| 54 | currentPage = WikiPage() |
| 55 | |
| 56 | def build_tool(config) -> Tool: |
| 57 | tool = Tool( |
| 58 | "Wikipedia Search", |
| 59 | "Plugin for wikipedia", |
| 60 | name_for_model="Wikipedia", |
| 61 | name_for_human="Wikipedia", |
| 62 | description_for_model="A tool to search entity, view content and disambiguate entity on Wikipedia.\nCurrent endpoint for each function is simple and you should only use exact entity name as input for search and disambiguate. And the keyword input to lookup api should also be simple like one or two words.\nSome Tips to use the APIs bertter:\n1. When the search api doesn't find the corresponding page, you should search a related entity in the return list.\n2. You can only search one entity name in each action, so, don't concat multiple entity names in one search input.\n3. The lookup api can only be used after search api since it depends on the result page of search.\n4. When search api result in an entity page that is not related, you should disambiguate the searched entity to find other entities with the same name.\n5. Don't over rely one this simple tool, you may figure out the next action based on your own knowledge.", |
| 63 | description_for_human="A tool to search entity, view content and disambiguate entity on Wikipedia.\nCurrent endpoint for each function is simple and you should only use exact entity name as input for search and disambiguate. And the keyword input to lookup api should also be simple like one or two words.", |
| 64 | logo_url="https://your-app-url.com/.well-known/logo.png", |
| 65 | contact_email="hello@contact.com", |
| 66 | legal_info_url="hello@legal.com" |
| 67 | ) |
| 68 | |
| 69 | @tool.get("/search") |
| 70 | def search(entity : str, request : Request): |
| 71 | '''The input is an exact entity name. The action will search this entity name on Wikipedia and returns the first five sentences if it exists. If not, it will return some related entities to search next. |
| 72 | ''' |
| 73 | |
| 74 | entity_ = entity.replace(" ", "+") |
| 75 | search_url = f"https://en.wikipedia.org/w/index.php?search={entity_}" |
| 76 | response_text = requests.get(search_url).text |
| 77 | soup = BeautifulSoup(response_text, features="html.parser") |
| 78 | result_divs = soup.find_all("div", {"class": "mw-search-result-heading"}) |
| 79 | if result_divs: # mismatch |
| 80 | result_titles = [clean_str(div.get_text().strip()) for div in result_divs] |
| 81 | obs = f"Could not find {entity}. Similar: {result_titles[:5]}." |
| 82 | else: |
| 83 | local_page = [p.get_text().strip() for p in soup.find_all("p") + soup.find_all("ul")] |
| 84 | if any("may refer to:" in p for p in local_page): |
| 85 | obs = search("[" + entity + "]", request) |
| 86 | else: |
| 87 | currentPage.reset_page() |
| 88 | page = "" |
| 89 | for p in local_page: |
| 90 | if len(p.split(" ")) > 2: |
| 91 | page += clean_str(p) |
| 92 | if not p.endswith("\n"): |
| 93 | page += "\n" |
| 94 | obs = currentPage.get_page_obs(page) |
| 95 | return obs |
| 96 | |
| 97 | @tool.get("/lookup") |
| 98 | def lookup(keyword : str, request : Request) -> str: |
| 99 | '''The input is a keyword. This action will look up in the current passage and return the next several sentences containing the keyword in current passage. |
| 100 | ''' |
| 101 | # lookup_keyword = request.session["lookup_keyword"] |
| 102 | # lookup_list = request.session["lookup_list"] |
| 103 | # lookup_cnt = request.session["lookup_cnt"] |
| 104 | # sentences = request.session["sentences"] |
| 105 | lookup_keyword = currentPage.lookup_keyword |
| 106 | if lookup_keyword != keyword: # reset lookup |
| 107 | currentPage.construct_lookup_list(keyword) |
| 108 | lookup_list = currentPage.lookup_list |
| 109 | lookup_cnt = currentPage.lookup_cnt |
| 110 | sentences = currentPage.sentences |
| 111 | if lookup_cnt >= len(lookup_list): |
| 112 | obs = "No more results." |
| 113 | else: |