.115.4
 fastapi-cli==0.0.5
 frozenlist==1.5.0