|
from smolagents import CodeAgent,DuckDuckGoSearchTool, HfApiModel,load_tool,tool |
|
import datetime |
|
import requests |
|
import pytz |
|
import yaml |
|
from tools.final_answer import FinalAnswerTool |
|
import pandas as pd |
|
from huggingface_hub import InferenceClient |
|
|
|
|
|
from Gradio_UI import GradioUI |
|
|
|
final_answer = FinalAnswerTool() |
|
|
|
|
|
|
|
|
|
model = HfApiModel( |
|
max_tokens=2096, |
|
temperature=0.5, |
|
model_id='Qwen/Qwen2.5-Coder-32B-Instruct', |
|
custom_role_conversions=None, |
|
) |
|
|
|
@tool |
|
def f1_tackinfo_getter(country: str)-> str: |
|
""" |
|
Returns data for a specified race |
|
Args: |
|
country: A string respresenting a valid country name from the 2024 F1 calendar. |
|
Returns: |
|
A string with information about the given race |
|
""" |
|
df = pd.read_csv('./Formula1_2024season_raceResults.csv') |
|
|
|
df = df[['Track', 'Position', 'Driver', 'Team']] |
|
info = str(df.groupby('Track').get_group(country).iloc[:10]) |
|
client = InferenceClient("meta-llama/Llama-3.2-3B-Instruct") |
|
system_prompt = "You are an expert in F1 race analysis. You will be given data about a race and your goal is to provide a very concise analysis of these results. Your answer should be no more that 5 sentences long. Your answer should only be based on the provided information, do not add anything that is not in the data" |
|
output = client.chat.completions.create( |
|
messages = [ |
|
{'role': 'system', 'content': system_prompt}, |
|
{'role': 'user', 'content': f'Here is the data about the race: {info}'} |
|
] |
|
) |
|
return output.choices[0].message.content |
|
|
|
with open("prompts.yaml", 'r') as stream: |
|
prompt_templates = yaml.safe_load(stream) |
|
|
|
|
|
agent = CodeAgent( |
|
model=model, |
|
tools=[f1_tackinfo_getter, final_answer], |
|
max_steps=6, |
|
verbosity_level=1, |
|
grammar=None, |
|
planning_interval=None, |
|
name='PitBot', |
|
description="A Formula 1 race analysis assistant that provides detailed insights about the 2024 F1 season races.", |
|
prompt_templates=prompt_templates |
|
) |
|
|
|
|
|
GradioUI(agent).launch() |