notion-md-crawler

A library to recursively retrieve and serialize Notion pages and databases with customization for machine learning applications.


🌟 Features

- 🕷️ Crawling Pages and Databases: Dig deep into Notion's hierarchical structure with ease.
- 📝 Serialize to Markdown: Seamlessly convert Notion pages to Markdown for easy use in machine learning and other.
- 🛠️ Custom Serialization: Adapt the serialization process to fit your specific machine learning needs.
- ⏳ Async Generator: Yields results on a page-by-page basis, so even huge documents can be made memory efficient.

🛠️ Installation

@notionhq/client must also be installed.

Using npm 📦:

``bash
npm install notion-md-crawler @notionhq/client
`

Using yarn 🧶:

`bash
yarn add notion-md-crawler @notionhq/client
`

Using pnpm 🚀:

`bash
pnpm add notion-md-crawler @notionhq/client
`

🚀 Quick Start

> ⚠️ Note: Before getting started, create an integration and find the token. Details on methods can be found in API section

Leveraging the power of JavaScript generators, this library is engineered to handle even the most extensive Notion documents with ease. It's designed to yield results page-by-page, allowing for efficient memory usage and real-time processing.

`ts
import { Client } from "@notionhq/client";
import { crawler, pageToString } from "notion-md-crawler";

// Need init notion client with credential.
const client = new Client({ auth: process.env.NOTION_API_KEY });

const crawl = crawler({ client });

const main = async () => {
const rootPageId = "**";
for await (const result of crawl(rootPageId)) {
if (result.success) {
const pageText = pageToString(result.page);
console.log(pageText);
}
}
};

main();
`

🌐 API

$3

Recursively crawl the Notion Page. dbCrawler should be used if the Root is a Notion Database.

> Note: It tries to continue crawling as much as possible even if it fails to retrieve a particular Notion Page.

#### Parameters:

- options (CrawlerOptions): Crawler options.
- rootPageId (string): Id of the root page to be crawled.

#### Returns:

- AsyncGenerator: Crawling results with failed information.

$3

Recursively crawl the Notion Database. crawler should be used if the Root is a Notion Page.

#### Parameters:

- options (CrawlerOptions): Crawler options.
- rootDatabaseId (string): Id of the root page to be crawled.

#### Returns:

- AsyncGenerator: Crawling results with failed information.

$3

| Option | Description | Type | Default |
| ------------------------ | --------------------------------------------------------------------------------------------------- | --------------------------------------------- | ----------- |
| client | Instance of Notion Client. Set up an instance of the Client class imported from @notionhq/client. | Notion Client | - |
| serializers? | Used for custom serialization of Block and Property objects. | Object | undefined |
| serializers?.block? | Map of Notion block type and BlockSerializer. | BlockSerializers | undefined |
| serializers?.property? | Map of Notion Property Type and PropertySerializer. | PropertySerializers | undefined |
| metadataBuilder? | The metadata generation process can be customize. | MetadataBuilder | undefined |
| urlMask? | If specified, the url is masked with the string. | string \| false | false |
| skipPageIds? | List of page Ids to skip crawling (also skips descendant pages) | string[] | undefined |

#### BlockSerializers

Map with Notion block type (like "heading_1", "to_do", "code") as key and BlockSerializer as value.

#### BlockSerializer

BlockSerializer that takes a Notion block object as argument. Returning false will skip serialization of that Notion block.

[Type]

`ts
type BlockSerializer = (
block: NotionBlock,
) => string | false | Promise;
`

#### PropertySerializers

Map with Notion Property Type (like "heading_1", "to_do", "code") as key and PropertySerializer as value.

#### PropertySerializer

PropertySerializer that takes a Notion property object as argument. Returning false will skip serialization of that Notion property.

[Type]

`ts
type PropertySerializer = (
name: string,
block: NotionBlock,
) => string | false | Promise;
`

#### MetadataBuilder

Retrieving metadata is sometimes very important, but the information you want to retrieve will vary depending on the context. MetadataBuilder allows you to customize it according to your use case.

[Example]

`ts
import { crawler, MetadataBuilderParams } from "notion-md-crawler";

const getUrl = (id: string) => https://www.notion.so/${id.replace(/-/g, "")};

const metadataBuilder = ({ page }: MetadataBuilderParams) => ({
url: getUrl(page.metadata.id),
});

const crawl = crawler({ client, metadataBuilder });

for await (const result of crawl("notion-page-id")) {
if (result.success) {
console.log(result.page.metadata.url); // "https://www.notion.so/**"
}
}
`

📊 Use Metadata

Since crawler returns Page objects and Page object contain metadata, you can be used it for machine learning.

🛠️ Custom Serialization

notion-md-crawler gives you the flexibility to customize the serialization logic for various Notion objects to cater to the unique requirements of your machine learning model or any other use case.

$3

You can define your own custom serializer. You can also use the utility function for convenience.

`ts
import { BlockSerializer, crawler, serializer } from "notion-md-crawler";

const customEmbedSerializer: BlockSerializer<"embed"> = (block) => {
if (block.embed.url) return "";

// You can use serializer utility.
const caption = serializer.utils.fromRichText(block.embed.caption);

return

const crawl = crawler({ client, serializers });
`

$3

Returning false in the serializer allows you to skip the serialize of that block. This is useful when you want to omit unnecessary information.

`ts
const image: BlockSerializer<"image"> = () => false;
const crawl = crawler({ client, serializers: { block: { image } } });
`

$3

If you want to customize serialization only in specific cases, you can use the default serializer in a custom serializer.

`ts
import { BlockSerializer, crawler, serializer } from "notion-md-crawler";

const defaultImageSerializer = serializer.block.defaults.image;

const customImageSerializer: BlockSerializer<"image"> = (block) => {
// Utility function to retrieve the link
const { title, href } = serializer.utils.fromLink(block.image);

// If the image is from a specific domain, wrap it in a special div
if (href.includes("special-domain.com")) {
return

const crawl = crawler({ client, serializers });
``

🔍 Supported Blocks and Database properties

$3

| Block Type | Supported |
| ------------------ | --------- |
| Text | ✅ Yes |
| Bookmark | ✅ Yes |
| Bulleted List | ✅ Yes |
| Numbered List | ✅ Yes |
| Heading 1 | ✅ Yes |
| Heading 2 | ✅ Yes |
| Heading 3 | ✅ Yes |
| Quote | ✅ Yes |
| Callout | ✅ Yes |
| Equation (block) | ✅ Yes |
| Equation (inline) | ✅ Yes |
| Todos (checkboxes) | ✅ Yes |
| Table Of Contents | ✅ Yes |
| Divider | ✅ Yes |
| Column | ✅ Yes |
| Column List | ✅ Yes |
| Toggle | ✅ Yes |
| Image | ✅ Yes |
| Embed | ✅ Yes |
| Video | ✅ Yes |
| Figma | ✅ Yes |
| PDF | ✅ Yes |
| Audio | ✅ Yes |
| File | ✅ Yes |
| Link | ✅ Yes |
| Page Link | ✅ Yes |
| External Page Link | ✅ Yes |
| Code (block) | ✅ Yes |
| Code (inline) | ✅ Yes |

$3

| Property Type | Supported |
| ---------------- | --------- |
| Checkbox | ✅ Yes |
| Created By | ✅ Yes |
| Created Time | ✅ Yes |
| Date | ✅ Yes |
| Email | ✅ Yes |
| Files | ✅ Yes |
| Formula | ✅ Yes |
| Last Edited By | ✅ Yes |
| Last Edited Time | ✅ Yes |
| Multi Select | ✅ Yes |
| Number | ✅ Yes |
| People | ✅ Yes |
| Phone Number | ✅ Yes |
| Relation | ✅ Yes |
| Rich Text | ✅ Yes |
| Rollup | ✅ Yes |
| Select | ✅ Yes |
| Status | ✅ Yes |
| Title | ✅ Yes |
| Unique Id | ✅ Yes |
| Url | ✅ Yes |
| Verification | □ No |

💬 Issues and Feedback

For any issues, feedback, or feature requests, please file an issue on GitHub.

📜 License

MIT

---

Made with ❤️ by TomPenguin.