import os import gradio as gr # Get Hugging Face API Key huggingface_token = os.environ.get('HUGGINGFACE_API_KEY') # PubMed Fetcher from pymed import PubMed from typing import List from haystack import component from haystack import Document def predict( message, history): pubmed = PubMed(tool="Tydei", email="info@tydei.io") def documentize(article): return Document(content=article.abstract, meta={'title': article.title, 'keywords': article.keywords, 'journal': article.journal}) @component class PubMedFetcher(): @component.output_types(articles=List[Document]) def run(self, queries: list[str]): cleaned_queries = queries[0].strip().split('\n') articles = [] try: for query in cleaned_queries: response = pubmed.query(query, max_results=3) documents = [documentize(article) for article in response] articles.extend(documents) except Exception as e: print(e) print(f"Couldn't fetch articles for queries: {queries}") results = {'articles': articles} print("-----------------------------") print ("results : ", results) # Initialize an empty list to store all keywords all_keywords = [] # Iterate over each Document in the articles list for document in articles: # Extract the keywords from the meta attribute of each Document keywords = document.meta.get('keywords', []) # Extend the all_keywords list with the extracted keywords all_keywords.extend(keywords) # Add the extracted keywords to the results dictionary results['keywords'] = all_keywords # Print all extracted keywords print("Extracted Keywords:", all_keywords) return results # Pipeline Setup from haystack.components.generators import HuggingFaceTGIGenerator from haystack import Pipeline from haystack.components.builders.prompt_builder import PromptBuilder keyword_llm = HuggingFaceTGIGenerator("mistralai/Mixtral-8x7B-Instruct-v0.1", token=huggingface_token) keyword_llm.warm_up() llm = HuggingFaceTGIGenerator("mistralai/Mixtral-8x7B-Instruct-v0.1", token=huggingface_token) llm.warm_up() from haystack import Pipeline from haystack.components.builders.prompt_builder import PromptBuilder keyword_prompt_template = """ Your task is to convert the follwing question into 5 keywords that can be used to find relevant medical research papers on PubMed. --- question: {{ question }} keywords: """ prompt_template = """ Answer the question truthfully based on the given documents. Only use data from the given documents to answer the question. q: {{ question }} Articles: {% for article in articles %} {{article.content}} keywords: {{article.meta['keywords']}} title: {{article.meta['title']}} journal: {{article.journal}} {% endfor %} Only provide an answer based on the information give. Do not make anything up. At the end of your answer create a Source List with the name of Article title and Journal. Retrieve your information from the article meta data DO not repeat yourself in the answer and ONLY use information from the given articles. DO NOT USE Your Own Knowledge base or you will be punished """ keyword_prompt_builder = PromptBuilder(template=keyword_prompt_template) prompt_builder = PromptBuilder(template=prompt_template) fetcher = PubMedFetcher() pipe = Pipeline() pipe.add_component("keyword_prompt_builder", keyword_prompt_builder) pipe.add_component("keyword_llm", keyword_llm) pipe.add_component("pubmed_fetcher", fetcher) pipe.add_component("prompt_builder", prompt_builder) pipe.add_component("llm", llm) pipe.connect("keyword_prompt_builder.prompt", "keyword_llm.prompt") pipe.connect("keyword_llm.replies", "pubmed_fetcher.queries") pipe.connect("pubmed_fetcher.articles", "prompt_builder.articles") pipe.connect("prompt_builder.prompt", "llm.prompt") # Function to Ask Questions def ask(question): result = pipe.run(data={"keyword_prompt_builder":{"question":question}, "prompt_builder":{"question": question}, "llm":{"generation_kwargs": {"max_new_tokens": 1500}}}) print("-----------------------------") print("Question is: ", question) print("-----------------------------") print(result) output = result['llm']['replies'][0] keywords = result['pubmed_fetcher']['keywords'] return keywords, output keywords , output= ask(message) return keywords, output # Launch Gradio Interface interface = gr.Interface( fn=predict, inputs= [gr.Textbox(label="Research Pubmed...", lines=2, placeholder="Type Question here..."), gr.Dropdown(["ortho", "anthroscopy", "osteotomy", "fracture repair"], value=["spinal fusion", "bone grafting"], multiselect=True, label="Keyword Entry")], outputs=[gr.Textbox(label="Keywords used in researching articles", lines=5), gr.Textbox(label="Pubmed Insights", lines=20)], allow_flagging="never" ) interface.queue() interface.launch()