Upload 4 files

llm.py

@@ -0,0 +1,16 @@
+from langchain_openai import OpenAI
+from langchain_community.llms import HuggingFaceHub
+from dotenv import load_dotenv
+import os
+
+load_dotenv()
+
+openai_key = os.getenv("OPENAI_API_KEY")
+hf_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
+
+llm = OpenAI(temperature=0.6, openai_api_key=openai_key)
+
+#! Alternatively, can use HuggingFace hub's LLM
+# llm = HuggingFaceHub(
+#     repo_id='google/flan-t5-large', model_kwargs={"temperature": 0.7, "max_length": 256}
+# )

main.py

@@ -0,0 +1,36 @@
+from fastapi import FastAPI
+from langserve import add_routes
+from langchain.chains import ConversationChain
+from memory import vectorstore_as_memory
+from prompt import PROMPT
+from llm import llm
+
+app = FastAPI(title="Retrieval App")
+
+# Initialize the conversation chain with a default memory
+memory = vectorstore_as_memory("USER1")
+final_chain = ConversationChain(
+    llm=llm,
+    prompt=PROMPT,
+    memory=memory,
+    verbose=False
+)
+
+# Define a function to update the memory associated with the final_chain
+def update_memory(username):
+    memory = vectorstore_as_memory(username)
+    final_chain.memory = memory
+
+# Define a route to handle API calls
+@app.post("/api/{username}")
+async def api_endpoint(username: str):
+    update_memory(username)
+    return {"message": f"Memory updated successfully with username: {username}"}
+
+# Add routes to the FastAPI app
+add_routes(app, final_chain)
+
+if __name__ == "__main__":
+    import uvicorn
+
+    uvicorn.run(app, host="localhost", port=8000)

memory.py

@@ -0,0 +1,66 @@
+from dotenv import load_dotenv
+import os
+from langchain.memory import VectorStoreRetrieverMemory
+from langchain_community.vectorstores.redis import Redis
+from langchain.embeddings import OpenAIEmbeddings
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain_core.runnables import ConfigurableField
+
+load_dotenv()
+
+redis_url = os.getenv("REDIS_URL")
+openai_key = os.getenv("OPENAI_API_KEY")
+hf_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
+
+
+embedding_fn = OpenAIEmbeddings(openai_api_key=openai_key)
+
+
+#! Alternatively, can use Hugging Face embeddings if you don't have one
+# modelPath = "HuggingFaceH4/zephyr-7b-beta"
+# model_kwargs = {'device':'cpu'}
+# encode_kwargs = {'normalize_embeddings':False}
+# embedding_fn = HuggingFaceEmbeddings(
+#   model_name = modelPath,  
+#   model_kwargs = model_kwargs,
+#   encode_kwargs=encode_kwargs
+# )
+
+schema = {'text': [{'name': 'content',
+   'weight': 1,
+   'no_stem': False,
+   'withsuffixtrie': False,
+   'no_index': False,
+   'sortable': False}],
+ 'vector': [{'name': 'content_vector',
+   'dims': 1536,
+   'algorithm': 'FLAT',
+   'datatype': 'FLOAT32',
+   'distance_metric': 'COSINE'}]}
+
+def vectorstore_as_memory(username):
+    try:
+        new_rds = Redis.from_existing_index(
+            embedding=embedding_fn,
+            index_name=username,
+            redis_url=redis_url,
+            # schema=rds.schema,
+            schema=schema,
+        )
+
+        retriever = new_rds.as_retriever(search_type="similarity", search_kwargs={"k": 3})
+        memory = VectorStoreRetrieverMemory(retriever=retriever)
+        return memory
+
+    except ValueError:
+        rds = Redis.from_texts(
+            texts=["Hi there"],
+            embedding=embedding_fn,
+            redis_url=redis_url,
+            index_name=username
+        )
+
+        retriever = rds.as_retriever(search_type="similarity", search_kwargs={"k": 3})
+        memory = VectorStoreRetrieverMemory(retriever=retriever)
+        return memory
+

prompt.py

@@ -0,0 +1,18 @@
+from langchain.prompts import PromptTemplate
+
+
+_DEFAULT_TEMPLATE = """The following is a friendly conversation between a human and an AI.
+The AI is talkative and provides lots of specific details from its context.
+If the AI does not know the answer to a question, it truthfully says it does not know.
+
+Relevant pieces of previous conversation:
+{history}
+
+(Note that you do not need to use these pieces of information if not relevant)
+
+Current conversation:
+Human: {input}
+AI:"""
+PROMPT = PromptTemplate(
+    input_variables=["history", "input"], template=_DEFAULT_TEMPLATE
+)