import logging import warnings from concurrent.futures import ThreadPoolExecutor from datetime import datetime, timezone from decimal import Decimal from functools import cached_property from itertools import cycle, islice from random import choice from threading import Event from types import TracebackType from typing import Dict, List, Optional, Tuple, Type, Union, cast import asyncio import json import aiohttp import pyreqwest_impersonate as pri try: from lxml.etree import _Element from lxml.html import HTMLParser as LHTMLParser from lxml.html import document_fromstring LXML_AVAILABLE = True except ImportError: LXML_AVAILABLE = False import re from decimal import Decimal from html import unescape from math import atan2, cos, radians, sin, sqrt from typing import Any, Dict, List, Union from urllib.parse import unquote try: HAS_ORJSON = True import orjson except ImportError: HAS_ORJSON = False import json REGEX_STRIP_TAGS = re.compile("<.*?>") def json_dumps(obj: Any) -> str: try: return ( orjson.dumps(obj, option=orjson.OPT_INDENT_2).decode() if HAS_ORJSON else json.dumps(obj, ensure_ascii=False, indent=2) ) except Exception as ex: raise WebscoutE(f"{type(ex).__name__}: {ex}") from ex def json_loads(obj: Union[str, bytes]) -> Any: try: return orjson.loads(obj) if HAS_ORJSON else json.loads(obj) except Exception as ex: raise WebscoutE(f"{type(ex).__name__}: {ex}") from ex def _extract_vqd(html_bytes: bytes, keywords: str) -> str: """Extract vqd from html bytes.""" for c1, c1_len, c2 in ( (b'vqd="', 5, b'"'), (b"vqd=", 4, b"&"), (b"vqd='", 5, b"'"), ): try: start = html_bytes.index(c1) + c1_len end = html_bytes.index(c2, start) return html_bytes[start:end].decode() except ValueError: pass raise WebscoutE(f"_extract_vqd() {keywords=} Could not extract vqd.") def _text_extract_json(html_bytes: bytes, keywords: str) -> List[Dict[str, str]]: """text(backend="api") -> extract json from html.""" try: start = html_bytes.index(b"DDG.pageLayout.load('d',") + 24 end = html_bytes.index(b");DDG.duckbar.load(", start) data = html_bytes[start:end] result: List[Dict[str, str]] = json_loads(data) return result except Exception as ex: raise WebscoutE(f"_text_extract_json() {keywords=} {type(ex).__name__}: {ex}") from ex raise WebscoutE(f"_text_extract_json() {keywords=} return None") def _normalize(raw_html: str) -> str: """Strip HTML tags from the raw_html string.""" return unescape(REGEX_STRIP_TAGS.sub("", raw_html)) if raw_html else "" def _normalize_url(url: str) -> str: """Unquote URL and replace spaces with '+'.""" return unquote(url.replace(" ", "+")) if url else "" def _calculate_distance(lat1: Decimal, lon1: Decimal, lat2: Decimal, lon2: Decimal) -> float: """Calculate distance between two points in km. Haversine formula.""" R = 6371.0087714 # Earth's radius in km rlat1, rlon1, rlat2, rlon2 = map(radians, [float(lat1), float(lon1), float(lat2), float(lon2)]) dlon, dlat = rlon2 - rlon1, rlat2 - rlat1 a = sin(dlat / 2) ** 2 + cos(rlat1) * cos(rlat2) * sin(dlon / 2) ** 2 c = 2 * atan2(sqrt(a), sqrt(1 - a)) return R * c class WebscoutE(Exception): """Base exception class for search.""" class RatelimitE(Exception): """Raised for rate limit exceeded errors during API requests.""" class TimeoutE(Exception): """Raised for timeout errors during API requests.""" class FailedToGenerateResponseError(Exception): """Provider failed to fetch response""" class AllProvidersFailure(Exception): """None of the providers generated response successfully""" pass logger = logging.getLogger("webscout.WEBS") class WEBS: """webscout class to get search results from duckduckgo.com.""" _executor: ThreadPoolExecutor = ThreadPoolExecutor() _impersonates = ( "chrome_99", "chrome_100", "chrome_101", "chrome_104", "chrome_105", "chrome_106", "chrome_108", "chrome_107", "chrome_109", "chrome_114", "chrome_116", "chrome_117", "chrome_118", "chrome_119", "chrome_120", #"chrome_123", "chrome_124", "chrome_126", "safari_ios_16.5", "safari_ios_17.2", "safari_ios_17.4.1", "safari_15.3", "safari_15.5", "safari_15.6.1", "safari_16", "safari_16.5", "safari_17.2.1", "safari_17.4.1", "safari_17.5", #"okhttp_3.9", "okhttp_3.11", "okhttp_3.13", "okhttp_3.14", "okhttp_4.9", "okhttp_4.10", "okhttp_5", "edge_99", "edge_101", "edge_122", ) # fmt: skip def __init__( self, headers: Optional[Dict[str, str]] = None, proxy: Optional[str] = None, proxies: Union[Dict[str, str], str, None] = None, # deprecated timeout: Optional[int] = 10, ) -> None: """Initialize the WEBS object. Args: headers (dict, optional): Dictionary of headers for the HTTP client. Defaults to None. proxy (str, optional): proxy for the HTTP client, supports http/https/socks5 protocols. example: "http://user:pass@example.com:3128". Defaults to None. timeout (int, optional): Timeout value for the HTTP client. Defaults to 10. """ self.proxy: Optional[str] = proxy assert self.proxy is None or isinstance(self.proxy, str), "proxy must be a str" if not proxy and proxies: warnings.warn("'proxies' is deprecated, use 'proxy' instead.", stacklevel=1) self.proxy = proxies.get("http") or proxies.get("https") if isinstance(proxies, dict) else proxies self.headers = headers if headers else {} self.headers["Referer"] = "https://duckduckgo.com/" self.client = pri.Client( headers=self.headers, proxy=self.proxy, timeout=timeout, cookie_store=True, referer=True, impersonate=choice(self._impersonates), follow_redirects=False, verify=False, ) self._exception_event = Event() self._chat_messages: List[Dict[str, str]] = [] self._chat_vqd: str = "" def __enter__(self) -> "WEBS": return self def __exit__( self, exc_type: Optional[Type[BaseException]] = None, exc_val: Optional[BaseException] = None, exc_tb: Optional[TracebackType] = None, ) -> None: pass @cached_property def parser(self) -> "LHTMLParser": """Get HTML parser.""" return LHTMLParser(remove_blank_text=True, remove_comments=True, remove_pis=True, collect_ids=False) def _get_url( self, method: str, url: str, params: Optional[Dict[str, str]] = None, content: Optional[bytes] = None, data: Optional[Union[Dict[str, str], bytes]] = None, ) -> bytes: if self._exception_event.is_set(): raise WebscoutE("Exception occurred in previous call.") try: resp = self.client.request(method, url, params=params, content=content, data=data) except Exception as ex: self._exception_event.set() if "time" in str(ex).lower(): raise TimeoutE(f"{url} {type(ex).__name__}: {ex}") from ex raise WebscoutE(f"{url} {type(ex).__name__}: {ex}") from ex logger.debug(f"_get_url() {resp.url} {resp.status_code} {len(resp.content)}") if resp.status_code == 200: return cast(bytes, resp.content) self._exception_event.set() if resp.status_code in (202, 301, 403): raise RatelimitE(f"{resp.url} {resp.status_code} Ratelimit") raise WebscoutE(f"{resp.url} return None. {params=} {content=} {data=}") def _get_vqd(self, keywords: str) -> str: """Get vqd value for a search query.""" resp_content = self._get_url("POST", "https://duckduckgo.com", data={"q": keywords}) return _extract_vqd(resp_content, keywords) def chat(self, keywords: str, model: str = "gpt-3.5", timeout: int = 20) -> str: """Initiates a chat session with DuckDuckGo AI. Args: keywords (str): The initial message or question to send to the AI. model (str): The model to use: "gpt-3.5", "claude-3-haiku", "llama-3-70b", "mixtral-8x7b". Defaults to "gpt-3.5". timeout (int): Timeout value for the HTTP client. Defaults to 20. Returns: str: The response from the AI. """ models = { "claude-3-haiku": "claude-3-haiku-20240307", "gpt-3.5": "gpt-3.5-turbo-0125", "llama-3-70b": "meta-llama/Llama-3-70b-chat-hf", "mixtral-8x7b": "mistralai/Mixtral-8x7B-Instruct-v0.1", "gpt-4o-mini": "gpt-4o-mini", } # vqd if not self._chat_vqd: resp = self.client.get("https://duckduckgo.com/duckchat/v1/status", headers={"x-vqd-accept": "1"}) self._chat_vqd = resp.headers.get("x-vqd-4", "") self._chat_messages.append({"role": "user", "content": keywords}) json_data = { "model": models[model], "messages": self._chat_messages, } resp = self.client.post( "https://duckduckgo.com/duckchat/v1/chat", headers={"x-vqd-4": self._chat_vqd}, json=json_data, timeout=timeout, ) self._chat_vqd = resp.headers.get("x-vqd-4", "") data = ",".join(x for line in resp.text.rstrip("[DONE]\n").split("data:") if (x := line.strip())) result = "".join(x.get("message", "") for x in json_loads("[" + data + "]")) self._chat_messages.append({"role": "assistant", "content": result}) return result def text( self, keywords: str, region: str = "wt-wt", safesearch: str = "moderate", timelimit: Optional[str] = None, backend: str = "api", max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo text search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". safesearch: on, moderate, off. Defaults to "moderate". timelimit: d, w, m, y. Defaults to None. backend: api, html, lite. Defaults to api. api - collect data from https://duckduckgo.com, html - collect data from https://html.duckduckgo.com, lite - collect data from https://lite.duckduckgo.com. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with search results, or None if there was an error. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ if LXML_AVAILABLE is False and backend != "api": backend = "api" warnings.warn("lxml is not installed. Using backend='api'.", stacklevel=2) if backend == "api": results = self._text_api(keywords, region, safesearch, timelimit, max_results) elif backend == "html": results = self._text_html(keywords, region, timelimit, max_results) elif backend == "lite": results = self._text_lite(keywords, region, timelimit, max_results) return results def _text_api( self, keywords: str, region: str = "wt-wt", safesearch: str = "moderate", timelimit: Optional[str] = None, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo text search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". safesearch: on, moderate, off. Defaults to "moderate". timelimit: d, w, m, y. Defaults to None. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with search results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" vqd = self._get_vqd(keywords) payload = { "q": keywords, "kl": region, "l": region, "p": "", "s": "0", "df": "", "vqd": vqd, "bing_market": f"{region[3:]}-{region[:2].upper()}", "ex": "", } safesearch = safesearch.lower() if safesearch == "moderate": payload["ex"] = "-1" elif safesearch == "off": payload["ex"] = "-2" elif safesearch == "on": # strict payload["p"] = "1" if timelimit: payload["df"] = timelimit cache = set() results: List[Dict[str, str]] = [] def _text_api_page(s: int) -> List[Dict[str, str]]: payload["s"] = f"{s}" resp_content = self._get_url("GET", "https://links.duckduckgo.com/d.js", params=payload) page_data = _text_extract_json(resp_content, keywords) page_results = [] for row in page_data: href = row.get("u", None) if href and href not in cache and href != f"http://www.google.com/search?q={keywords}": cache.add(href) body = _normalize(row["a"]) if body: result = { "title": _normalize(row["t"]), "href": _normalize_url(href), "body": body, } page_results.append(result) return page_results slist = [0] if max_results: max_results = min(max_results, 2023) slist.extend(range(23, max_results, 50)) try: for r in self._executor.map(_text_api_page, slist): results.extend(r) except Exception as e: raise e return list(islice(results, max_results)) def _text_html( self, keywords: str, region: str = "wt-wt", timelimit: Optional[str] = None, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo text search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". timelimit: d, w, m, y. Defaults to None. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with search results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" payload = { "q": keywords, "s": "0", "o": "json", "api": "d.js", "vqd": "", "kl": region, "bing_market": region, } if timelimit: payload["df"] = timelimit if max_results and max_results > 20: vqd = self._get_vqd(keywords) payload["vqd"] = vqd cache = set() results: List[Dict[str, str]] = [] def _text_html_page(s: int) -> List[Dict[str, str]]: payload["s"] = f"{s}" resp_content = self._get_url("POST", "https://html.duckduckgo.com/html", data=payload) if b"No results." in resp_content: return [] page_results = [] tree = document_fromstring(resp_content, self.parser) elements = tree.xpath("//div[h2]") if not isinstance(elements, List): return [] for e in elements: if isinstance(e, _Element): hrefxpath = e.xpath("./a/@href") href = str(hrefxpath[0]) if hrefxpath and isinstance(hrefxpath, List) else None if ( href and href not in cache and not href.startswith( ("http://www.google.com/search?q=", "https://duckduckgo.com/y.js?ad_domain") ) ): cache.add(href) titlexpath = e.xpath("./h2/a/text()") title = str(titlexpath[0]) if titlexpath and isinstance(titlexpath, List) else "" bodyxpath = e.xpath("./a//text()") body = "".join(str(x) for x in bodyxpath) if bodyxpath and isinstance(bodyxpath, List) else "" result = { "title": _normalize(title), "href": _normalize_url(href), "body": _normalize(body), } page_results.append(result) return page_results slist = [0] if max_results: max_results = min(max_results, 2023) slist.extend(range(23, max_results, 50)) try: for r in self._executor.map(_text_html_page, slist): results.extend(r) except Exception as e: raise e return list(islice(results, max_results)) def _text_lite( self, keywords: str, region: str = "wt-wt", timelimit: Optional[str] = None, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo text search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". timelimit: d, w, m, y. Defaults to None. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with search results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" payload = { "q": keywords, "s": "0", "o": "json", "api": "d.js", "vqd": "", "kl": region, "bing_market": region, } if timelimit: payload["df"] = timelimit cache = set() results: List[Dict[str, str]] = [] def _text_lite_page(s: int) -> List[Dict[str, str]]: payload["s"] = f"{s}" resp_content = self._get_url("POST", "https://lite.duckduckgo.com/lite/", data=payload) if b"No more results." in resp_content: return [] page_results = [] tree = document_fromstring(resp_content, self.parser) elements = tree.xpath("//table[last()]//tr") if not isinstance(elements, List): return [] data = zip(cycle(range(1, 5)), elements) for i, e in data: if isinstance(e, _Element): if i == 1: hrefxpath = e.xpath(".//a//@href") href = str(hrefxpath[0]) if hrefxpath and isinstance(hrefxpath, List) else None if ( href is None or href in cache or href.startswith( ("http://www.google.com/search?q=", "https://duckduckgo.com/y.js?ad_domain") ) ): [next(data, None) for _ in range(3)] # skip block(i=1,2,3,4) else: cache.add(href) titlexpath = e.xpath(".//a//text()") title = str(titlexpath[0]) if titlexpath and isinstance(titlexpath, List) else "" elif i == 2: bodyxpath = e.xpath(".//td[@class='result-snippet']//text()") body = ( "".join(str(x) for x in bodyxpath).strip() if bodyxpath and isinstance(bodyxpath, List) else "" ) if href: result = { "title": _normalize(title), "href": _normalize_url(href), "body": _normalize(body), } page_results.append(result) return page_results slist = [0] if max_results: max_results = min(max_results, 2023) slist.extend(range(23, max_results, 50)) try: for r in self._executor.map(_text_lite_page, slist): results.extend(r) except Exception as e: raise e return list(islice(results, max_results)) def images( self, keywords: str, region: str = "wt-wt", safesearch: str = "moderate", timelimit: Optional[str] = None, size: Optional[str] = None, color: Optional[str] = None, type_image: Optional[str] = None, layout: Optional[str] = None, license_image: Optional[str] = None, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo images search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". safesearch: on, moderate, off. Defaults to "moderate". timelimit: Day, Week, Month, Year. Defaults to None. size: Small, Medium, Large, Wallpaper. Defaults to None. color: color, Monochrome, Red, Orange, Yellow, Green, Blue, Purple, Pink, Brown, Black, Gray, Teal, White. Defaults to None. type_image: photo, clipart, gif, transparent, line. Defaults to None. layout: Square, Tall, Wide. Defaults to None. license_image: any (All Creative Commons), Public (PublicDomain), Share (Free to Share and Use), ShareCommercially (Free to Share and Use Commercially), Modify (Free to Modify, Share, and Use), ModifyCommercially (Free to Modify, Share, and Use Commercially). Defaults to None. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with images search results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" vqd = self._get_vqd(keywords) safesearch_base = {"on": "1", "moderate": "1", "off": "-1"} timelimit = f"time:{timelimit}" if timelimit else "" size = f"size:{size}" if size else "" color = f"color:{color}" if color else "" type_image = f"type:{type_image}" if type_image else "" layout = f"layout:{layout}" if layout else "" license_image = f"license:{license_image}" if license_image else "" payload = { "l": region, "o": "json", "q": keywords, "vqd": vqd, "f": f"{timelimit},{size},{color},{type_image},{layout},{license_image}", "p": safesearch_base[safesearch.lower()], } cache = set() results: List[Dict[str, str]] = [] def _images_page(s: int) -> List[Dict[str, str]]: payload["s"] = f"{s}" resp_content = self._get_url("GET", "https://duckduckgo.com/i.js", params=payload) resp_json = json_loads(resp_content) page_data = resp_json.get("results", []) page_results = [] for row in page_data: image_url = row.get("image") if image_url and image_url not in cache: cache.add(image_url) result = { "title": row["title"], "image": _normalize_url(image_url), "thumbnail": _normalize_url(row["thumbnail"]), "url": _normalize_url(row["url"]), "height": row["height"], "width": row["width"], "source": row["source"], } page_results.append(result) return page_results slist = [0] if max_results: max_results = min(max_results, 500) slist.extend(range(100, max_results, 100)) try: for r in self._executor.map(_images_page, slist): results.extend(r) except Exception as e: raise e return list(islice(results, max_results)) def videos( self, keywords: str, region: str = "wt-wt", safesearch: str = "moderate", timelimit: Optional[str] = None, resolution: Optional[str] = None, duration: Optional[str] = None, license_videos: Optional[str] = None, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo videos search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". safesearch: on, moderate, off. Defaults to "moderate". timelimit: d, w, m. Defaults to None. resolution: high, standart. Defaults to None. duration: short, medium, long. Defaults to None. license_videos: creativeCommon, youtube. Defaults to None. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with videos search results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" vqd = self._get_vqd(keywords) safesearch_base = {"on": "1", "moderate": "-1", "off": "-2"} timelimit = f"publishedAfter:{timelimit}" if timelimit else "" resolution = f"videoDefinition:{resolution}" if resolution else "" duration = f"videoDuration:{duration}" if duration else "" license_videos = f"videoLicense:{license_videos}" if license_videos else "" payload = { "l": region, "o": "json", "q": keywords, "vqd": vqd, "f": f"{timelimit},{resolution},{duration},{license_videos}", "p": safesearch_base[safesearch.lower()], } cache = set() results: List[Dict[str, str]] = [] def _videos_page(s: int) -> List[Dict[str, str]]: payload["s"] = f"{s}" resp_content = self._get_url("GET", "https://duckduckgo.com/v.js", params=payload) resp_json = json_loads(resp_content) page_data = resp_json.get("results", []) page_results = [] for row in page_data: if row["content"] not in cache: cache.add(row["content"]) page_results.append(row) return page_results slist = [0] if max_results: max_results = min(max_results, 400) slist.extend(range(60, max_results, 60)) try: for r in self._executor.map(_videos_page, slist): results.extend(r) except Exception as e: raise e return list(islice(results, max_results)) def news( self, keywords: str, region: str = "wt-wt", safesearch: str = "moderate", timelimit: Optional[str] = None, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo news search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". safesearch: on, moderate, off. Defaults to "moderate". timelimit: d, w, m. Defaults to None. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with news search results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" vqd = self._get_vqd(keywords) safesearch_base = {"on": "1", "moderate": "-1", "off": "-2"} payload = { "l": region, "o": "json", "noamp": "1", "q": keywords, "vqd": vqd, "p": safesearch_base[safesearch.lower()], } if timelimit: payload["df"] = timelimit cache = set() results: List[Dict[str, str]] = [] def _news_page(s: int) -> List[Dict[str, str]]: payload["s"] = f"{s}" resp_content = self._get_url("GET", "https://duckduckgo.com/news.js", params=payload) resp_json = json_loads(resp_content) page_data = resp_json.get("results", []) page_results = [] for row in page_data: if row["url"] not in cache: cache.add(row["url"]) image_url = row.get("image", None) result = { "date": datetime.fromtimestamp(row["date"], timezone.utc).isoformat(), "title": row["title"], "body": _normalize(row["excerpt"]), "url": _normalize_url(row["url"]), "image": _normalize_url(image_url), "source": row["source"], } page_results.append(result) return page_results slist = [0] if max_results: max_results = min(max_results, 120) slist.extend(range(30, max_results, 30)) try: for r in self._executor.map(_news_page, slist): results.extend(r) except Exception as e: raise e return list(islice(results, max_results)) def answers(self, keywords: str) -> List[Dict[str, str]]: """DuckDuckGo instant answers. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query, Returns: List of dictionaries with instant answers results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" payload = { "q": f"what is {keywords}", "format": "json", } resp_content = self._get_url("GET", "https://api.duckduckgo.com/", params=payload) page_data = json_loads(resp_content) results = [] answer = page_data.get("AbstractText") url = page_data.get("AbstractURL") if answer: results.append( { "icon": None, "text": answer, "topic": None, "url": url, } ) # related payload = { "q": f"{keywords}", "format": "json", } resp_content = self._get_url("GET", "https://api.duckduckgo.com/", params=payload) resp_json = json_loads(resp_content) page_data = resp_json.get("RelatedTopics", []) for row in page_data: topic = row.get("Name") if not topic: icon = row["Icon"].get("URL") results.append( { "icon": f"https://duckduckgo.com{icon}" if icon else "", "text": row["Text"], "topic": None, "url": row["FirstURL"], } ) else: for subrow in row["Topics"]: icon = subrow["Icon"].get("URL") results.append( { "icon": f"https://duckduckgo.com{icon}" if icon else "", "text": subrow["Text"], "topic": topic, "url": subrow["FirstURL"], } ) return results def suggestions(self, keywords: str, region: str = "wt-wt") -> List[Dict[str, str]]: """DuckDuckGo suggestions. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query. region: wt-wt, us-en, uk-en, ru-ru, etc. Defaults to "wt-wt". Returns: List of dictionaries with suggestions results. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" payload = { "q": keywords, "kl": region, } resp_content = self._get_url("GET", "https://duckduckgo.com/ac/", params=payload) page_data = json_loads(resp_content) return [r for r in page_data] def maps( self, keywords: str, place: Optional[str] = None, street: Optional[str] = None, city: Optional[str] = None, county: Optional[str] = None, state: Optional[str] = None, country: Optional[str] = None, postalcode: Optional[str] = None, latitude: Optional[str] = None, longitude: Optional[str] = None, radius: int = 0, max_results: Optional[int] = None, ) -> List[Dict[str, str]]: """DuckDuckGo maps search. Query params: https://duckduckgo.com/params. Args: keywords: keywords for query place: if set, the other parameters are not used. Defaults to None. street: house number/street. Defaults to None. city: city of search. Defaults to None. county: county of search. Defaults to None. state: state of search. Defaults to None. country: country of search. Defaults to None. postalcode: postalcode of search. Defaults to None. latitude: geographic coordinate (north-south position). Defaults to None. longitude: geographic coordinate (east-west position); if latitude and longitude are set, the other parameters are not used. Defaults to None. radius: expand the search square by the distance in kilometers. Defaults to 0. max_results: max number of results. If None, returns results only from the first response. Defaults to None. Returns: List of dictionaries with maps search results, or None if there was an error. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" vqd = self._get_vqd(keywords) # if longitude and latitude are specified, skip the request about bbox to the nominatim api if latitude and longitude: lat_t = Decimal(latitude.replace(",", ".")) lat_b = Decimal(latitude.replace(",", ".")) lon_l = Decimal(longitude.replace(",", ".")) lon_r = Decimal(longitude.replace(",", ".")) if radius == 0: radius = 1 # otherwise request about bbox to nominatim api else: if place: params = { "q": place, "polygon_geojson": "0", "format": "jsonv2", } else: params = { "polygon_geojson": "0", "format": "jsonv2", } if street: params["street"] = street if city: params["city"] = city if county: params["county"] = county if state: params["state"] = state if country: params["country"] = country if postalcode: params["postalcode"] = postalcode # request nominatim api to get coordinates box resp_content = self._get_url( "GET", "https://nominatim.openstreetmap.org/search.php", params=params, ) if resp_content == b"[]": raise WebscoutE("maps() Coordinates are not found, check function parameters.") resp_json = json_loads(resp_content) coordinates = resp_json[0]["boundingbox"] lat_t, lon_l = Decimal(coordinates[1]), Decimal(coordinates[2]) lat_b, lon_r = Decimal(coordinates[0]), Decimal(coordinates[3]) # if a radius is specified, expand the search square lat_t += Decimal(radius) * Decimal(0.008983) lat_b -= Decimal(radius) * Decimal(0.008983) lon_l -= Decimal(radius) * Decimal(0.008983) lon_r += Decimal(radius) * Decimal(0.008983) logger.debug(f"bbox coordinates\n{lat_t} {lon_l}\n{lat_b} {lon_r}") cache = set() results: List[Dict[str, str]] = [] def _maps_page( bbox: Tuple[Decimal, Decimal, Decimal, Decimal], ) -> Optional[List[Dict[str, str]]]: if max_results and len(results) >= max_results: return None lat_t, lon_l, lat_b, lon_r = bbox params = { "q": keywords, "vqd": vqd, "tg": "maps_places", "rt": "D", "mkexp": "b", "wiki_info": "1", "is_requery": "1", "bbox_tl": f"{lat_t},{lon_l}", "bbox_br": f"{lat_b},{lon_r}", "strict_bbox": "1", } resp_content = self._get_url("GET", "https://duckduckgo.com/local.js", params=params) resp_json = json_loads(resp_content) page_data = resp_json.get("results", []) page_results = [] for res in page_data: r_name = f'{res["name"]} {res["address"]}' if r_name in cache: continue else: cache.add(r_name) result = { "title": res["name"], "address": res["address"], "country_code": res["country_code"], "url": _normalize_url(res["website"]), "phone": res["phone"] or "", "latitude": res["coordinates"]["latitude"], "longitude": res["coordinates"]["longitude"], "source": _normalize_url(res["url"]), "image": x.get("image", "") if (x := res["embed"]) else "", "desc": x.get("description", "") if (x := res["embed"]) else "", "hours": res["hours"] or "", "category": res["ddg_category"] or "", "facebook": f"www.facebook.com/profile.php?id={x}" if (x := res["facebook_id"]) else "", "instagram": f"https://www.instagram.com/{x}" if (x := res["instagram_id"]) else "", "twitter": f"https://twitter.com/{x}" if (x := res["twitter_id"]) else "", } page_results.append(result) return page_results # search squares (bboxes) start_bbox = (lat_t, lon_l, lat_b, lon_r) work_bboxes = [start_bbox] while work_bboxes: queue_bboxes = [] # for next iteration, at the end of the iteration work_bboxes = queue_bboxes tasks = [] for bbox in work_bboxes: tasks.append(bbox) # if distance between coordinates > 1, divide the square into 4 parts and save them in queue_bboxes if _calculate_distance(lat_t, lon_l, lat_b, lon_r) > 1: lat_t, lon_l, lat_b, lon_r = bbox lat_middle = (lat_t + lat_b) / 2 lon_middle = (lon_l + lon_r) / 2 bbox1 = (lat_t, lon_l, lat_middle, lon_middle) bbox2 = (lat_t, lon_middle, lat_middle, lon_r) bbox3 = (lat_middle, lon_l, lat_b, lon_middle) bbox4 = (lat_middle, lon_middle, lat_b, lon_r) queue_bboxes.extend([bbox1, bbox2, bbox3, bbox4]) # gather tasks using asyncio.wait_for and timeout work_bboxes_results = [] try: for r in self._executor.map(_maps_page, tasks): if r: work_bboxes_results.extend(r) except Exception as e: raise e for x in work_bboxes_results: if isinstance(x, list): results.extend(x) elif isinstance(x, dict): results.append(x) work_bboxes = queue_bboxes if not max_results or len(results) >= max_results or len(work_bboxes_results) == 0: break return list(islice(results, max_results)) def translate( self, keywords: Union[List[str], str], from_: Optional[str] = None, to: str = "en" ) -> List[Dict[str, str]]: """DuckDuckGo translate. Args: keywords: string or list of strings to translate. from_: translate from (defaults automatically). Defaults to None. to: what language to translate. Defaults to "en". Returns: List od dictionaries with translated keywords. Raises: WebscoutE: Base exception for webscout errors. RatelimitE: Inherits from WebscoutE, raised for exceeding API request rate limits. TimeoutE: Inherits from WebscoutE, raised for API request timeouts. """ assert keywords, "keywords is mandatory" vqd = self._get_vqd("translate") payload = { "vqd": vqd, "query": "translate", "to": to, } if from_: payload["from"] = from_ def _translate_keyword(keyword: str) -> Dict[str, str]: resp_content = self._get_url( "POST", "https://duckduckgo.com/translation.js", params=payload, content=keyword.encode(), ) page_data: Dict[str, str] = json_loads(resp_content) page_data["original"] = keyword return page_data if isinstance(keywords, str): keywords = [keywords] results = [] try: for r in self._executor.map(_translate_keyword, keywords): results.append(r) except Exception as e: raise e return results import requests import http.cookiejar as cookiejar import sys import json from xml.etree import ElementTree import re from requests import HTTPError import html.parser html_parser = html.parser.HTMLParser() import html def unescape(string): return html.unescape(string) WATCH_URL = 'https://www.youtube.com/watch?v={video_id}' class TranscriptRetrievalError(Exception): """ Base class for exceptions raised when a transcript cannot be retrieved. """ ERROR_MESSAGE = '\nCould not retrieve a transcript for the video {video_url}!' CAUSE_MESSAGE_INTRO = ' This is most likely caused by:\n\n{cause}' CAUSE_MESSAGE = '' GITHUB_REFERRAL = ( '\n\nIf you are sure that the described cause is not responsible for this error ' 'and that a transcript should be retrievable, please create an issue at ' 'https://github.com/OE-LUCIFER/Webscout/issues. ' 'Please add which version of webscout you are using ' 'and provide the information needed to replicate the error. ' ) def __init__(self, video_id): self.video_id = video_id super(TranscriptRetrievalError, self).__init__(self._build_error_message()) def _build_error_message(self): cause = self.cause error_message = self.ERROR_MESSAGE.format(video_url=WATCH_URL.format(video_id=self.video_id)) if cause: error_message += self.CAUSE_MESSAGE_INTRO.format(cause=cause) + self.GITHUB_REFERRAL return error_message @property def cause(self): return self.CAUSE_MESSAGE class YouTubeRequestFailedError(TranscriptRetrievalError): CAUSE_MESSAGE = 'Request to YouTube failed: {reason}' def __init__(self, video_id, http_error): self.reason = str(http_error) super(YouTubeRequestFailedError, self).__init__(video_id) @property def cause(self): return self.CAUSE_MESSAGE.format(reason=self.reason) class VideoUnavailableError(TranscriptRetrievalError): CAUSE_MESSAGE = 'The video is no longer available' class InvalidVideoIdError(TranscriptRetrievalError): CAUSE_MESSAGE = ( 'You provided an invalid video id. Make sure you are using the video id and NOT the url!\n\n' 'Do NOT run: `YouTubeTranscriptApi.get_transcript("https://www.youtube.com/watch?v=1234")`\n' 'Instead run: `YouTubeTranscriptApi.get_transcript("1234")`' ) class TooManyRequestsError(TranscriptRetrievalError): CAUSE_MESSAGE = ( 'YouTube is receiving too many requests from this IP and now requires solving a captcha to continue. ' 'One of the following things can be done to work around this:\n\ - Manually solve the captcha in a browser and export the cookie. ' 'Read here how to use that cookie with ' 'youtube-transcript-api: https://github.com/jdepoix/youtube-transcript-api#cookies\n\ - Use a different IP address\n\ - Wait until the ban on your IP has been lifted' ) class TranscriptsDisabledError(TranscriptRetrievalError): CAUSE_MESSAGE = 'Subtitles are disabled for this video' class NoTranscriptAvailableError(TranscriptRetrievalError): CAUSE_MESSAGE = 'No transcripts are available for this video' class NotTranslatableError(TranscriptRetrievalError): CAUSE_MESSAGE = 'The requested language is not translatable' class TranslationLanguageNotAvailableError(TranscriptRetrievalError): CAUSE_MESSAGE = 'The requested translation language is not available' class CookiePathInvalidError(TranscriptRetrievalError): CAUSE_MESSAGE = 'The provided cookie file was unable to be loaded' class CookiesInvalidError(TranscriptRetrievalError): CAUSE_MESSAGE = 'The cookies provided are not valid (may have expired)' class FailedToCreateConsentCookieError(TranscriptRetrievalError): CAUSE_MESSAGE = 'Failed to automatically give consent to saving cookies' class NoTranscriptFoundError(TranscriptRetrievalError): CAUSE_MESSAGE = ( 'No transcripts were found for any of the requested language codes: {requested_language_codes}\n\n' '{transcript_data}' ) def __init__(self, video_id, requested_language_codes, transcript_data): self._requested_language_codes = requested_language_codes self._transcript_data = transcript_data super(NoTranscriptFoundError, self).__init__(video_id) @property def cause(self): return self.CAUSE_MESSAGE.format( requested_language_codes=self._requested_language_codes, transcript_data=str(self._transcript_data), ) def _raise_http_errors(response, video_id): try: response.raise_for_status() return response except HTTPError as error: raise YouTubeRequestFailedError(error, video_id) class TranscriptListFetcher(object): def __init__(self, http_client): self._http_client = http_client def fetch(self, video_id): return TranscriptList.build( self._http_client, video_id, self._extract_captions_json(self._fetch_video_html(video_id), video_id), ) def _extract_captions_json(self, html, video_id): splitted_html = html.split('"captions":') if len(splitted_html) <= 1: if video_id.startswith('http://') or video_id.startswith('https://'): raise InvalidVideoIdError(video_id) if 'class="g-recaptcha"' in html: raise TooManyRequestsError(video_id) if '"playabilityStatus":' not in html: raise VideoUnavailableError(video_id) raise TranscriptsDisabledError(video_id) captions_json = json.loads( splitted_html[1].split(',"videoDetails')[0].replace('\n', '') ).get('playerCaptionsTracklistRenderer') if captions_json is None: raise TranscriptsDisabledError(video_id) if 'captionTracks' not in captions_json: raise TranscriptsDisabledError(video_id) return captions_json def _create_consent_cookie(self, html, video_id): match = re.search('name="v" value="(.*?)"', html) if match is None: raise FailedToCreateConsentCookieError(video_id) self._http_client.cookies.set('CONSENT', 'YES+' + match.group(1), domain='.youtube.com') def _fetch_video_html(self, video_id): html = self._fetch_html(video_id) if 'action="https://consent.youtube.com/s"' in html: self._create_consent_cookie(html, video_id) html = self._fetch_html(video_id) if 'action="https://consent.youtube.com/s"' in html: raise FailedToCreateConsentCookieError(video_id) return html def _fetch_html(self, video_id): response = self._http_client.get(WATCH_URL.format(video_id=video_id), headers={'Accept-Language': 'en-US'}) return unescape(_raise_http_errors(response, video_id).text) class TranscriptList(object): """ This object represents a list of transcripts. It can be iterated over to list all transcripts which are available for a given YouTube video. Also it provides functionality to search for a transcript in a given language. """ def __init__(self, video_id, manually_created_transcripts, generated_transcripts, translation_languages): """ The constructor is only for internal use. Use the static build method instead. :param video_id: the id of the video this TranscriptList is for :type video_id: str :param manually_created_transcripts: dict mapping language codes to the manually created transcripts :type manually_created_transcripts: dict[str, Transcript] :param generated_transcripts: dict mapping language codes to the generated transcripts :type generated_transcripts: dict[str, Transcript] :param translation_languages: list of languages which can be used for translatable languages :type translation_languages: list[dict[str, str]] """ self.video_id = video_id self._manually_created_transcripts = manually_created_transcripts self._generated_transcripts = generated_transcripts self._translation_languages = translation_languages @staticmethod def build(http_client, video_id, captions_json): """ Factory method for TranscriptList. :param http_client: http client which is used to make the transcript retrieving http calls :type http_client: requests.Session :param video_id: the id of the video this TranscriptList is for :type video_id: str :param captions_json: the JSON parsed from the YouTube pages static HTML :type captions_json: dict :return: the created TranscriptList :rtype TranscriptList: """ translation_languages = [ { 'language': translation_language['languageName']['simpleText'], 'language_code': translation_language['languageCode'], } for translation_language in captions_json.get('translationLanguages', []) ] manually_created_transcripts = {} generated_transcripts = {} for caption in captions_json['captionTracks']: if caption.get('kind', '') == 'asr': transcript_dict = generated_transcripts else: transcript_dict = manually_created_transcripts transcript_dict[caption['languageCode']] = Transcript( http_client, video_id, caption['baseUrl'], caption['name']['simpleText'], caption['languageCode'], caption.get('kind', '') == 'asr', translation_languages if caption.get('isTranslatable', False) else [], ) return TranscriptList( video_id, manually_created_transcripts, generated_transcripts, translation_languages, ) def __iter__(self): return iter(list(self._manually_created_transcripts.values()) + list(self._generated_transcripts.values())) def find_transcript(self, language_codes): """ Finds a transcript for a given language code. Manually created transcripts are returned first and only if none are found, generated transcripts are used. If you only want generated transcripts use `find_manually_created_transcript` instead. :param language_codes: A list of language codes in a descending priority. For example, if this is set to ['de', 'en'] it will first try to fetch the german transcript (de) and then fetch the english transcript (en) if it fails to do so. :type languages: list[str] :return: the found Transcript :rtype Transcript: :raises: NoTranscriptFound """ return self._find_transcript(language_codes, [self._manually_created_transcripts, self._generated_transcripts]) def find_generated_transcript(self, language_codes): """ Finds an automatically generated transcript for a given language code. :param language_codes: A list of language codes in a descending priority. For example, if this is set to ['de', 'en'] it will first try to fetch the german transcript (de) and then fetch the english transcript (en) if it fails to do so. :type languages: list[str] :return: the found Transcript :rtype Transcript: :raises: NoTranscriptFound """ return self._find_transcript(language_codes, [self._generated_transcripts]) def find_manually_created_transcript(self, language_codes): """ Finds a manually created transcript for a given language code. :param language_codes: A list of language codes in a descending priority. For example, if this is set to ['de', 'en'] it will first try to fetch the german transcript (de) and then fetch the english transcript (en) if it fails to do so. :type languages: list[str] :return: the found Transcript :rtype Transcript: :raises: NoTranscriptFound """ return self._find_transcript(language_codes, [self._manually_created_transcripts]) def _find_transcript(self, language_codes, transcript_dicts): for language_code in language_codes: for transcript_dict in transcript_dicts: if language_code in transcript_dict: return transcript_dict[language_code] raise NoTranscriptFoundError( self.video_id, language_codes, self ) def __str__(self): return ( 'For this video ({video_id}) transcripts are available in the following languages:\n\n' '(MANUALLY CREATED)\n' '{available_manually_created_transcript_languages}\n\n' '(GENERATED)\n' '{available_generated_transcripts}\n\n' '(TRANSLATION LANGUAGES)\n' '{available_translation_languages}' ).format( video_id=self.video_id, available_manually_created_transcript_languages=self._get_language_description( str(transcript) for transcript in self._manually_created_transcripts.values() ), available_generated_transcripts=self._get_language_description( str(transcript) for transcript in self._generated_transcripts.values() ), available_translation_languages=self._get_language_description( '{language_code} ("{language}")'.format( language=translation_language['language'], language_code=translation_language['language_code'], ) for translation_language in self._translation_languages ) ) def _get_language_description(self, transcript_strings): description = '\n'.join(' - {transcript}'.format(transcript=transcript) for transcript in transcript_strings) return description if description else 'None' class Transcript(object): def __init__(self, http_client, video_id, url, language, language_code, is_generated, translation_languages): """ You probably don't want to initialize this directly. Usually you'll access Transcript objects using a TranscriptList. :param http_client: http client which is used to make the transcript retrieving http calls :type http_client: requests.Session :param video_id: the id of the video this TranscriptList is for :type video_id: str :param url: the url which needs to be called to fetch the transcript :param language: the name of the language this transcript uses :param language_code: :param is_generated: :param translation_languages: """ self._http_client = http_client self.video_id = video_id self._url = url self.language = language self.language_code = language_code self.is_generated = is_generated self.translation_languages = translation_languages self._translation_languages_dict = { translation_language['language_code']: translation_language['language'] for translation_language in translation_languages } def fetch(self, preserve_formatting=False): """ Loads the actual transcript data. :param preserve_formatting: whether to keep select HTML text formatting :type preserve_formatting: bool :return: a list of dictionaries containing the 'text', 'start' and 'duration' keys :rtype [{'text': str, 'start': float, 'end': float}]: """ response = self._http_client.get(self._url, headers={'Accept-Language': 'en-US'}) return _TranscriptParser(preserve_formatting=preserve_formatting).parse( _raise_http_errors(response, self.video_id).text, ) def __str__(self): return '{language_code} ("{language}"){translation_description}'.format( language=self.language, language_code=self.language_code, translation_description='[TRANSLATABLE]' if self.is_translatable else '' ) @property def is_translatable(self): return len(self.translation_languages) > 0 def translate(self, language_code): if not self.is_translatable: raise NotTranslatableError(self.video_id) if language_code not in self._translation_languages_dict: raise TranslationLanguageNotAvailableError(self.video_id) return Transcript( self._http_client, self.video_id, '{url}&tlang={language_code}'.format(url=self._url, language_code=language_code), self._translation_languages_dict[language_code], language_code, True, [], ) class _TranscriptParser(object): _FORMATTING_TAGS = [ 'strong', # important 'em', # emphasized 'b', # bold 'i', # italic 'mark', # marked 'small', # smaller 'del', # deleted 'ins', # inserted 'sub', # subscript 'sup', # superscript ] def __init__(self, preserve_formatting=False): self._html_regex = self._get_html_regex(preserve_formatting) def _get_html_regex(self, preserve_formatting): if preserve_formatting: formats_regex = '|'.join(self._FORMATTING_TAGS) formats_regex = r'<\/?(?!\/?(' + formats_regex + r')\b).*?\b>' html_regex = re.compile(formats_regex, re.IGNORECASE) else: html_regex = re.compile(r'<[^>]*>', re.IGNORECASE) return html_regex def parse(self, plain_data): return [ { 'text': re.sub(self._html_regex, '', unescape(xml_element.text)), 'start': float(xml_element.attrib['start']), 'duration': float(xml_element.attrib.get('dur', '0.0')), } for xml_element in ElementTree.fromstring(plain_data) if xml_element.text is not None ] WATCH_URL = 'https://www.youtube.com/watch?v={video_id}' class transcriber(object): @classmethod def list_transcripts(cls, video_id, proxies=None, cookies=None): with requests.Session() as http_client: if cookies: http_client.cookies = cls._load_cookies(cookies, video_id) http_client.proxies = proxies if proxies else {} return TranscriptListFetcher(http_client).fetch(video_id) @classmethod def get_transcripts(cls, video_ids, languages=('en',), continue_after_error=False, proxies=None, cookies=None, preserve_formatting=False): assert isinstance(video_ids, list), "`video_ids` must be a list of strings" data = {} unretrievable_videos = [] for video_id in video_ids: try: data[video_id] = cls.get_transcript(video_id, languages, proxies, cookies, preserve_formatting) except Exception as exception: if not continue_after_error: raise exception unretrievable_videos.append(video_id) return data, unretrievable_videos @classmethod def get_transcript(cls, video_id, languages=('en',), proxies=None, cookies=None, preserve_formatting=False): assert isinstance(video_id, str), "`video_id` must be a string" return cls.list_transcripts(video_id, proxies, cookies).find_transcript(languages).fetch(preserve_formatting=preserve_formatting) @classmethod def _load_cookies(cls, cookies, video_id): try: cookie_jar = cookiejar.MozillaCookieJar() cookie_jar.load(cookies) if not cookie_jar: raise CookiesInvalidError(video_id) return cookie_jar except: raise CookiePathInvalidError(video_id) class LLM: def __init__(self, model: str, system_message: str = "You are a Helpful AI."): self.model = model self.conversation_history = [{"role": "system", "content": system_message}] def chat(self, messages: List[Dict[str, str]]) -> Union[str, None]: url = "https://api.deepinfra.com/v1/openai/chat/completions" headers = { 'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36', 'Accept-Language': 'en,fr-FR;q=0.9,fr;q=0.8,es-ES;q=0.7,es;q=0.6,en-US;q=0.5,am;q=0.4,de;q=0.3', 'Cache-Control': 'no-cache', 'Connection': 'keep-alive', 'Content-Type': 'application/json', 'Origin': 'https://deepinfra.com', 'Pragma': 'no-cache', 'Referer': 'https://deepinfra.com/', 'Sec-Fetch-Dest': 'empty', 'Sec-Fetch-Mode': 'cors', 'Sec-Fetch-Site': 'same-site', 'X-Deepinfra-Source': 'web-embed', 'accept': 'text/event-stream', 'sec-ch-ua': '"Google Chrome";v="119", "Chromium";v="119", "Not?A_Brand";v="24"', 'sec-ch-ua-mobile': '?0', 'sec-ch-ua-platform': '"macOS"' } data = json.dumps( { 'model': self.model, 'messages': messages, 'temperature': 0.7, 'max_tokens': 16000, 'stop': [], 'stream': False #dont change it }, separators=(',', ':') ) try: result = requests.post(url=url, data=data, headers=headers) return result.json()['choices'][0]['message']['content'] except: return None def fastai(user, model="llama3-70b", system="Answer as concisely as possible."): env_type = "tp16405b" if "405b" in model else "tp16" data = {'body': {'messages': [{'role': 'system', 'content': system}, {'role': 'user', 'content': user}], 'stream': True, 'model': model}, 'env_type': env_type} with requests.post('https://fast.snova.ai/api/completion', headers={'content-type': 'application/json'}, json=data, stream=True) as response: output = '' for line in response.iter_lines(decode_unicode=True): if line.startswith('data:'): try: data = json.loads(line[len('data: '):]) output += data.get("choices", [{}])[0].get("delta", {}).get("content", '') except json.JSONDecodeError: if line[len('data: '):] == '[DONE]': break return output async def fastai_stream(user, model="llama3-8b", system="Answer as concisely as possible."): env_type = "tp16405b" if "405b" in model else "tp16" data = {'body': {'messages': [{'role': 'system', 'content': system}, {'role': 'user', 'content': user}], 'stream': True, 'model': model}, 'env_type': env_type} async with aiohttp.ClientSession() as session: async with session.post('https://fast.snova.ai/api/completion', headers={'content-type': 'application/json'}, json=data) as response: async for line in response.content: line = line.decode('utf-8').strip() if line.startswith('data:'): try: data = json.loads(line[len('data: '):]) content = data.get("choices", [{}])[0].get("delta", {}).get("content", '') if content: yield f"data: {json.dumps({'response': content})}\n\n" except json.JSONDecodeError: if line[len('data: '):] == '[DONE]': break