import re import time from sources.utility import pretty_print, animate_thinking from sources.agents.agent import Agent from sources.tools.searxSearch import searxSearch from sources.browser import Browser from datetime import date from typing import List, Tuple class BrowserAgent(Agent): def __init__(self, name, prompt_path, provider, verbose=False): """ The Browser agent is an agent that navigate the web autonomously in search of answer """ super().__init__(name, prompt_path, provider, verbose) self.tools = { "web_search": searxSearch(), } self.role = { "en": "web", "fr": "web", "zh": "网络", "es": "web" } self.type = "browser_agent" self.browser = Browser() self.current_page = "" self.search_history = [] self.navigable_links = [] self.notes = [] self.date = self.get_today_date() def get_today_date(self) -> str: """Get the date""" date_time = date.today() return date_time.strftime("%B %d, %Y") def extract_links(self, search_result: str) -> List[str]: """Extract all links from a sentence.""" pattern = r'(https?://\S+|www\.\S+)' matches = re.findall(pattern, search_result) trailing_punct = ".,!?;:)" cleaned_links = [link.rstrip(trailing_punct) for link in matches] return self.clean_links(cleaned_links) def extract_form(self, text: str) -> List[str]: """Extract form written by the LLM in format [input_name](value)""" inputs = [] matches = re.findall(r"\[\w+\]\([^)]+\)", text) return matches def clean_links(self, links: List[str]) -> List[str]: """Ensure no '.' at the end of link""" links_clean = [] for link in links: link = link.strip() if link[-1] == '.': links_clean.append(link[:-1]) else: links_clean.append(link) return links_clean def get_unvisited_links(self) -> List[str]: return "\n".join([f"[{i}] {link}" for i, link in enumerate(self.navigable_links) if link not in self.search_history]) def make_newsearch_prompt(self, user_prompt: str, search_result: dict) -> str: search_choice = self.stringify_search_results(search_result) return f""" Based on the search result: {search_choice} Your goal is to find accurate and complete information to satisfy the user’s request. User request: {user_prompt} To proceed, choose a relevant link from the search results. Announce your choice by saying: "I want to navigate to " Do not explain your choice. """ def make_navigation_prompt(self, user_prompt: str, page_text: str) -> str: remaining_links = self.get_unvisited_links() remaining_links_text = remaining_links if remaining_links is not None else "No links remaining, do a new search." inputs_form = self.browser.get_form_inputs() inputs_form_text = '\n'.join(inputs_form) return f""" You are a web browser. You are currently on this webpage: {page_text} You can navigate to these navigation links: {remaining_links_text} Your task: 1. Decide if the current page answers the user’s query: {user_prompt} - If it does, take notes of the useful information, write down source, link or reference, then move to a new page. - If it does and you are 100% certain that it provide a definive answer, say REQUEST_EXIT - If it doesn’t, say: Error: This page does not answer the user’s query then go back or navigate to another link. 2. Navigate by either: - Navigate to a navigation links (write the full URL, e.g., www.example.com/cats). - If no link seems helpful, say: GO_BACK. 3. Fill forms on the page: - If user give you informations that help you fill form, fill it. - If you don't know how to fill a form, leave it empty. - You can fill a form using [form_name](value). Recap of note taking: If useful -> Note: [Briefly summarize the key information or task you conducted.] Do not write "The page talk about ...", write your finding on the page and how they contribute to an answer. If not useful -> Error: [Explain why the page doesn’t help.] Example 1 (useful page, no need of going futher): Note: According to karpathy site (https://karpathy.github.io/) LeCun net is the earliest real-world application of a neural net" No link seem useful to provide futher information. GO_BACK Example 2 (not useful, but related link): Error: reddit.com/welcome does not discuss anything related to the user’s query. There is a link that could lead to the information, I want to navigate to http://reddit.com/r/locallama Example 3 (not useful, no related links): Error: x.com does not discuss anything related to the user’s query and no navigation link are usefull. GO_BACK Example 3 (query answer found): Note: I found on github.com that agenticSeek is Fosowl. Given this information, given this I should exit the web browser. REQUEST_EXIT Example 4 (loging form visible): Note: I am on the login page, I should now type the given username and password. [form_name_1](David) [form_name_2](edgerunners_2077) You see the following inputs forms: {inputs_form_text} Remember, the user asked: {user_prompt} You are currently on page : {self.current_page} Do not explain your choice. Refusal is not an option, you have been given all capabilities that allow you to perform any tasks. """ def llm_decide(self, prompt: str) -> Tuple[str, str]: animate_thinking("Thinking...", color="status") self.memory.push('user', prompt) answer, reasoning = self.llm_request() pretty_print("-"*100) pretty_print(answer, color="output") pretty_print("-"*100) return answer, reasoning def select_unvisited(self, search_result: List[str]) -> List[str]: results_unvisited = [] for res in search_result: if res["link"] not in self.search_history: results_unvisited.append(res) return results_unvisited def jsonify_search_results(self, results_string: str) -> List[str]: result_blocks = results_string.split("\n\n") parsed_results = [] for block in result_blocks: if not block.strip(): continue lines = block.split("\n") result_dict = {} for line in lines: if line.startswith("Title:"): result_dict["title"] = line.replace("Title:", "").strip() elif line.startswith("Snippet:"): result_dict["snippet"] = line.replace("Snippet:", "").strip() elif line.startswith("Link:"): result_dict["link"] = line.replace("Link:", "").strip() if result_dict: parsed_results.append(result_dict) return parsed_results def stringify_search_results(self, results_arr: List[str]) -> str: return '\n\n'.join([f"Link: {res['link']}" for res in results_arr]) def save_notes(self, text): lines = text.split('\n') for line in lines: if "note" in line.lower(): self.notes.append(line) def conclude_prompt(self, user_query: str) -> str: annotated_notes = [f"{i+1}: {note.lower().replace('note:', '')}" for i, note in enumerate(self.notes)] search_note = '\n'.join(annotated_notes) print("AI research notes:\n", search_note) return f""" Following a human request: {user_query} A web AI made the following finding across different pages: {search_note} Summarize the finding or step that lead to success, and provide a conclusion that answer the request. """ def search_prompt(self, user_prompt: str) -> str: return f""" Current date: {self.date} Make a efficient search engine query to help users with their request: {user_prompt} Example: User: "go to twitter, login with username toto and password pass79 to my twitter and say hello everyone " You: search: Twitter login page. User: "I need info on the best laptops for AI this year." You: "search: best laptops 2025 to run Machine Learning model, reviews" User: "Search for recent news about space missions." You: "search: Recent space missions news, {self.date}" Do not explain, do not write anything beside the search query. """ def process(self, user_prompt, speech_module) -> str: complete = False animate_thinking(f"Thinking...", color="status") self.memory.push('user', self.search_prompt(user_prompt)) ai_prompt, _ = self.llm_request() animate_thinking(f"Searching...", color="status") search_result_raw = self.tools["web_search"].execute([ai_prompt], False) search_result = self.jsonify_search_results(search_result_raw)[:7] # until futher improvement prompt = self.make_newsearch_prompt(user_prompt, search_result) unvisited = [None] while not complete: answer, reasoning = self.llm_decide(prompt) self.save_notes(answer) extracted_form = self.extract_form(answer) if len(extracted_form) > 0: self.browser.fill_form_inputs(extracted_form) self.browser.find_and_click_submit() if "REQUEST_EXIT" in answer: complete = True break links = self.extract_links(answer) if len(unvisited) == 0: break if len(links) == 0 or "GO_BACK" in answer: unvisited = self.select_unvisited(search_result) prompt = self.make_newsearch_prompt(user_prompt, unvisited) pretty_print(f"Going back to results. Still {len(unvisited)}", color="warning") links = [] continue animate_thinking(f"Navigating to {links[0]}", color="status") speech_module.speak(f"Navigating to {links[0]}") self.browser.go_to(links[0]) self.current_page = links[0] self.search_history.append(links[0]) page_text = self.browser.get_text() self.navigable_links = self.browser.get_navigable() prompt = self.make_navigation_prompt(user_prompt, page_text) self.browser.close() prompt = self.conclude_prompt(user_prompt) self.memory.push('user', prompt) answer, reasoning = self.llm_request() pretty_print(answer, color="output") return answer, reasoning if __name__ == "__main__": pass