[BUG] Notion API Error: Unsupported block type ai_block

[tuboihirokidesu]

Hello LlamaIndex team,

I am using the LlamaIndex library to interact with the Notion API and I have encountered an issue. When I try to load pages from Notion using the NotionReader class, I receive an error message stating that the ai_block type is not supported by the API.

Here is the error message I receive:

Error: Notion API error: block type `ai_block` is not supported via the API.

Here is the relevant code from my route.ts file:

route.ts

```ts import { Message, StreamingTextResponse } from 'ai'; import { OpenAI, NotionReader } from 'llamaindex'; import { NextRequest, NextResponse } from 'next/server'; import { createChatEngine } from './engine'; import { LlamaIndexStream } from './llamaindex-stream'; import { Client } from '@notionhq/client'; export const runtime: string = 'nodejs'; export const dynamic: string = 'force-dynamic'; type Page = { metadata: { id: string; title: string; createdTime: string; lastEditedTime: string; parentId?: string; }; properties: string[]; lines: string[]; }; type Pages = Record; export async function POST(request: NextRequest) { try { const body = await request.json(); const { messages }: { messages: Message[] } = body; const lastMessage = messages.pop(); if (!messages || !lastMessage || lastMessage.role !== 'user') { return NextResponse.json( { error: 'messages are required in the request body and the last message must be from the user', }, { status: 400 } ); } const pageLoader = new NotionReader({ client: new Client({ auth: process.env.NOTION_INTEGRATION_TOKEN, }), }); const pageDocs = await pageLoader.loadPages('rootPageId'); const { llm, relevantInfo } = await extractRelevantInfo( pageDocs, lastMessage.content ); const chatEngine = await createChatEngine(llm); const responseText = generateResponse(relevantInfo); const response = await chatEngine.chat(responseText, messages, true); const stream = LlamaIndexStream(response); return new StreamingTextResponse(stream); } catch (error) { console.error('[LlamaIndex]', error); return NextResponse.json( { error: (error as Error).message, }, { status: 500, } ); } } async function extractRelevantInfo( pageDocs: Pages, query: string ): Promise<{ relevantInfo: string; llm: OpenAI; }> { const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY, model: 'gpt-4-1106-preview', temperature: 0.7, maxTokens: 800, }); let combinedContent = ''; for (const pageId in pageDocs) { const page = pageDocs[pageId]; if (page && page.lines) { combinedContent += page.lines.join('\n') + '\n\n'; } } const prompt = `This AI extracts information based on the following content. It cannot answer questions about information that is not related to the Notion page.\n\nContent:\n${combinedContent}\n\nQuestion: ${query}\n\nRelevant information:`; const response = await openai.complete(prompt); return { relevantInfo: response.message.content.trim(), llm: openai, }; } function generateResponse(relevantInfo: string): string { if (relevantInfo) { return relevantInfo; } else { return 'Sorry, I could not find any relevant information.'; } } ```

I understand that this error is due to the Notion API not supporting the ai_block type. However, I am unable to find a way to exclude this block type from the API request or to filter it out after the data is received.

I would appreciate it if you could provide a solution or workaround for this issue. It would be helpful if there was a way to exclude certain block types when using the NotionReader class, or if the class could automatically handle unsupported block types.

Thank you for your assistance.

[himself65]

This looks like an upstream issue but upstream notion-to-md has not updated for a long time.

[himself65]

llama-hub has this reader but it's python

https://github.com/run-llama/llama-hub/blob/2c95b021246b54b0542bf9ed9289828cc9da6654/llama_hub/notion/base.py