klbr/n8nworkflows.xyz
1
0
Fork 0
mirror of https://github.com/khoaliber/n8nworkflows.xyz.git synced 2026-04-28 00:29:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

creation

Browse Source
This commit is contained in:
nusquama
2025-11-13 14:08:43 +01:00
parent 9c93aacd8c
commit d5edcdd95c
1 changed files with 288 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
workflows/Analyse papers from Hugging Face with AI and store them in Notion-2804/readme-2804.md
+288
View File
@@ -0,0 +1,288 @@
Analyse papers from Hugging Face with AI and store them in Notion
https://n8nworkflows.xyz/workflows/analyse-papers-from-hugging-face-with-ai-and-store-them-in-notion-2804
# Analyse papers from Hugging Face with AI and store them in Notion
### 1. Workflow Overview
This workflow automates the retrieval, analysis, and storage of academic paper summaries from Hugging Face into Notion. It is designed to run automatically on weekdays at 8 AM, fetching the latest papers, analyzing their abstracts with OpenAI’s GPT model, and saving structured insights into a Notion database. The workflow is logically divided into the following blocks:
- **1.1 Scheduled Trigger:** Initiates the workflow automatically on weekdays at 8 AM.
- **1.2 Fetching Paper Data:** Retrieves the list of recent papers from Hugging Face using their API.
- **1.3 Extracting Paper URLs:** Parses the HTML response to extract individual paper URLs.
- **1.4 Looping Over Papers:** Processes each paper URL one by one.
- **1.5 Duplicate Check in Notion:** Checks if the paper already exists in the Notion database to avoid duplicates.
- **1.6 Fetching Paper Details:** For new papers, fetches detailed HTML content of the paper page.
- **1.7 Extracting Paper Abstract and Title:** Extracts the abstract and title from the paper’s HTML.
- **1.8 OpenAI Content Analysis:** Sends the abstract to OpenAI GPT-4o for detailed analysis and structured extraction.
- **1.9 Storing Results in Notion:** Saves the analyzed data and metadata into the Notion database.
---
### 2. Block-by-Block Analysis
#### 2.1 Scheduled Trigger
- **Overview:** Automatically triggers the workflow at 8 AM on weekdays (Monday to Friday).
- **Nodes Involved:** `Schedule Trigger`
- **Node Details:**
- Type: Schedule Trigger (Cron-like)
- Configuration: Runs weekly on days 1-5 (Mon-Fri) at hour 8.
- Inputs: None (trigger node)
- Outputs: Initiates the HTTP request to fetch Hugging Face papers.
- Edge Cases: Misconfiguration could cause missed runs or runs on weekends.
- Version: 1.2
#### 2.2 Fetching Paper Data
- **Overview:** Sends an HTTP GET request to Hugging Face’s papers endpoint, querying papers from the previous day.
- **Nodes Involved:** `Request Hugging Face Paper`
- **Node Details:**
- Type: HTTP Request
- Configuration: URL `https://huggingface.co/papers` with query parameter `date` set dynamically to yesterday’s date in `yyyy-MM-dd` format.
- Inputs: Trigger from Schedule Trigger
- Outputs: Raw HTML response containing paper listings.
- Edge Cases: API downtime, rate limiting, or changes in Hugging Face’s API/HTML structure.
- Version: 4.2
#### 2.3 Extracting Paper URLs
- **Overview:** Parses the HTML response to extract URLs of individual papers using CSS selectors.
- **Nodes Involved:** `Extract Hugging Face Paper`
- **Node Details:**
- Type: HTML Extract
- Configuration: Extracts `href` attributes from elements matching `.line-clamp-3` CSS selector, returning an array of URLs.
- Inputs: Output of HTTP Request node
- Outputs: List of paper URLs for further processing.
- Edge Cases: Changes in HTML structure or CSS classes could break extraction.
- Version: 1.2
#### 2.4 Splitting and Looping Over Papers
- **Overview:** Splits the list of paper URLs into individual items and processes them sequentially.
- **Nodes Involved:** `Split Out`, `Loop Over Items`
- **Node Details:**
- `Split Out`:
- Type: Split Out
- Configuration: Splits out the `url` field from the extracted URLs.
- Inputs: Extracted URLs
- Outputs: Individual paper URLs
- Version: 1
- `Loop Over Items`:
- Type: Split In Batches
- Configuration: Processes one paper URL at a time without resetting.
- Inputs: Output from Split Out
- Outputs: Single paper URL per iteration
- Edge Cases: Large number of papers could slow processing.
- Version: 3
#### 2.5 Duplicate Check in Notion
- **Overview:** Checks if the current paper URL already exists in the Notion database to avoid duplicate entries.
- **Nodes Involved:** `Check Paper URL Existed`, `If`
- **Node Details:**
- `Check Paper URL Existed`:
- Type: Notion (Get All Pages)
- Configuration: Queries Notion database filtering pages where the `URL` property equals the full Hugging Face paper URL.
- Inputs: Single paper URL from Loop Over Items
- Outputs: List of matching pages (empty if none)
- Credentials: Notion API token required
- Edge Cases: Notion API rate limits, incorrect database ID or property keys.
- Version: 2.2
- `If`:
- Type: Conditional
- Configuration: Checks if the output from Notion query is empty (paper not found).
- Inputs: Output from Notion check
- Outputs: If true (paper new) proceeds to fetch details; if false loops to next paper.
- Edge Cases: Expression errors if data format changes.
- Version: 2.2
#### 2.6 Fetching Paper Details
- **Overview:** For new papers, fetches the detailed HTML page of the paper from Hugging Face.
- **Nodes Involved:** `Request Hugging Face Paper Detail`
- **Node Details:**
- Type: HTTP Request
- Configuration: URL dynamically constructed by concatenating `https://huggingface.co` with the paper URL.
- Inputs: Paper URL from If node (new papers only)
- Outputs: HTML content of the paper detail page.
- Edge Cases: Network errors, page not found, or HTML structure changes.
- Version: 4.2
#### 2.7 Extracting Paper Abstract and Title
- **Overview:** Extracts the abstract and title text from the paper’s HTML detail page.
- **Nodes Involved:** `Extract Hugging Face Paper Abstract`
- **Node Details:**
- Type: HTML Extract
- Configuration: Extracts text content from `.text-gray-700` (abstract) and `.text-2xl` (title) CSS selectors.
- Inputs: HTML content from paper detail request
- Outputs: JSON with `abstract` and `title` fields.
- Edge Cases: Changes in CSS selectors or page layout.
- Version: 1.2
#### 2.8 OpenAI Content Analysis
- **Overview:** Sends the extracted abstract to OpenAI GPT-4o for detailed structured analysis, extracting key insights in JSON format.
- **Nodes Involved:** `OpenAI Analysis Abstract`
- **Node Details:**
- Type: OpenAI (Langchain node)
- Configuration:
- Model: GPT-4o-2024-11-20
- System prompt instructs the model to extract:
- Core Introduction
- Keywords (2-5)
- Key Data and Results
- Technical Details
- Classification (academic categories)
- User prompt: The paper abstract text.
- Output: JSON with structured fields.
- Inputs: Abstract text from HTML extraction
- Outputs: JSON with analyzed content
- Credentials: OpenAI API key required
- Edge Cases: API rate limits, prompt failures, or unexpected output format.
- Version: 1.8
#### 2.9 Storing Results in Notion
- **Overview:** Saves the analyzed paper data and metadata into a Notion database for future reference.
- **Nodes Involved:** `Store Abstract Notion`
- **Node Details:**
- Type: Notion (Create Database Page)
- Configuration:
- Database ID: Specified Notion database for Hugging Face abstracts.
- Properties mapped:
- URL: Full Hugging Face paper URL
- Title: Extracted paper title
- Abstract: Abstract text truncated to 2000 characters
- Scrap Date: Current date (no time)
- Classification, Technical Details, Data and Results, Keywords, Core Introduction: From OpenAI JSON output
- Inputs: Output from OpenAI analysis node
- Credentials: Notion API token required
- Edge Cases: API limits, property mismatches, data truncation issues.
- Version: 2.2
---
### 3. Summary Table
| Node Name | Node Type | Functional Role | Input Node(s) | Output Node(s) | Sticky Note |
|-----------------------------|-------------------------------|------------------------------------|---------------------------------|---------------------------------|---------------------------------------------------------------------------------------------------|
| Schedule Trigger | Schedule Trigger | Initiates workflow on weekdays 8AM | None | Request Hugging Face Paper | |
| Request Hugging Face Paper | HTTP Request | Fetches paper list HTML from API | Schedule Trigger | Extract Hugging Face Paper | |
| Extract Hugging Face Paper | HTML Extract | Extracts paper URLs from HTML | Request Hugging Face Paper | Split Out | |
| Split Out | Split Out | Splits URLs into individual items | Extract Hugging Face Paper | Loop Over Items | |
| Loop Over Items | Split In Batches | Processes each paper URL sequentially | Split Out | Check Paper URL Existed | |
| Check Paper URL Existed | Notion (Get All Pages) | Checks if paper URL exists in Notion | Loop Over Items | If | |
| If | If | Conditional: proceed if paper new | Check Paper URL Existed | Request Hugging Face Paper Detail (true), Loop Over Items (false) | |
| Request Hugging Face Paper Detail | HTTP Request | Fetches detailed paper HTML page | If (true branch) | Extract Hugging Face Paper Abstract | |
| Extract Hugging Face Paper Abstract | HTML Extract | Extracts abstract and title from HTML | Request Hugging Face Paper Detail | OpenAI Analysis Abstract | |
| OpenAI Analysis Abstract | OpenAI (Langchain) | Analyzes abstract with GPT-4o | Extract Hugging Face Paper Abstract | Store Abstract Notion | |
| Store Abstract Notion | Notion (Create Database Page) | Stores analyzed data in Notion | OpenAI Analysis Abstract | Loop Over Items | |
---
### 4. Reproducing the Workflow from Scratch
1. **Create a Schedule Trigger node:**
- Type: Schedule Trigger
- Set to run weekly on Monday to Friday at 8 AM.
- No credentials needed.
2. **Add an HTTP Request node named "Request Hugging Face Paper":**
- URL: `https://huggingface.co/papers`
- Method: GET
- Query Parameter: `date` = `={{ $now.minus(1,'days').format('yyyy-MM-dd') }}`
- Connect Schedule Trigger output to this node.
3. **Add an HTML Extract node named "Extract Hugging Face Paper":**
- Operation: Extract HTML content
- Extraction Values:
- Key: `url`
- CSS Selector: `.line-clamp-3`
- Attribute: `href`
- Return Array: true
- Connect "Request Hugging Face Paper" output to this node.
4. **Add a Split Out node named "Split Out":**
- Field to split out: `url`
- Connect "Extract Hugging Face Paper" output to this node.
5. **Add a Split In Batches node named "Loop Over Items":**
- Default batch size (process one item at a time)
- Connect "Split Out" output to this node.
6. **Add a Notion node named "Check Paper URL Existed":**
- Resource: Database Page
- Operation: Get All
- Database ID: Your Notion database ID for storing papers
- Filter: Property `URL` equals `={{ 'https://huggingface.co'+$json.url }}`
- Credentials: Configure Notion API credentials
- Connect "Loop Over Items" output to this node.
7. **Add an If node named "If":**
- Condition: Check if the output from "Check Paper URL Existed" is empty (no existing page)
- Expression: `={{ $json.length === 0 }}`
- Connect "Check Paper URL Existed" output to this node.
8. **Add an HTTP Request node named "Request Hugging Face Paper Detail":**
- URL: `={{ 'https://huggingface.co'+$('Split Out').item.json.url }}`
- Method: GET
- Connect "If" node’s true output to this node.
9. **Add an HTML Extract node named "Extract Hugging Face Paper Abstract":**
- Operation: Extract HTML content
- Extraction Values:
- Key: `abstract`, CSS Selector: `.text-gray-700`
- Key: `title`, CSS Selector: `.text-2xl`
- Connect "Request Hugging Face Paper Detail" output to this node.
10. **Add an OpenAI node named "OpenAI Analysis Abstract":**
- Model: GPT-4o-2024-11-20 (or latest GPT-4 variant)
- System Prompt: Provide instructions to extract Core Introduction, Keywords, Data and Results, Technical Details, Classification in JSON format.
- User Prompt: `={{ $json.abstract }}`
- Enable JSON output parsing.
- Credentials: Configure OpenAI API credentials.
- Connect "Extract Hugging Face Paper Abstract" output to this node.
11. **Add a Notion node named "Store Abstract Notion":**
- Resource: Database Page
- Operation: Create
- Database ID: Same as in step 6
- Map properties:
- URL: `={{ 'https://huggingface.co'+$('Split Out').item.json.url }}`
- Title: `={{ $('Extract Hugging Face Paper Abstract').item.json.title }}`
- Abstract: `={{ $('Extract Hugging Face Paper Abstract').item.json.abstract.substring(0,2000) }}`
- Scrap Date: `={{ $today.format('yyyy-MM-dd') }}`
- Classification: `={{ $json.message.content.Classification.join(',') }}`
- Technical Details: `={{ $json.message.content.Technical_Details }}`
- Data and Results: `={{ $json.message.content.Data_and_Results }}`
- Keywords: `={{ $json.message.content.Keywords.join(',') }}`
- Core Introduction: `={{ $json.message.content.Core_Introduction }}`
- Credentials: Notion API credentials
- Connect "OpenAI Analysis Abstract" output to this node.
12. **Connect the false output of the "If" node back to "Loop Over Items" to continue processing remaining papers.**
13. **Ensure all credentials (OpenAI, Notion) are properly set up in n8n.**
14. **Activate the workflow.**
---
### 5. General Notes & Resources
| Note Content | Context or Link |
|---------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------|
| The workflow uses GPT-4o-2024-11-20, a GPT-4 variant optimized for academic and technical content analysis. | OpenAI model selection in the OpenAI node. |
| Notion database must have properties matching the keys used in the workflow (URL, title, abstract, classification, etc.) | Notion database schema setup. |
| Hugging Face papers page HTML structure is critical for CSS selectors `.line-clamp-3`, `.text-gray-700`, `.text-2xl` | Changes in Hugging Face website may require updates to HTML extraction nodes. |
| Scheduled trigger uses n8n’s built-in cron-like scheduling for weekday runs at 8 AM. | n8n Schedule Trigger documentation: https://docs.n8n.io/nodes/n8n-nodes-base.scheduleTrigger/ |
| OpenAI prompt is designed to output JSON for easy parsing and storage. | Prompt engineering is key to consistent output format. |
| Image illustrating workflow result is referenced as `huggingface.png` (fileId:918) | Visual asset for documentation or UI display. |
---
This document fully describes the workflow structure, node configurations, and setup instructions to enable reproduction, modification, and troubleshooting by advanced users or automation agents.