from langchain.docstore.document import Document from langchain.vectorstores import FAISS from langchain.embeddings.openai import OpenAIEmbeddings from langchain.memory.simple import SimpleMemory from langchain.chains import ConversationChain, LLMChain, SequentialChain from langchain.memory import ConversationBufferMemory from langchain.prompts import ChatPromptTemplate, PromptTemplate from langchain.document_loaders import UnstructuredFileLoader from langchain.chat_models import ChatOpenAI from langchain.llms import OpenAI from langchain.memory import ConversationSummaryMemory from langchain.callbacks import PromptLayerCallbackHandler from langchain.prompts.chat import ( ChatPromptTemplate, SystemMessagePromptTemplate, AIMessagePromptTemplate, HumanMessagePromptTemplate, ) from langchain.schema import AIMessage, HumanMessage, SystemMessage from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler from langchain.callbacks.base import BaseCallbackHandler import gradio as gr from threading import Thread from queue import Queue, Empty from threading import Thread from collections.abc import Generator from langchain.llms import OpenAI from langchain.callbacks.base import BaseCallbackHandler import itertools import time import os import getpass import json import sys from typing import Any, Dict, List, Union import promptlayer import openai import gradio as gr from pydantic import BaseModel, Field, validator #Load the FAISS Model ( vector ) openai.api_key = os.environ["OPENAI_API_KEY"] db = FAISS.load_local("db", OpenAIEmbeddings()) #API Keys promptlayer.api_key = os.environ["PROMPTLAYER"] from langchain.callbacks import PromptLayerCallbackHandler from langchain.prompts.chat import ( ChatPromptTemplate, SystemMessagePromptTemplate, AIMessagePromptTemplate, HumanMessagePromptTemplate, ) from langchain.memory import ConversationSummaryMemory # Defined a QueueCallback, which takes as a Queue object during initialization. Each new token is pushed to the queue. class QueueCallback(BaseCallbackHandler): """Callback handler for streaming LLM responses to a queue.""" def __init__(self, q): self.q = q def on_llm_new_token(self, token: str, **kwargs: Any) -> None: self.q.put(token) def on_llm_end(self, *args, **kwargs: Any) -> None: return self.q.empty() class DDSAgent: def __init__(self, name, db, prompt_template='', model_name='gpt-4', verbose=False, temp=0.2): self.db = db self.verbose = verbose self.llm = ChatOpenAI( model_name="gpt-4", temperature=temp ) #The zero shot prompt provided at creation self.prompt_template = prompt_template #The LLM used for conversation summarization self.summary_llm = ChatOpenAI( model_name=model_name, max_tokens=25, callbacks=[PromptLayerCallbackHandler(pl_tags=["froebel"])], streaming=False, ) #Reviews convesation history and summarizes it to keep the token count down. self.memory = ConversationSummaryMemory(llm=self.summary_llm, max_token_limit=200, memory_key="memory", input_key="input") def chain(self, prompt: PromptTemplate, llm: ChatOpenAI) -> LLMChain: return LLMChain( llm=llm, prompt=prompt, verbose=self.verbose, memory=self.memory ) def lookup(self, input, num_docs=5): docs = self.db.similarity_search(input, k=num_docs) docs_to_string = "" for doc in docs: docs_to_string += str(doc.page_content) return docs_to_string def stream(self, input) -> Generator: # Create a Queue q = Queue() job_done = object() #RAG docs = self.lookup(input,5) llm = ChatOpenAI( model_name='gpt-4', callbacks=[QueueCallback(q), PromptLayerCallbackHandler(pl_tags=["froebel"])], streaming=True, ) prompt = PromptTemplate( input_variables=['input','docs','history'], template=self.prompt_template # partial_variables={"format_instructions": self.parser.get_format_instructions()} ) # Create a funciton to call - this will run in a thread def task(): resp = self.chain(prompt,llm).run( {'input':input, 'docs':docs, 'history':self.memory}) q.put(job_done) # Create a thread and start the function t = Thread(target=task) t.start() content = "" # Get each new token from the queue and yield for our generator while True: try: next_token = q.get(True, timeout=1) if next_token is job_done: break content += next_token yield next_token, content except Empty: continue agent_prompt = """ Roleplay You are a UBD ( Understanding by Design ) coach. Educators come to you to develop UBD based learning experiences and curriculum. This is the conversation up until now: {history} The teacher says: {input} As a result, following standards were matched: {docs} Respond to the teacher message. You have three objectives: a) to help them through the design process b) to help simplify the process for the educator c) to help build confidence and understand in the ubd process Take it step by step and keep. Keep focused on the current task at hand. Close with a single guiding step in the form of a question. Be encouraging. Do not start with "AI:" or any self identifying text. """ dds = DDSAgent('agent', db, prompt_template=agent_prompt) def ask_agent(input, history): for next_token, content in dds.stream(input): yield(content) gr.ChatInterface(ask_agent, title="UBD Coach", description=""" Using the Understanding By Design framework? I can help. (/◕ヮ◕)/ """, theme="monochrome", retry_btn=None, undo_btn=None, clear_btn=None ).queue().launch(debug=True)