wehub-resource-sync
433 lines
13 KiB
Plaintext
433 lines
13 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"import os\n",
|
|
"from crewai import Agent, Task, Crew\n",
|
|
"\n",
|
|
"from langchain_openai import ChatOpenAI\n",
|
|
"\n",
|
|
"llm = ChatOpenAI(\n",
|
|
" openai_api_base=\"https://api.groq.com/openai/v1\",\n",
|
|
" openai_api_key=os.environ['GROQ_API_KEY'],\n",
|
|
" model_name=\"llama3-8b-8192\",\n",
|
|
" temperature=0,\n",
|
|
" max_tokens=1000,\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"from crewai_tools import PDFSearchTool\n",
|
|
"\n",
|
|
"rag_tool = PDFSearchTool(pdf='/content/17.pdf',\n",
|
|
" config=dict(\n",
|
|
" llm=dict(\n",
|
|
" provider=\"groq\", # or google, openai, anthropic, llama2, ...\n",
|
|
" config=dict(\n",
|
|
" model=\"llama3-8b-8192\",\n",
|
|
" # temperature=0.5,\n",
|
|
" # top_p=1,\n",
|
|
" # stream=true,\n",
|
|
" ),\n",
|
|
" ),\n",
|
|
" embedder=dict(\n",
|
|
" provider=\"huggingface\", # or openai, ollama, ...\n",
|
|
" config=dict(\n",
|
|
" model=\"BAAI/bge-small-en-v1.5\",\n",
|
|
" #task_type=\"retrieval_document\",\n",
|
|
" # title=\"Embeddings\",\n",
|
|
" ),\n",
|
|
" ),\n",
|
|
" )\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"rag_tool.run(\"How does exercise price determine for ESOP?\")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"import os\n",
|
|
"from langchain_community.tools.tavily_search import TavilySearchResults\n",
|
|
"os.environ['TAVILY_API_KEY'] = userdata.get('TAVILY_API_KEY')\n",
|
|
"web_search_tool = TavilySearchResults(k=3)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": []
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"web_search_tool.run(\"How does exercise price determine for ESOP?\")\n"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"from crewai_tools import tool\n",
|
|
"@tool\n",
|
|
"def router_tool(question):\n",
|
|
" \"\"\"Router Function\"\"\"\n",
|
|
" if 'ESOP' in question:\n",
|
|
" return 'vectorstore'\n",
|
|
" else:\n",
|
|
" return 'web_search'"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Router Agent"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"Router_Agent = Agent(\n",
|
|
" role='Router',\n",
|
|
" goal='Route user question to a vectorstore or web search',\n",
|
|
" backstory=(\n",
|
|
" \"You are an expert at routing a user question to a vectorstore or web search.\"\n",
|
|
" \"Use the vectorstore for questions on concepta related to Retrieval-Augmented Generation.\"\n",
|
|
" \"You do not need to be stringent with the keywords in the question related to these topics. Otherwise, use web-search.\"\n",
|
|
" ),\n",
|
|
" verbose=True,\n",
|
|
" allow_delegation=False,\n",
|
|
" llm=llm,\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Retriever_Agent"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"Retriever_Agent = Agent(\n",
|
|
"role=\"Retriever\",\n",
|
|
"goal=\"Use the information retrieved from the vectorstore to answer the question\",\n",
|
|
"backstory=(\n",
|
|
" \"You are an assistant for question-answering tasks.\"\n",
|
|
" \"Use the information present in the retrieved context to answer the question.\"\n",
|
|
" \"You have to provide a clear concise answer.\"\n",
|
|
"),\n",
|
|
"verbose=True,\n",
|
|
"allow_delegation=False,\n",
|
|
"llm=llm,\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Grader Agent"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"Grader_agent = Agent(\n",
|
|
" role='Answer Grader',\n",
|
|
" goal='Filter out erroneous retrievals',\n",
|
|
" backstory=(\n",
|
|
" \"You are a grader assessing relevance of a retrieved document to a user question.\"\n",
|
|
" \"If the document contains keywords related to the user question, grade it as relevant.\"\n",
|
|
" \"It does not need to be a stringent test.You have to make sure that the answer is relevant to the question.\"\n",
|
|
" ),\n",
|
|
" verbose=True,\n",
|
|
" allow_delegation=False,\n",
|
|
" llm=llm,\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Hallucination Grader Agent"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"hallucination_grader = Agent(\n",
|
|
" role=\"Hallucination Grader\",\n",
|
|
" goal=\"Filter out hallucination\",\n",
|
|
" backstory=(\n",
|
|
" \"You are a hallucination grader assessing whether an answer is grounded in / supported by a set of facts.\"\n",
|
|
" \"Make sure you meticulously review the answer and check if the response provided is in alignmnet with the question asked\"\n",
|
|
" ),\n",
|
|
" verbose=True,\n",
|
|
" allow_delegation=False,\n",
|
|
" llm=llm,\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Answer Grader Agent"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"answer_grader = Agent(\n",
|
|
" role=\"Answer Grader\",\n",
|
|
" goal=\"Filter out hallucination from the answer.\",\n",
|
|
" backstory=(\n",
|
|
" \"You are a grader assessing whether an answer is useful to resolve a question.\"\n",
|
|
" \"Make sure you meticulously review the answer and check if it makes sense for the question asked\"\n",
|
|
" \"If the answer is relevant generate a clear and concise response.\"\n",
|
|
" \"If the answer gnerated is not relevant then perform a websearch using 'web_search_tool'\"\n",
|
|
" ),\n",
|
|
" verbose=True,\n",
|
|
" allow_delegation=False,\n",
|
|
" llm=llm,\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Router Task"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"router_task = Task(\n",
|
|
" description=(\"Analyse the keywords in the question {question}\"\n",
|
|
" \"Based on the keywords decide whether it is eligible for a vectorstore search or a web search.\"\n",
|
|
" \"Return a single word 'vectorstore' if it is eligible for vectorstore search.\"\n",
|
|
" \"Return a single word 'websearch' if it is eligible for web search.\" \n",
|
|
" \"Do not provide any other premable or explaination.\"\n",
|
|
" ),\n",
|
|
" expected_output=(\"Give a binary choice 'websearch' or 'vectorstore' based on the question\"\n",
|
|
" \"Do not provide any other premable or explaination.\"),\n",
|
|
" agent=Router_Agent,\n",
|
|
" tools=[router_tool],\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Retriever Task"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"retriever_task = Task(\n",
|
|
" description=(\"Based on the response from the router task extract information for the question {question} with the help of the respective tool.\"\n",
|
|
" \"Use the web_serach_tool to retrieve information from the web in case the router task output is 'websearch'.\"\n",
|
|
" \"Use the rag_tool to retrieve information from the vectorstore in case the router task output is 'vectorstore'.\"\n",
|
|
" ),\n",
|
|
" expected_output=(\"You should analyse the output of the 'router_task'\"\n",
|
|
" \"If the response is 'websearch' then use the web_search_tool to retrieve information from the web.\"\n",
|
|
" \"If the response is 'vectorstore' then use the rag_tool to retrieve information from the vectorstore.\"\n",
|
|
" \"Return a claer and consise text as response.\"),\n",
|
|
" agent=Retriever_Agent,\n",
|
|
" context=[router_task],\n",
|
|
" #tools=[retriever_tool],\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Grader Task"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"grader_task = Task(\n",
|
|
" description=(\"Based on the response from the retriever task for the quetion {question} evaluate whether the retrieved content is relevant to the question.\"\n",
|
|
" ),\n",
|
|
" expected_output=(\"Binary score 'yes' or 'no' score to indicate whether the document is relevant to the question\"\n",
|
|
" \"You must answer 'yes' if the response from the 'retriever_task' is in alignment with the question asked.\"\n",
|
|
" \"You must answer 'no' if the response from the 'retriever_task' is not in alignment with the question asked.\"\n",
|
|
" \"Do not provide any preamble or explanations except for 'yes' or 'no'.\"),\n",
|
|
" agent=Grader_agent,\n",
|
|
" context=[retriever_task],\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Hallucination Grader Task"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"hallucination_task = Task(\n",
|
|
" description=(\"Based on the response from the grader task for the quetion {question} evaluate whether the answer is grounded in / supported by a set of facts.\"),\n",
|
|
" expected_output=(\"Binary score 'yes' or 'no' score to indicate whether the answer is sync with the question asked\"\n",
|
|
" \"Respond 'yes' if the answer is in useful and contains fact about the question asked.\"\n",
|
|
" \"Respond 'no' if the answer is not useful and does not contains fact about the question asked.\"\n",
|
|
" \"Do not provide any preamble or explanations except for 'yes' or 'no'.\"),\n",
|
|
" agent=hallucination_grader,\n",
|
|
" context=[grader_task],\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Answer grader Task"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"answer_task = Task( \n",
|
|
" description=(\"Based on the response from the hallucination task for the quetion {question} evaluate whether the answer is useful to resolve the question.\"\n",
|
|
" \"If the answer is 'yes' return a clear and concise answer.\"\n",
|
|
" \"If the answer is 'no' then perform a 'websearch' and return the response\"),\n",
|
|
" expected_output=(\"Return a clear and concise response if the response from 'hallucination_task' is 'yes'.\"\n",
|
|
" \"Perform a web search using 'web_search_tool' and return ta clear and concise response only if the response from 'hallucination_task' is 'no'.\"\n",
|
|
" \"Otherwise respond as 'Sorry! unable to find a valid response'.\"), \n",
|
|
" context=[hallucination_task],\n",
|
|
" agent=answer_grader,\n",
|
|
" #tools=[answer_grader_tool],\n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Setup the Crew"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"rag_crew = Crew(\n",
|
|
" agents=[Router_Agent, Retriever_Agent, Grader_agent, hallucination_grader, answer_grader],\n",
|
|
" tasks=[router_task, retriever_task, grader_task, hallucination_task, answer_task],\n",
|
|
" verbose=True,\n",
|
|
" \n",
|
|
")"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"inputs ={\"question\":\"Does the ESOP supplement the salary of an employee?\"}"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"result = rag_crew.kickoff(inputs=inputs)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"result"
|
|
]
|
|
}
|
|
],
|
|
"metadata": {
|
|
"kernelspec": {
|
|
"display_name": "env_crewai",
|
|
"language": "python",
|
|
"name": "python3"
|
|
},
|
|
"language_info": {
|
|
"name": "python",
|
|
"version": "3.10.15"
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 2
|
|
}
|