Spaces:
Sleeping
Sleeping
| import os | |
| import re | |
| from pathlib import Path | |
| import tempfile | |
| import openai | |
| import pandas as pd | |
| from tabulate import tabulate | |
| from smolagents.tools import Tool, PipelineTool | |
| # Load your OpenAI API key | |
| openai.api_key = os.getenv("OPENAI_API_KEY") | |
| SYSTEM_PROMPT = """You are a general AI assistant. I will ask you a question. Report your thoughts, and finish your answer with the following template: | |
| FINAL ANSWER: [YOUR FINAL ANSWER]. | |
| If you have enough confidence (≥70%) in your internal knowledge, answer directly without calling any other tool. | |
| Otherwise, decide which tool (wiki, web search, python, audio, excel) is best and use it. | |
| YOUR FINAL ANSWER should be: | |
| - a number OR as few words as possible OR a comma separated list of numbers and/or strings. | |
| - If you are asked for a number, don't use commas, symbols or units (e.g. %, $, km) unless explicitly asked. | |
| - If you are asked for a string, don't use articles ("a", "the"), abbreviations (e.g. "NYC"), or extra words; write digits in plain text unless specified otherwise. | |
| - If you are asked for a comma separated list, apply the above rules to each element. | |
| """ | |
| class AnswerTool(Tool): | |
| name = "answer_tool" | |
| description = "Answer GAIA Level 1 questions in exact-match format." | |
| inputs = {"question": {"type": "string", "description": "The GAIA question text."}} | |
| output_type = "string" | |
| def forward(self, question: str) -> str: | |
| messages = [ | |
| {"role": "system", "content": SYSTEM_PROMPT}, | |
| {"role": "user", "content": question}, | |
| ] | |
| resp = openai.chat.completions.create( | |
| model="gpt-4o", | |
| messages=messages, | |
| temperature=0.0, | |
| max_tokens=64, | |
| ) | |
| text = resp.choices[0].message.content.strip() | |
| if "FINAL ANSWER:" in text: | |
| return text.split("FINAL ANSWER:")[-1].strip() | |
| return text | |
| class SpeechToTextTool(PipelineTool): | |
| """Transcribes audio files using OpenAI Whisper.""" | |
| name = "transcriber" | |
| description = "Transcribes a local audio file to text using Whisper API." | |
| inputs = {"audio": {"type": "string", "description": "Path to audio file."}} | |
| output_type = "string" | |
| default_checkpoint = "openai/whisper-1" | |
| def __call__(self, audio: str) -> str: | |
| return self._transcribe(audio) | |
| def _transcribe(audio_path: str) -> str: | |
| path = Path(audio_path).expanduser().resolve() | |
| if not path.is_file(): | |
| raise FileNotFoundError(f"Audio file not found: {path}") | |
| with path.open("rb") as fp: | |
| response = openai.audio.transcriptions.create( | |
| file=fp, | |
| model="whisper-1", | |
| response_format="text" | |
| ) | |
| return response | |
| class ExcelToTextTool(Tool): | |
| """Renders an Excel worksheet as Markdown text.""" | |
| name = "excel_to_text" | |
| description = "Read an Excel file and return a Markdown table." | |
| inputs = { | |
| "excel_path": {"type": "string", "description": "Path to Excel file."}, | |
| "sheet_name": {"type": "string", "description": "Sheet name or index as string.", "nullable": True}, | |
| } | |
| output_type = "string" | |
| def forward(self, excel_path: str, sheet_name: str = None) -> str: | |
| path = Path(excel_path).expanduser().resolve() | |
| if not path.exists(): | |
| return f"Error: Excel file not found at {path}" | |
| sheet = 0 if not sheet_name or sheet_name == "" else ( | |
| int(sheet_name) if sheet_name.isdigit() else sheet_name | |
| ) | |
| df = pd.read_excel(path, sheet_name=sheet) | |
| if hasattr(df, "to_markdown"): | |
| return df.to_markdown(index=False) | |
| return tabulate(df, headers="keys", tablefmt="github", showindex=False) | |