wehub-resource-sync
517 lines
19 KiB
Plaintext
517 lines
19 KiB
Plaintext
{
|
||
"cells": [
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"# 100% local Agentic RAG"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 1,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"name": "stderr",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"/Users/akshaypachaar/miniconda3/envs/env_crewai/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
|
||
" from .autonotebook import tqdm as notebook_tqdm\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"import warnings\n",
|
||
"from crewai import Agent, Crew, Task, LLM, Process\n",
|
||
"from src.agentic_rag.tools.custom_tool import DocumentSearchTool\n",
|
||
"from src.agentic_rag.tools.custom_tool import FireCrawlWebSearchTool\n",
|
||
"\n",
|
||
"warnings.filterwarnings(\"ignore\")"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Setup LLM\n"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 2,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"llm = LLM(\n",
|
||
" model=\"ollama/llama3.2\",\n",
|
||
" base_url=\"http://localhost:11434\"\n",
|
||
")"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Setup Tools"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 3,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"pdf_tool = DocumentSearchTool(file_path='/Users/akshaypachaar/Eigen/ai-engineering/agentic_rag/knowledge/dspy.pdf')"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"Make sure you have the API key for FireCrawl in your environment variables."
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 4,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"web_search_tool = FireCrawlWebSearchTool()"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Agents"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 5,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"retriever_agent = Agent(\n",
|
||
" role=\"\"\"Retrieve relevant information to answer the user query: {query}\"\"\",\n",
|
||
" goal=\"\"\"Retrieve the most relevant information from the available sources \n",
|
||
" for the user query: {query}, always try to use the pdf search tool first. \n",
|
||
" If you are not able to retrieve the information from the pdf search tool \n",
|
||
" then try to use the web search tool.\"\"\",\n",
|
||
" backstory=\"\"\"You're a meticulous analyst with a keen eye for detail. \n",
|
||
" You're known for your ability understand the user query: {query} \n",
|
||
" and retrieve knowlege from the most suitable knowledge base.\"\"\",\n",
|
||
" verbose=True,\n",
|
||
" tools=[\n",
|
||
" pdf_tool,\n",
|
||
" web_search_tool\n",
|
||
" ],\n",
|
||
" # llm=llm\n",
|
||
")\n",
|
||
"\n",
|
||
"response_synthesizer_agent = Agent(\n",
|
||
" role=\"\"\"Response synthesizer agent for the user query: {query}\"\"\",\n",
|
||
" goal=\"\"\"Synthesize the retrieved information into a concise and coherent response \n",
|
||
" based on the user query: {query}. If you are not able to retrieve the \n",
|
||
" information then respond with \"I'm sorry, I couldn't find the information \n",
|
||
" you're looking for.\"\"\",\n",
|
||
" backstory=\"\"\"You're a skilled communicator with a knack for turning complex \n",
|
||
" information into clear and concise responses.\"\"\",\n",
|
||
" verbose=True,\n",
|
||
" # llm=llm\n",
|
||
")\n"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Tasks"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 6,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"retrieval_task = Task(\n",
|
||
" description=\"\"\"Retrieve the most relevant information from the available \n",
|
||
" sources for the user query: {query}\"\"\",\n",
|
||
" expected_output=\"\"\"The most relevant information in form of text as retrieved\n",
|
||
" from the sources.\"\"\",\n",
|
||
" agent=retriever_agent\n",
|
||
")\n",
|
||
"\n",
|
||
"response_task = Task(\n",
|
||
" description=\"\"\"Synthesize the final response for the user query: {query}\"\"\",\n",
|
||
" expected_output=\"\"\"A concise and coherent response based on the retrieved infromation\n",
|
||
" from the right source for the user query: {query}. If you are not \n",
|
||
" able to retrieve the information then respond with \n",
|
||
" I'm sorry, I couldn't find the information you're looking for.\"\"\",\n",
|
||
" agent=response_synthesizer_agent\n",
|
||
")"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Initialize Crew"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 7,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"crew = Crew(\n",
|
||
"\t\t\tagents=[retriever_agent, response_synthesizer_agent], \n",
|
||
"\t\t\ttasks=[retrieval_task, response_task],\n",
|
||
"\t\t\tprocess=Process.sequential,\n",
|
||
"\t\t\tverbose=True,\n",
|
||
"\t\t\t# process=Process.hierarchical, # In case you wanna use that instead https://docs.crewai.com/how-to/Hierarchical/\n",
|
||
"\t\t)"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Kickoff Crew"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 10,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"\u001b[1m\u001b[95m# Agent:\u001b[00m \u001b[1m\u001b[92mRetrieve relevant information to answer the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\u001b[95m## Task:\u001b[00m \u001b[92mRetrieve the most relevant information from the available \n",
|
||
" sources for the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\n",
|
||
"\n",
|
||
"\u001b[1m\u001b[95m# Agent:\u001b[00m \u001b[1m\u001b[92mRetrieve relevant information to answer the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\u001b[95m## Thought:\u001b[00m \u001b[92mI will start by searching the document for information regarding the dates of the Australian Open in 2025.\u001b[00m\n",
|
||
"\u001b[95m## Using tool:\u001b[00m \u001b[92mDocumentSearchTool\u001b[00m\n",
|
||
"\u001b[95m## Tool Input:\u001b[00m \u001b[92m\n",
|
||
"\"{\\\"query\\\": \\\"Australian Open 2025 dates\\\"}\"\u001b[00m\n",
|
||
"\u001b[95m## Tool Output:\u001b[00m \u001b[92m\n",
|
||
"02406, 2022.\n",
|
||
"\n",
|
||
"\n",
|
||
"___\n",
|
||
" In International Conference on Machine\n",
|
||
"Learning, pp.\n",
|
||
"___\n",
|
||
" arXiv preprint arXiv:2305.03495, 2023.\n",
|
||
"\n",
|
||
"Peng Qi, Xiaowen Lin, Leo Mehr, Zijian Wang, and Christopher D. Manning. Answering complex\n",
|
||
"In Proceedings of the 2019 Con-\n",
|
||
"\n",
|
||
"___\n",
|
||
"Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. Optuna:\n",
|
||
"In Proceedings of the 25th ACM\n",
|
||
"A next-generation hyperparameter optimization framework.\n",
|
||
"\n",
|
||
"___\n",
|
||
" arXiv preprint arXiv:1809.09600, 2018.\n",
|
||
"\n",
|
||
"\n",
|
||
"___\n",
|
||
"Q: Olivia has $23. She bought five bagels for $3 each. How much money does she have left?\n",
|
||
"\n",
|
||
"# solution in Python:\n",
|
||
"\n",
|
||
"def solution():\n",
|
||
"\n",
|
||
"\"\"\"Olivia has $23. She bought five bagels for $3 each. How much money does she have left?\"\"\"\n",
|
||
"money initial = 23\n",
|
||
"bagels = 5\n",
|
||
"bagel cost = 3\n",
|
||
"money spent = bagels * bagel cost\n",
|
||
"money left = money initial - money spent\n",
|
||
"result = money left\n",
|
||
"return result\n",
|
||
"\n",
|
||
"Q: Michael had 58 golf balls. On tuesday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls did he\n",
|
||
"have at the end of wednesday?\n",
|
||
"\n",
|
||
"# solution in Python:\n",
|
||
"\n",
|
||
"def solution():\n",
|
||
"\n",
|
||
"\"\"\"Michael had 58 golf balls. On tuesday, he lost 23 golf balls. On wednesday, he lost 2 more. How many golf balls\n",
|
||
"\n",
|
||
"did he have at the end of wednesday?\"\"\"\n",
|
||
"\n",
|
||
"golf balls initial = 58\n",
|
||
"golf balls lost tuesday = 23\n",
|
||
"golf balls lost wednesday = 2\n",
|
||
"golf balls left = golf balls initial - golf balls lost tuesday - golf balls lost wednesday\n",
|
||
"result = golf balls left\n",
|
||
"return result\n",
|
||
"\n",
|
||
"Q: There were nine computers in the server room. Five more computers were installed each day, from monday to thursday.\n",
|
||
"How many computers are now in the server room?\n",
|
||
"\n",
|
||
"# solution in Python:\n",
|
||
"\n",
|
||
"def solution():\n",
|
||
"\n",
|
||
"\"\"\"There were nine computers in the server room. Five more computers were installed each day, from monday to thursday.\n",
|
||
"\n",
|
||
"How many computers are now in the server room?\"\"\"\n",
|
||
"\n",
|
||
"computers initial = 9\n",
|
||
"computers per day = 5\n",
|
||
"num days = 4\n",
|
||
"computers added = computers per day * num days\n",
|
||
"computers total = computers initial + computers added\n",
|
||
"result = computers total\n",
|
||
"return result\n",
|
||
"\n",
|
||
"Q: Shawn has five toys. For Christmas, he got two toys each from his mom and dad. How many toys does he have now?\n",
|
||
"\n",
|
||
"# solution in Python:\n",
|
||
"\n",
|
||
"def solution():\n",
|
||
"\n",
|
||
"\"\"\"Shawn has five toys. For Christmas, he got two toys each from his mom and dad. How many toys does he have now?\"\"\"\n",
|
||
"toys initial = 5\n",
|
||
"mom toys = 2\n",
|
||
"dad toys = 2\n",
|
||
"total received = mom toys + dad toys\n",
|
||
"total toys = toys initial + total received\n",
|
||
"\n",
|
||
"___\n",
|
||
"I checked: {query}\n",
|
||
"\n",
|
||
"___\n",
|
||
"======== table info ========\n",
|
||
"{table info}\n",
|
||
"Question: {input}\n",
|
||
"SQLQuery:\n",
|
||
"\n",
|
||
"\n",
|
||
"___\n",
|
||
"# Step 2: Evaluate the generated candidate program .\n",
|
||
"score = evaluate_program ( candidate_program , self . metric , valset )\n",
|
||
"candidates . append (( score , candidate_program ) )\n",
|
||
"\n",
|
||
"# Create a new basic bootstrap few - shot program .\n",
|
||
"shuffled_trainset = shuffle ( trainset , seed = seed )\n",
|
||
"tp = BootstrapFewShot ( metric = metric , max_bootstrap_demos = random_size () )\n",
|
||
"candidate_program = tp . compile ( student , shuffled_trainset , teacher )\n",
|
||
"\n",
|
||
"28\n",
|
||
"\n",
|
||
"\fPreprint\n",
|
||
"\n",
|
||
"E.3 BOOTSTRAPFEWSHOTWITHOPTUNA\n",
|
||
"\n",
|
||
"pool = self . pool\n",
|
||
"\n",
|
||
"def objective ( self , trial ):\n",
|
||
"\n",
|
||
"def __init__ ( self , metric , trials =16) :\n",
|
||
"\n",
|
||
"self . metric = metric\n",
|
||
"self . trials = trials\n",
|
||
"\n",
|
||
"# Step 1: Create copy of student program .\n",
|
||
"candidate_program = self . student . reset_copy ()\n",
|
||
"\n",
|
||
"\n",
|
||
"___\n",
|
||
"1 class SimplifiedBootstrapFewShotWithOptuna ( Teleprompter ) :\n",
|
||
"2\n",
|
||
"3\n",
|
||
"4\n",
|
||
"5\n",
|
||
"6\n",
|
||
"7\n",
|
||
"8\n",
|
||
"9\n",
|
||
"10\n",
|
||
"11\n",
|
||
"12\n",
|
||
"13\n",
|
||
"14\n",
|
||
"15\n",
|
||
"16\n",
|
||
"17\n",
|
||
"18\n",
|
||
"19\n",
|
||
"20\n",
|
||
"21\n",
|
||
"22\n",
|
||
"23\n",
|
||
"24\n",
|
||
"25\n",
|
||
"26\n",
|
||
"27\n",
|
||
"28\n",
|
||
"29\n",
|
||
"30\n",
|
||
"31\n",
|
||
"32\n",
|
||
"33\n",
|
||
"34\n",
|
||
"35\n",
|
||
"36\n",
|
||
"37\n",
|
||
"38\n",
|
||
"39\n",
|
||
"40\n",
|
||
"41\n",
|
||
"42\n",
|
||
"43\n",
|
||
"44\n",
|
||
"\n",
|
||
"print ( ’ Best score : ’, best_program . score )\n",
|
||
"print ( ’ Best program : ’, best_program )\n",
|
||
"\u001b[00m\n",
|
||
"\n",
|
||
"\n",
|
||
"\u001b[1m\u001b[95m# Agent:\u001b[00m \u001b[1m\u001b[92mRetrieve relevant information to answer the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\u001b[95m## Thought:\u001b[00m \u001b[92mThought: It seems that the document search did not yield any relevant information regarding the dates for the Australian Open 2025. I will now search the internet to find the information.\u001b[00m\n",
|
||
"\u001b[95m## Using tool:\u001b[00m \u001b[92mSearch the internet\u001b[00m\n",
|
||
"\u001b[95m## Tool Input:\u001b[00m \u001b[92m\n",
|
||
"\"{\\\"search_query\\\": \\\"Australian Open 2025 schedule dates\\\"}\"\u001b[00m\n",
|
||
"\u001b[95m## Tool Output:\u001b[00m \u001b[92m\n",
|
||
"\n",
|
||
"Search results: Title: 2025 Australian Open Tennis Schedule - Roadtrips\n",
|
||
"Link: https://www.roadtrips.com/tennis-packages/australian-open/schedule/\n",
|
||
"Snippet: 2025 Australian Open Schedule ; 9, Day, Mon, January 20 ; 9, Night, Mon, January 20 ; 10, Day, Tues, January 21 ; 10, Night, Tues, January 21 ...\n",
|
||
"---\n",
|
||
"Title: 2025 Australian Open Schedule of Play - Grand Slam Tennis Tours\n",
|
||
"Link: https://www.grandslamtennistours.com/australian-open/schedule-of-play\n",
|
||
"Snippet: 2025 Australian Open Schedule of Play ; 22 · 23 ; 5:00 PM · 10:00 AM.\n",
|
||
"---\n",
|
||
"Title: Australian Open 2025 dates announced | AO\n",
|
||
"Link: https://ausopen.com/articles/news/australian-open-2025-dates-announced\n",
|
||
"Snippet: Australian Open 2025 dates are set for 6-26 January at Melbourne Park, guaranteeing fans three weeks of thrilling Grand Slam tennis.\n",
|
||
"---\n",
|
||
"Title: Australian Open 2025 draw: How to watch and follow along | AO\n",
|
||
"Link: https://ausopen.com/articles/news/australian-open-2025-draw-how-watch-and-follow-along\n",
|
||
"Snippet: Thursday 9 January marks the date the Australian Open men's and women's singles draws will be revealed. From 2.30pm AEDT, the draw will be ...\n",
|
||
"---\n",
|
||
"Title: Australian Open 2025: Schedule, format and how to watch\n",
|
||
"Link: https://www.usatoday.com/story/sports/tennis/aus/2025/01/02/australian-open-schedule-format-how-to-watch/77360328007/\n",
|
||
"Snippet: Date: Sunday, Jan. 5 to Saturday, Jan. 25 · TV: ESPN family of networks, Tennis Channel · Streaming: ESPN+, Fubo · Location: Multiple venues ( ...\n",
|
||
"---\n",
|
||
"Title: Official Australian Open 2025 Packages & Tickets | Book Now\n",
|
||
"Link: https://events.com.au/tennis/australian-open-2025/\n",
|
||
"Snippet: Australian Open 2025 will start on Sunday, 12 January 2025 and finish Sunday, 26 January 2025. It's a two week long tournament. Where is Australian ...\n",
|
||
"---\n",
|
||
"Title: 2025 Australian Open: Dates, schedule, draw, and odds\n",
|
||
"Link: https://tenngrand.com/2025-australian-open-dates-schedule-draw-and-odds/\n",
|
||
"Snippet: The Australian Open will take place January 13-26 at Melbourne Park in Melbourne, Australia. Qualifying begins on Monday, January 6. Schedule.\n",
|
||
"---\n",
|
||
"Title: Australian Open 2025 provisional schedule: Plan your three weeks ...\n",
|
||
"Link: https://ausopen.com/articles/news/australian-open-2025-provisional-schedule-plan-your-three-weeks-melbourne-park\n",
|
||
"Snippet: The men's singles semifinals – one beginning in the afternoon, one that same evening – are scheduled at Rod Laver Arena on Friday 24 January, ...\n",
|
||
"---\n",
|
||
"\n",
|
||
"\n",
|
||
"\n",
|
||
"You ONLY have access to the following tools, and should NEVER make up tools that are not listed here:\n",
|
||
"\n",
|
||
"Tool Name: DocumentSearchTool\n",
|
||
"Tool Arguments: {'query': {'description': 'Query to search the document.', 'type': 'str'}}\n",
|
||
"Tool Description: Search the document for the given query.\n",
|
||
"Tool Name: Search the internet\n",
|
||
"Tool Arguments: {'search_query': {'description': 'Mandatory search query you want to use to search the internet', 'type': 'str'}}\n",
|
||
"Tool Description: A tool that can be used to search the internet with a search_query.\n",
|
||
"\n",
|
||
"Use the following format:\n",
|
||
"\n",
|
||
"Thought: you should always think about what to do\n",
|
||
"Action: the action to take, only one name of [DocumentSearchTool, Search the internet], just the name, exactly as it's written.\n",
|
||
"Action Input: the input to the action, just a simple python dictionary, enclosed in curly braces, using \" to wrap keys and values.\n",
|
||
"Observation: the result of the action\n",
|
||
"\n",
|
||
"Once all necessary information is gathered:\n",
|
||
"\n",
|
||
"Thought: I now know the final answer\n",
|
||
"Final Answer: the final answer to the original input question\n",
|
||
"\u001b[00m\n",
|
||
"\n",
|
||
"\n",
|
||
"\u001b[1m\u001b[95m# Agent:\u001b[00m \u001b[1m\u001b[92mRetrieve relevant information to answer the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\u001b[95m## Final Answer:\u001b[00m \u001b[92m\n",
|
||
"The Australian Open 2025 will take place from January 6 to January 26, 2025, at Melbourne Park in Melbourne, Australia.\u001b[00m\n",
|
||
"\n",
|
||
"\n",
|
||
"\u001b[1m\u001b[95m# Agent:\u001b[00m \u001b[1m\u001b[92mResponse synthesizer agent for the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\u001b[95m## Task:\u001b[00m \u001b[92mSynthesize the final response for the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\n",
|
||
"\n",
|
||
"\u001b[1m\u001b[95m# Agent:\u001b[00m \u001b[1m\u001b[92mResponse synthesizer agent for the user query: When is Australian open 2025 happening?\u001b[00m\n",
|
||
"\u001b[95m## Final Answer:\u001b[00m \u001b[92m\n",
|
||
"The Australian Open 2025 will take place from January 6 to January 26, 2025, at Melbourne Park in Melbourne, Australia.\u001b[00m\n",
|
||
"\n",
|
||
"\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"result = crew.kickoff(inputs={\"query\": \"When is Australian open 2025 happening?\"})"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 11,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"The Australian Open 2025 will take place from January 6 to January 26, 2025, at Melbourne Park in Melbourne, Australia.\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"print(result)"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": []
|
||
}
|
||
],
|
||
"metadata": {
|
||
"kernelspec": {
|
||
"display_name": "env_crewai",
|
||
"language": "python",
|
||
"name": "python3"
|
||
},
|
||
"language_info": {
|
||
"codemirror_mode": {
|
||
"name": "ipython",
|
||
"version": 3
|
||
},
|
||
"file_extension": ".py",
|
||
"mimetype": "text/x-python",
|
||
"name": "python",
|
||
"nbconvert_exporter": "python",
|
||
"pygments_lexer": "ipython3",
|
||
"version": "3.10.15"
|
||
}
|
||
},
|
||
"nbformat": 4,
|
||
"nbformat_minor": 2
|
||
}
|