John is 12 years old, and wears a blue polo shirt.
His favorite color is blue, and likes to dance.
When he grows up, his dream is to be a farmer.

Use the following context section and only that context to answer the question at the end.
Do not use your internal knowledge. If you don't know the answer, just say that you don't know, don't try to make up an answer.

Context: {context}
Question: {query}
Answer:

{
  "model_file" : "ggml-model-gpt4all-falcon-q4_0.bin",
  "model_tokens" : 1000,
  "doc_file" : "2a-dummy.txt",
  "doc_chunks" : 512,
  "doc_overlap" : 20,
  "ws_host" : "localhost",
  "ws_port" : 5678
}

# (A) LOAD MODULES
import os, json
from langchain.document_loaders import TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import Chroma
from langchain.embeddings import GPT4AllEmbeddings
from langchain.llms import GPT4All
from langchain import PromptTemplate, LLMChain

# (B) SETTINGS
# (B1) SYSTEM PATH
set = { "path" : os.path.dirname(os.path.realpath(__file__)) }

# (B2) LOAD SETTINGS
with open(os.path.join(set["path"], "2c-settings.json"), "r") as fs:
  set.update(json.load(fs))
  fs.close()
  set["model_file"] = os.path.join(set["path"], "models", set["model_file"])
  set["doc_file"] = os.path.join(set["path"], set["doc_file"])

# (B3) LOAD PROMPT TEMPLATE
with open(os.path.join(set["path"], "2b-prompt.txt"), "r") as fs:
  set["prompt"] = fs.read()
  fs.close()

# (C) INIT
# (C1) CREATE VECTOR STORE FROM DOCUMENT
vstore = Chroma.from_documents(
  documents = RecursiveCharacterTextSplitter(
    chunk_size = set["doc_chunks"], chunk_overlap = set["doc_overlap"]
  ).split_documents(TextLoader(set["doc_file"]).load()),
  embedding = GPT4AllEmbeddings()
)

# (C2) LLM
llm = GPT4All(model = set["model_file"], max_tokens = set["model_tokens"])

# (D) Q&A SESSION
while True:
  # (D1) LISTEN TO QUERIES
  query = input("\nEnter a query: ")
  if query == "exit":
    break
  if query.strip() == "":
    continue

  # (D2) BUILD CONTEXT FROM QUERY & VECTOR STORE
  context = ""
  for doc in vstore.similarity_search(query):
    context = context + doc.page_content + " \n\n "

  # (D3) LLM CHAIN
  llm_chain = LLMChain(
    llm = llm,
    prompt = PromptTemplate(
      template = set["prompt"], input_variables = ["context", "query"]
    ).partial(context = context)
  )

   # (D4) OUTPUT RESULT
  print(llm_chain.run(query))

# A TO C ARE PRETTY MUCH THE SAME
# ONLY AN ADDITIONAL "IMPORT WEBSOCKETS"

# (D) Q&A SESSION
# (D1) ANSWER QUERY
async def qna (websocket):
  async for query in websocket:
    # (D1-1) BUILD CONTEXT FROM QUERY & VECTOR STORE
    context = ""
    for doc in vstore.similarity_search(query):
      context = context + doc.page_content + " \n\n "

    # (D1-2) LLM CHAIN
    llm_chain = LLMChain(
      llm = llm,
      prompt = PromptTemplate(
        template = set["prompt"], input_variables = ["context", "query"]
      ).partial(context = context)
    )

    # (D1-3) RESPOND
    await websocket.send(llm_chain.run(query))

# (D2) GO!
async def main ():
  async with websockets.serve(qna, set["ws_host"], set["ws_port"]):
    print("Server deployed at " + set["ws_host"] + ":" + str(set["ws_port"]))
    await asyncio.Future()

if __name__ == "__main__":
  asyncio.run(main())

<!-- (A) CHAT HISTORY -->
<div id="messages"></div>

<!-- (B) QUESTION -->
<form id="question" onsubmit="return chat.send()">
  <input type="text" id="qTxt" placeholder="Question" required disabled>
  <input type="submit" id="qGo" value="Go" disabled>
</form>

var chat = {
  // (A) SETTINGS & FLAGS
  host : "ws://localhost:5678/", socket : null,
  hMsg : null, hTxt : null, hGo : null,

  // (B) INIT
  init : () => {
    // (B1) GET HTML ELEMENTS
    chat.hMsg = document.getElementById("messages");
    chat.hTxt = document.getElementById("qTxt");
    chat.hGo = document.getElementById("qGo");

    // (B2) CONNECT TO WEBSOCKET
    chat.socket = new WebSocket(chat.host);

    // (B3) ON CONNECT - ENABLE QUERY FORM
    chat.socket.addEventListener("open", () => {
      chat.controls(1);
      chat.draw("Ready!", "sys");
    });

    // (B4) ON RECEIVE MESSAGE - DRAW IN HTML
    chat.socket.addEventListener("message", e => chat.draw(e.data, "bot"));

    // (B5) ON ERROR & CONNECTION LOST
    chat.socket.addEventListener("close", () => {
      chat.controls();
      chat.draw("Websocket connection lost!", "sys");
    });
    chat.socket.addEventListener("error", err => {
      chat.controls();
      console.log(err);
      chat.draw("Websocket connection error!", "sys");
    });
  },

  // (C) TOGGLE HTML CONTROLS
  controls : enable => {
    if (enable) {
      chat.hTxt.disabled = false;
      chat.hGo.disabled = false;
    } else {
      chat.hTxt.disabled = true;
      chat.hGo.disabled = true;
    }
  },

  // (D) SEND MESSAGE TO CHAT SERVER
  send : () => {
    chat.controls(); // disable question form until bot replies
    chat.draw(chat.hTxt.value, "human");
    chat.socket.send(chat.hTxt.value);
    chat.hTxt.value = "";
    return false;
  },

  // (E) DRAW MESSAGE IN HTML
  draw : (msg, css) => {
    let row = document.createElement("div");
    row.className = css;
    row.innerHTML = `<div class="chatName">${css}</div> <div class="chatMsg">${msg}</div>`;
    chat.hMsg.appendChild(row);
    if (css=="bot") { chat.controls(1); } // enable form on bot reply
    // window.scrollTo(0, 0);
  }
};
window.addEventListener("load", chat.init);