|
| 1 | +"""Notion reader.""" |
| 2 | + |
| 3 | +from datetime import datetime |
| 4 | +import os |
| 5 | +from typing import Any, Dict, List, Optional |
| 6 | + |
| 7 | +import requests # type: ignore |
| 8 | +from llama_index.core.readers.base import BasePydanticReader |
| 9 | +from llama_index.core.schema import Document |
| 10 | + |
| 11 | +INTEGRATION_TOKEN_NAME = "NOTION_INTEGRATION_TOKEN" |
| 12 | +BLOCK_CHILD_URL_TMPL = "https://api.notion.com/v1/blocks/{block_id}/children" |
| 13 | +DATABASE_URL_TMPL = "https://api.notion.com/v1/databases/{database_id}/query" |
| 14 | +SEARCH_URL = "https://api.notion.com/v1/search" |
| 15 | + |
| 16 | + |
| 17 | +def utc_to_iso(utc_time: str) -> datetime: |
| 18 | + return datetime.fromisoformat(utc_time.replace("Z", "+00:00")) |
| 19 | + |
| 20 | + |
| 21 | +# TODO: Notion DB reader coming soon! |
| 22 | +class NotionPageReader(BasePydanticReader): |
| 23 | + """Notion Page reader. |
| 24 | +
|
| 25 | + Reads a set of Notion pages. |
| 26 | +
|
| 27 | + Args: |
| 28 | + integration_token (str): Notion integration token. |
| 29 | +
|
| 30 | + """ |
| 31 | + |
| 32 | + is_remote: bool = True |
| 33 | + token: str |
| 34 | + headers: Dict[str, str] |
| 35 | + |
| 36 | + def __init__(self, integration_token: Optional[str] = None) -> None: |
| 37 | + """Initialize with parameters.""" |
| 38 | + if integration_token is None: |
| 39 | + integration_token = os.getenv(INTEGRATION_TOKEN_NAME) |
| 40 | + if integration_token is None: |
| 41 | + raise ValueError( |
| 42 | + "Must specify `integration_token` or set environment " |
| 43 | + "variable `NOTION_INTEGRATION_TOKEN`." |
| 44 | + ) |
| 45 | + |
| 46 | + token = integration_token |
| 47 | + headers = { |
| 48 | + "Authorization": "Bearer " + token, |
| 49 | + "Content-Type": "application/json", |
| 50 | + "Notion-Version": "2022-06-28", |
| 51 | + } |
| 52 | + |
| 53 | + super().__init__(token=token, headers=headers) |
| 54 | + |
| 55 | + @classmethod |
| 56 | + def class_name(cls) -> str: |
| 57 | + """Get the name identifier of the class.""" |
| 58 | + return "NotionPageReader" |
| 59 | + |
| 60 | + def _read_block(self, block_id: str, num_tabs: int = 0) -> str: |
| 61 | + """Read a block.""" |
| 62 | + done = False |
| 63 | + result_lines_arr = [] |
| 64 | + cur_block_id = block_id |
| 65 | + most_recent_time = None |
| 66 | + while not done: |
| 67 | + block_url = BLOCK_CHILD_URL_TMPL.format(block_id=cur_block_id) |
| 68 | + query_dict: Dict[str, Any] = {} |
| 69 | + |
| 70 | + res = requests.request( |
| 71 | + "GET", block_url, headers=self.headers, json=query_dict |
| 72 | + ) |
| 73 | + data = res.json() |
| 74 | + |
| 75 | + for result in data["results"]: |
| 76 | + result_type = result["type"] |
| 77 | + result_obj = result[result_type] |
| 78 | + |
| 79 | + cur_result_text_arr = [] |
| 80 | + if "rich_text" in result_obj: |
| 81 | + for rich_text in result_obj["rich_text"]: |
| 82 | + # skip if doesn't have text object |
| 83 | + if "text" in rich_text: |
| 84 | + text = rich_text["text"]["content"] |
| 85 | + prefix = "\t" * num_tabs |
| 86 | + cur_result_text_arr.append(prefix + text) |
| 87 | + |
| 88 | + result_block_id = result["id"] |
| 89 | + has_children = result["has_children"] |
| 90 | + if has_children: |
| 91 | + children_text, _ = self._read_block( |
| 92 | + result_block_id, num_tabs=num_tabs + 1 |
| 93 | + ) |
| 94 | + cur_result_text_arr.append(children_text) |
| 95 | + |
| 96 | + cur_result_text = "\n".join(cur_result_text_arr) |
| 97 | + result_lines_arr.append(cur_result_text) |
| 98 | + last_edited_time = result["last_edited_time"] |
| 99 | + |
| 100 | + if most_recent_time is None or utc_to_iso( |
| 101 | + last_edited_time |
| 102 | + ) > utc_to_iso(most_recent_time): |
| 103 | + most_recent_time = last_edited_time |
| 104 | + |
| 105 | + if data["next_cursor"] is None: |
| 106 | + done = True |
| 107 | + break |
| 108 | + else: |
| 109 | + cur_block_id = data["next_cursor"] |
| 110 | + |
| 111 | + block_text = "\n".join(result_lines_arr) |
| 112 | + |
| 113 | + return block_text, most_recent_time |
| 114 | + |
| 115 | + def read_page(self, page_id: str) -> str: |
| 116 | + """Read a page.""" |
| 117 | + return self._read_block(page_id) |
| 118 | + |
| 119 | + def query_database( |
| 120 | + self, database_id: str, query_dict: Dict[str, Any] = {"page_size": 100} |
| 121 | + ) -> List[str]: |
| 122 | + """Get all the pages from a Notion database.""" |
| 123 | + pages = [] |
| 124 | + |
| 125 | + res = requests.post( |
| 126 | + DATABASE_URL_TMPL.format(database_id=database_id), |
| 127 | + headers=self.headers, |
| 128 | + json=query_dict, |
| 129 | + ) |
| 130 | + res.raise_for_status() |
| 131 | + data = res.json() |
| 132 | + |
| 133 | + pages.extend(data.get("results")) |
| 134 | + |
| 135 | + while data.get("has_more"): |
| 136 | + query_dict["start_cursor"] = data.get("next_cursor") |
| 137 | + res = requests.post( |
| 138 | + DATABASE_URL_TMPL.format(database_id=database_id), |
| 139 | + headers=self.headers, |
| 140 | + json=query_dict, |
| 141 | + ) |
| 142 | + res.raise_for_status() |
| 143 | + data = res.json() |
| 144 | + pages.extend(data.get("results")) |
| 145 | + |
| 146 | + return [page["id"] for page in pages] |
| 147 | + |
| 148 | + def search(self, query: str) -> List[str]: |
| 149 | + """Search Notion page given a text query.""" |
| 150 | + done = False |
| 151 | + next_cursor: Optional[str] = None |
| 152 | + page_ids = [] |
| 153 | + while not done: |
| 154 | + query_dict = { |
| 155 | + "query": query, |
| 156 | + } |
| 157 | + if next_cursor is not None: |
| 158 | + query_dict["start_cursor"] = next_cursor |
| 159 | + res = requests.post(SEARCH_URL, headers=self.headers, json=query_dict) |
| 160 | + data = res.json() |
| 161 | + for result in data["results"]: |
| 162 | + page_id = result["id"] |
| 163 | + page_ids.append(page_id) |
| 164 | + |
| 165 | + if data["next_cursor"] is None: |
| 166 | + done = True |
| 167 | + break |
| 168 | + else: |
| 169 | + next_cursor = data["next_cursor"] |
| 170 | + return page_ids |
| 171 | + |
| 172 | + def load_data( |
| 173 | + self, page_ids: List[str] = [], database_id: Optional[str] = None |
| 174 | + ) -> List[Document]: |
| 175 | + """Load data from the input directory. |
| 176 | +
|
| 177 | + Args: |
| 178 | + page_ids (List[str]): List of page ids to load. |
| 179 | + database_id (str): Database_id from which to load page ids. |
| 180 | +
|
| 181 | + Returns: |
| 182 | + List[Document]: List of documents. |
| 183 | +
|
| 184 | + """ |
| 185 | + if not page_ids and not database_id: |
| 186 | + raise ValueError("Must specify either `page_ids` or `database_id`.") |
| 187 | + docs = [] |
| 188 | + if database_id is not None: |
| 189 | + # get all the pages in the database |
| 190 | + page_ids = self.query_database(database_id) |
| 191 | + for page_id in page_ids: |
| 192 | + page_text, most_recent_time = self.read_page(page_id) |
| 193 | + docs.append( |
| 194 | + Document( |
| 195 | + text=page_text, |
| 196 | + id_=page_id, |
| 197 | + extra_info={"page_id": page_id, "updated_at": most_recent_time}, |
| 198 | + ) |
| 199 | + ) |
| 200 | + else: |
| 201 | + for page_id in page_ids: |
| 202 | + page_text, most_recent_time = self.read_page(page_id) |
| 203 | + docs.append( |
| 204 | + Document( |
| 205 | + text=page_text, |
| 206 | + id_=page_id, |
| 207 | + extra_info={"page_id": page_id, "updated_at": most_recent_time}, |
| 208 | + ) |
| 209 | + ) |
| 210 | + |
| 211 | + return docs |
0 commit comments