tools.py

from smolagents import DuckDuckGoSearchTool
from smolagents import Tool
from huggingface_hub import InferenceClient
import soundfile as sf
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset
from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info

class Web_research(Tool):
    name="web_research"
    description = "Web search on a specific topic."
    inputs = {
        "topic": {
            "type": "string",
            "description": "The topic on which the user wants the latest news"
        }
    }
    output_type = "string"

    def forward(self, topic: str):
        search_tool = DuckDuckGoSearchTool()
        # Example usage
        results = search_tool(f"{topic}")
        return f"Here is what we can find on the web for {topic} : str({results})"

class Find_wikipedia_URL(Tool):
    name="wiki_url"
    description = "Always use to check a wikipedia ENGLISH URL page before trying to acces the URL. For another langage, you just have to change the beginning of the url (here, it is en for english)"
    inputs = {
        "subject": {
            "type": "string",
            "description": "The name or topic on which you want the Wikipedia URL"
        }
    }
    output_type = "string"

    def forward(self, subject: str):
        words=subject.split()
        url_wiki="https://en.wikipedia.org/wiki/"
        for i in range(len(words)):
            if(i==0):
                url_wiki+=str(words[i])
            if(i!=0):
                url_wiki+='_'+str(words[i])
        return f"Here is what we url to use : str({url_wiki}). If it does not work, change the first letters of {subject} to be upper or lower, but never change anything else"

class translate_everything(Tool):
    name="translator"
    description = "You do not understand a sentence? It does not look like any language you know? Try this tool, maybe the sentence is just reversed!"
    inputs = {
        "sentence": {
            "type": "string",
            "description": "The sentence to translate"
        }
    }
    output_type = "string"

    def forward(self, sentence: str):
        # Input string
        reversed_words = sentence.split() #' '.join(s.split()[::-1])
        right_sentence=[]
        for word in reversed_words:
            right_sentence.append(word[::-1])
        
        translated_sentence = " ".join(right_sentence[::-1])
        return f"The translated sentence is : {translated_sentence}"

class multimodal_interpreter(Tool):
    name="multimodal_tool"
    description = "Allows you to answer any question which relies on image or video input."
    inputs = {
        'image': {"type": "image", "description": "the image or video of interest"},
        'prompt': {"type": "string", "description": "Any specific question you have on the image. For example, the prompt can be : Describe this image."}
    }
    output_type = "string"
    
    def forward(self, prompt, image):
        
        # default: Load the model on the available device(s)
        model = Qwen2VLForConditionalGeneration.from_pretrained(
            "Qwen/Qwen2-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
        )
        
        # We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
        # model = Qwen2VLForConditionalGeneration.from_pretrained(
        #     "Qwen/Qwen2-VL-7B-Instruct",
        #     torch_dtype=torch.bfloat16,
        #     attn_implementation="flash_attention_2",
        #     device_map="auto",
        # )
        
        # default processer
        processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
        
        # The default range for the number of visual tokens per image in the model is 4-16384. You can set min_pixels and max_pixels according to your needs, such as a token count range of 256-1280, to balance speed and memory usage.
        # min_pixels = 256*28*28
        # max_pixels = 1280*28*28
        # processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)
        
        messages = [
            {
                "role": "user",
                "content": [
                    {
                        "type": "image",
                        "image": {image},
                    },
                    {"type": "text", "text": {prompt}},
                ],
            }
        ]
        
        # Preparation for inference
        text = processor.apply_chat_template(
            messages, tokenize=False, add_generation_prompt=True
        )
        image_inputs, video_inputs = process_vision_info(messages)
        inputs = processor(
            text=[text],
            images=image_inputs,
            videos=video_inputs,
            padding=True,
            return_tensors="pt",
        )
        inputs = inputs.to("cuda")
        
        # Inference: Generation of the output
        generated_ids = model.generate(**inputs, max_new_tokens=128)
        generated_ids_trimmed = [
            out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
        ]
        output_text = processor.batch_decode(
            generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
        )
        return output_text

class audio_or_mp3__interpreter(Tool):
    name="multimodal_tool"
    description = "Allows you to convert audio into text. It uses Whisper, it is a state-of-the-art model for automatic speech recognition (ASR) and speech translation"
    inputs = {
        'audio': {"type": "audio", "description": "the audio of interest"}
    }
    output_type = "string"
    
    def forward(self, prompt, audio):
        device = "cuda:0" if torch.cuda.is_available() else "cpu"
        torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
        
        model_id = "openai/whisper-large-v3"
        
        model = AutoModelForSpeechSeq2Seq.from_pretrained(
            model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
        )
        model.to(device)
        
        processor = AutoProcessor.from_pretrained(model_id)
        
        pipe = pipeline(
            "automatic-speech-recognition",
            model=model,
            tokenizer=processor.tokenizer,
            feature_extractor=processor.feature_extractor,
            torch_dtype=torch_dtype,
            device=device,
        )
        
        sample = {audio} #sample must be of the type dataset[0]["audio"]
        
        result = pipe(sample)
        return result["text"]

class Wikipedia_reader(Tool):
    name="wikipedia_tool"
    description = "To be used whenever you need to read a Wikipedia page. Will return all the text of the Wikipedia page, to easily read it and find information"
    inputs = {
        "url": {
            "type": "string",
            "description": "The wikippedia url page"
        }
    }
    output_type = "string"

    def forward(self, url: str):
        try:
            page = requests.get(url)
        except Exception as e:
            print('Error downloading page: ',e)
        soup = BeautifulSoup(page.text, 'html.parser')
        return soup.text