langchain/docs/extras/integrations/llms/bittensor.ipynb

Bittensor

Bittensor is a mining network, similar to Bitcoin, that includes built-in incentives designed to encourage miners to contribute compute + knowledge.

NIBittensorLLM is developed by Neural Internet, powered by Bittensor.

This LLM showcases true potential of decentralized AI by giving you the best response(s) from the Bittensor protocol, which consist of various AI models such as OpenAI, LLaMA2 etc.

Users can view their logs, requests, and API keys on the Validator Endpoint Frontend. However, changes to the configuration are currently prohibited; otherwise, the user's queries will be blocked.

If you encounter any difficulties or have any questions, please feel free to reach out to our developer on GitHub, Discord or join our discord server for latest update and queries Neural Internet.

Different Parameter and response handling for NIBittensorLLM

In [ ]:

import langchain
from langchain.llms import NIBittensorLLM
import json
from pprint import pprint

langchain.debug = True

# System parameter in NIBittensorLLM is optional but you can set whatever you want to perform with model
llm_sys = NIBittensorLLM(
    system_prompt="Your task is to determine response based on user prompt.Explain me like I am technical lead of a project"
)
sys_resp = llm_sys(
    "What is bittensor and What are the potential benifits of decentralized AI?"
)
print(f"Response provided by LLM with system prompt set is : {sys_resp}")

# The top_responses parameter can give multiple responses based on its parameter value
# This below code retrive top 10 miner's response all the response are in format of json

# Json response structure is
""" {
    "choices":  [
                    {"index": Bittensor's Metagraph index number,
                    "uid": Unique Identifier of a miner,
                    "responder_hotkey": Hotkey of a miner,
                    "message":{"role":"assistant","content": Contains actual response},
                    "response_ms": Time in millisecond required to fetch response from a miner} 
                ]
    } """

multi_response_llm = NIBittensorLLM(top_responses=10)
multi_resp = multi_response_llm("What is Neural Network Feeding Mechanism?")
json_multi_resp = json.loads(multi_resp)
pprint(json_multi_resp)

Using NIBittensorLLM with LLMChain and PromptTemplate

In [ ]:

import langchain
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain
from langchain.llms import NIBittensorLLM

langchain.debug = True

template = """Question: {question}

Answer: Let's think step by step."""


prompt = PromptTemplate(template=template, input_variables=["question"])

# System parameter in NIBittensorLLM is optional but you can set whatever you want to perform with model
llm = NIBittensorLLM(system_prompt="Your task is to determine response based on user prompt.")

llm_chain = LLMChain(prompt=prompt, llm=llm)
question = "What is bittensor?"

llm_chain.run(question)

Using NIBittensorLLM with Conversational Agent and Google Search Tool

In [ ]:

from langchain.agents import (
    AgentType,
    initialize_agent,
    load_tools,
    ZeroShotAgent,
    Tool,
    AgentExecutor,
)
from langchain.memory import ConversationBufferMemory
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain.utilities import GoogleSearchAPIWrapper, SerpAPIWrapper
from langchain.llms import NIBittensorLLM

memory = ConversationBufferMemory(memory_key="chat_history")


prefix = """Answer prompt based on LLM if there is need to search something then use internet and observe internet result and give accurate reply of user questions also try to use authenticated sources"""
suffix = """Begin!
            {chat_history}
            Question: {input}
            {agent_scratchpad}"""

prompt = ZeroShotAgent.create_prompt(
    tools,
    prefix=prefix,
    suffix=suffix,
    input_variables=["input", "chat_history", "agent_scratchpad"],
)

llm = NIBittensorLLM(system_prompt="Your task is to determine response based on user prompt")

llm_chain = LLMChain(llm=llm, prompt=prompt)

memory = ConversationBufferMemory(memory_key="chat_history")

agent = ZeroShotAgent(llm_chain=llm_chain, tools=tools, verbose=True)
agent_chain = AgentExecutor.from_agent_and_tools(
    agent=agent, tools=tools, verbose=True, memory=memory
)

response = agent_chain.run(input=prompt)