Build a PDF Q&A system with LlamaIndex, OpenAI embeddings & Pinecone vector DB

Open on n8n.io

$20/month : Unlimited workflows

2500 executions/month

Try free

THE #1 IN WEB SCRAPING

Scrape any website without limits

Try free

HOSTINGER

Early Deal
DISCOUNT 20%

Self-hosted n8n

Unlimited workflows - from $4.99/mo

Try free

#1 hub for scraping, AI & automation

6000+ actors - $5 credits/mo

Try free

Important notice

This workflow is provided as-is. Please review and test before using in production.

1. Workflow Overview

Parse, Normalize, Extract, and Store PDF Content for RAG in Pinecone This workflow automates a full RAG pipeline for structured documents (like insurance policies). What it does Watches a Google Dr...

Best for

AI RAG automation workflows
Multimodal AI automation workflows
advanced n8n builders looking for reusable templates

Tools used

n8n-nodes-base.googledrivetrigger, n8n-nodes-base.googledrive, @n8n/n8n-nodes-langchain.documentdefaultdataloader, n8n-nodes-base.stickynote, n8n-nodes-base.wait, n8n-nodes-base.if, n8n-nodes-base.httprequest, n8n-nodes-base.code

Source and attribution

This workflow is cataloged by N8N Workflows and links back to its original n8n.io source page by Alok Kumar.

Original n8n.io source

1.1 Workflow description

Title: Build a PDF Q&A system with LlamaIndex, OpenAI embeddings & Pinecone vector DB
Workflow name: Build a PDF Q&A system with LlamaIndex, OpenAI embeddings & Pinecone vector DB

Parse, Normalize, Extract, and Store PDF Content for RAG in Pinecone

This workflow automates a full RAG pipeline for structured documents (like insurance policies).

What it does

Watches a Google Drive folder for new PDFs
Uploads to LlamaIndex Cloud for parsing → returns clean Markdown
Normalizes text (removes headers, footers, page numbers, formatting artifacts)
Splits text into chunks (~1200 chars with 150 overlap)
Generates embeddings with OpenAI
Stores vectors in Pinecone with metadata
Connects a Chat Agent that retrieves answers from Pinecone

Who’s it for

Developers building chatbots or Q&A systems for structured docs
Teams working with insurance, compliance, or legal PDFs
Anyone who needs to normalize & store documents for semantic search

Requirements

Google Drive connected (for source PDFs)
LlamaIndex Cloud account (parsing API key)
Pinecone account (vector DB)
OpenAI account (LLM and embeddings)

How to use and customize

Update the folder name in google drive trigger node.
Place a pdf file in the same folder in google drive.
Customize the Normalized Content function node to adjust regex for headers/footers specific to your documents.
Adjust chunk size or metadata namespace in the Pinecone node to fit your project needs.

1.2 Logical Blocks

This catalog entry is organized from the workflow JSON. The node-level section below shows the executable blocks available for review before importing the template.

2. Block-by-Block Analysis

Block 1 - Google Drive Trigger

Type / Role: n8n-nodes-base.googleDriveTrigger - googleDriveTrigger
Config choices: Version 1

Block 2 - Download file

Type / Role: n8n-nodes-base.googleDrive - googleDrive
Config choices: Version 3

Block 3 - Default Data Loader

Type / Role: @n8n/n8n-nodes-langchain.documentDefaultDataLoader - documentDefaultDataLoader
Config choices: Version 1.1

Block 4 - Sticky Note

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

Block 5 - Sticky Note1

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

Block 6 - Wait

Type / Role: n8n-nodes-base.wait - wait
Config choices: Version 1.1

Block 7 - If

Type / Role: n8n-nodes-base.if - if
Config choices: Version 2.2

Block 8 - Wait2

Type / Role: n8n-nodes-base.wait - wait
Config choices: Version 1.1

Block 9 - Sticky Note2

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

Block 10 - Sticky Note3

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

Block 11 - Upload to Llama Cloud

Type / Role: n8n-nodes-base.httpRequest - httpRequest
Config choices: Version 4.2

Block 12 - Check Parsing Status

Type / Role: n8n-nodes-base.httpRequest - httpRequest
Config choices: Version 4.2

Block 13 - Extract Markdown from Llama Cloud

Type / Role: n8n-nodes-base.httpRequest - httpRequest
Config choices: Version 4.2

Block 14 - Normalize Text

Type / Role: n8n-nodes-base.code - code
Config choices: Version 2

Block 15 - Chunk Text

Type / Role: @n8n/n8n-nodes-langchain.textSplitterRecursiveCharacterTextSplitter - textSplitterRecursiveCharacterTextSplitter
Config choices: Version 1

Block 16 - Generate Embeddings

Type / Role: @n8n/n8n-nodes-langchain.embeddingsOpenAi - embeddingsOpenAi
Config choices: Version 1.2

Block 17 - Store in Pinecone

Type / Role: @n8n/n8n-nodes-langchain.vectorStorePinecone - vectorStorePinecone
Config choices: Version 1.3

Block 18 - Sticky Note4

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

3. Summary Table

Workflow	Build a PDF Q&A system with LlamaIndex, OpenAI embeddings & Pinecone vector DB
Complexity	advanced
Nodes	18
Categories	AI RAG, Multimodal AI
Author	Alok Kumar
Published	22 Aug 2025

4. Reproducing the Workflow from Scratch

1. Download the workflow JSON

Use the JSON export at /data/workflows/7717/7717.json as the source template for this automation.
2. Import the template into n8n

Open n8n, import the downloaded JSON, and review each node before activating the workflow.
3. Configure credentials and variables

Replace placeholder credentials, API keys, webhook URLs, account IDs, and environment-specific values with your own settings.
4. Test with sample data

Run the workflow manually or in a staging workspace, inspect node output, and confirm downstream systems receive the expected data.
5. Activate and monitor

Enable the workflow only after testing, then monitor executions, errors, and rate limits during the first production runs.

5. General Notes & Resources

Review imported nodes carefully before activation. This catalog entry is intended to help you inspect the workflow structure, understand required services, and find related templates faster.

Node names, credentials, schedules, webhook paths, and external service limits may need adjustment for your workspace.

Download workflow JSON Original n8n.io source AI RAG workflows Multimodal AI workflows

Frequently asked questions

What does Build a PDF Q&A system with LlamaIndex, OpenAI embeddings & Pinecone vector DB do?

What do I need before importing this workflow?

Review the workflow JSON, configure any required credentials in n8n, and test the automation in a safe workspace before using it in production.

Can I customize this workflow?

Yes. Use the block-by-block analysis and the downloadable JSON to inspect each node, then adjust credentials, prompts, schedules, filters, or destinations for your AI RAG, Multimodal AI use case.

Alok Kumar

6 workflows

Nodes

n8n-nodes-base.googledrivetrigger n8n-nodes-base.googledrive @n8n/n8n-nodes-langchain.documentdefaultdataloader n8n-nodes-base.stickynote n8n-nodes-base.wait n8n-nodes-base.if n8n-nodes-base.httprequest n8n-nodes-base.code

Complexity

advanced

Published 22 Aug 2025

Likes 0

View on n8n.io Download Workflow

Install path: /data/workflows/7717/7717.json

Share Your Workflow

Have a useful automation to share? Publish it and help the community.

Submit Your Template How to Submit

Related Workflows

Route RAG customer support across WhatsApp and Outlook with Pinecone and OpenRouter

## Quick overview A production-ready 3-workflow system that handles customer support across WhatsApp and Email using RAG-powered AI. Automatically routes queries, detects escalation intent, logs handoffs to Google Sheets, alerts your team via Slack, and lets agents resolve cases with a single slash command. ## How it works 1. A form trigger accepts PDF, DOCX, and TXT documents, chunks them with a token splitter, embeds them using Google Gemini, and stores them in your Pinecone vector store to build the AI knowledge base. 2. Incoming messages from WhatsApp Business Cloud API and Microsoft Outlook are each normalized into a unified contact format — capturing channel, contact ID, name, and message text — before merging into a single pipeline. 3. Every message is checked against a Google Sheets handoff tracker. If the contact is currently assigned to a human agent, the message is forwarded directly to Slack without any AI involvement. 4. For active (non-handoff) contacts, the AI Support Agent queries Pinecone via RAG, pulls context from a sliding conversation memory window, and returns a structured JSON output containing a reply and a handoff_requested flag. 5. If the AI detects escalation intent or has no answer, it sets handoff_requested: true — the contact is logged to Google Sheets, AI goes silent for that contact, and a detailed alert is posted to your Slack support channel. 6. Answers are routed back through the originating channel — WhatsApp replies via Business Cloud API, email replies via Microsoft Outlook — keeping every conversation native to its platform. 7. Support agents run /resolve [contact_id] in Slack to close a handoff. The workflow validates the input, updates the Google Sheets row with agent name and timestamp, re-enables AI for that contact, and confirms resolution in the Slack channel. ## Setup 1. Pinecone + Embeddings: Create a Pinecone index and add your Pinecone credential to the "Insert into Vector Store" and "Knowledge Base" nodes. Set your Google Gemini API credential on both Embeddings nodes (one in Workflow 1, one in Workflow 2). 2. WhatsApp: Connect your WhatsApp Business Cloud API credential to the WhatsApp Trigger node. Replace YOUR_WHATSAPP_PHONE_NUMBER_ID in both Send WhatsApp Message nodes. 3. Email: Add your Microsoft Outlook OAuth2 credential to the Outlook Trigger and the Send Email Answer node. 4. Handoff Sheet: Create a Google Sheet with a tab named Handoffs and these column headers exactly: contact_id | channel | contact_name | status | handoff_since | resolved_by | resolved_at. Set your Google Sheets credential and paste your Sheet ID into all four Google Sheets nodes across Workflows 2 and 3. 5. Slack: Create a Slack bot app, add it to your support channel, and set the credential in all three Slack nodes. Update the channel name if yours differs from #support-handoff. 6. AI Model: Add your OpenRouter (or OpenAI) API key to the Chat Model node. Swap the model slug in node settings as needed. 7. Slash Command: Activate Workflow 3 first to generate its production webhook URL. Then in your Slack app settings go to Slash Commands → Create New → set command to /resolve → paste the webhook URL → reinstall app to workspace. 8. Activate order: Workflow 1 (KB ingestion) → Workflow 3 (resolve handler) → Workflow 2 (main support flow). ## Requirements - n8n (self-hosted or cloud) - WhatsApp Business Cloud API account (Meta Developer App) - Microsoft 365 / Outlook account - Pinecone account - Google Gemini API key - OpenRouter or OpenAI API key or Any other LLM - Google Sheets - Slack workspace with a custom bot app ## Customization - Edit the system prompt in the AI Support Agent node to match your brand voice, support scope, and escalation rules - Swap Google Gemini Embeddings for OpenAI Embeddings — just replace the Embeddings node and ensure your Pinecone index dimensions match - dd more channels (Telegram, Instagram DM, etc.) by normalizing their payloads into the same contact format and feeding them into the Merge node - Adjust conversation memory window (default: 8 messages) in the Conversation Memory node - Trigger the resolve webhook from your CRM or helpdesk instead of Slack if preferred ## Additional info All three workflows must be active simultaneously for the system to function Handoff state is persisted in Google Sheets and survives n8n restarts or server reboots Once a contact enters handoff, the AI is completely silent for that contact — no accidental double-responses Document metadata (filename, uploader name, timestamp) is stored alongside vectors in Pinecone for full traceability Default chunk settings: 500 tokens / 100 overlap — tune in the Token Splitter node based on your document style Contact / Support 🔗 n8n Templates: n8n.io/creators/salmanmehboob 💼 LinkedIn: linkedin.com/in/salman-mehboob-pro 📧 Email: [email protected] 💬 WhatsApp: +92 306 746 6499

View

Detect semantic duplicate website pages with Google Drive, Postgres and Ollama

## Quick overview This workflow scans HTML files in a Google Drive folder, extracts and stores page text in Postgres, generates local vector embeddings with Ollama, and uses PGVector similarity searches to produce CSV reports that flag semantically duplicate website pages. ## How it works 1. Starts manually and clears the existing PGVector embeddings table and the scraped page text table in Postgres. 2. Lists files in a specified Google Drive folder, filters to the target documents, and processes them in batches. 3. Downloads each HTML file from Google Drive, extracts the main body text, cleans it, and upserts the results into a Postgres table for scraped pages. 4. Reads the scraped page text back from Postgres in batches, splits it into overlapping chunks, and attaches page metadata (sheet_id, file_name, file_url) to each chunk. 5. Generates embeddings locally with Ollama and inserts the chunk vectors and metadata into Postgres (PGVector), deduplicating already-processed pages. 6. Builds an HNSW index in Postgres, computes chunk-to-chunk similarity matches and a pairwise page report, and exports the results as a CSV file. 7. Computes page-level centroid embeddings, finds highly similar page pairs, and exports a page-level duplicate report as a CSV file. ## Setup 1. Add Google Drive OAuth2 credentials and set the Google Drive folder URL/ID used to scan for your HTML files. 2. Add Postgres credentials for a database with the pgvector extension enabled and permissions to create/alter tables and indexes (including HNSW indexes). 3. Add an Ollama credential and ensure the embedding model mxbai-embed-large:latest is available on your Ollama instance. 4. Confirm your source files are HTML documents and that the workflow’s text extraction and similarity thresholds match your content and desired duplicate sensitivity. ## Requirements - Working instance of n8n, either self-hosted or on the cloud. Remember, this workflow can be computationally expensive. - Google Drive API (with OAuth setup in n8n credentials section) - Ollama (for open source models) or any Embedding model API - PostgreSQL with PGVector or any other vector database - PgAdmin (for PostgreSQL) or your interface to access database tables via SQL for troubleshooting (optional). ## Additional info Limitations and Enhancements: Physical system memory mxbai-embed-large Running through Ollama is free and private, but the embedding generation speed depends entirely on your hardware. The more system memory you have, the more data you can process in batches in the loop node. Similarity threshold and boilerplate content The cosine distance used in this workflow is 0.15 for chunk-level matching. And 0.05 (similarity above 95%) of the threshold is used for page-level centroid matching. This is only the starting point. Once you have the data, and especially if your data has more noise, you might need to tweak these thresholds for better matching. This workflow needs HTML files to extract text This workflow doesn't crawl a website or fetch pages by entering a URL. You need to download HTML files (rendered or source) for consumption. Use parallel processing and Cloud APIs Two sub-processes take the most time: Downloading HTML files from Google Drive Creating vector embeddings If you can use parallel processing in n8n and execute these sub-processes in parallel, the process will be done much faster. Additionally, if you can use cloud APIs for embedding, it may save some you some processing time as well. Use efficient SQL queries Since I am from a non-tech background and not a coder, I used a mix of Gemini, Perplexity and Claude to create SQL codes for this workflow. If you're better at it, you can run computationally efficient queries that would help you achieve better results with less computation expense and time.

View

Answer BigQuery questions in Microsoft Teams with Claude 3.5 dual agents

## Quick overview Turn Microsoft Teams into a natural language BI tool. This dual-agent Claude workflow intelligently classifies user questions, generates safe BigQuery SQL, executes the query, and returns a plain-English summary alongside a data preview directly to your Teams channel. ## How it works 1. The workflow listens for new messages in a designated Microsoft Teams channel or chat. 2. It checks a 24-hour static data cache for your BigQuery schema. If the cache is missing or expired, it automatically fetches the latest INFORMATION_SCHEMA metadata. 3. A main Claude 3.5 Sonnet agent classifies the user's intent (analytical, conversational, or unsafe) and identifies the relevant database tables. 4. If the intent is analytical, a specialized Claude SQL sub-agent generates a read-only SELECT query tailored exactly to your schema. 5. The workflow executes the query against BigQuery (safely constrained to a maximum of 100 rows). 6. A final Claude agent translates the raw query results into a concise, plain-English summary. 7. The answer is posted back to Teams as an Adaptive Card, complete with a formatted data preview and the generated SQL code. 8. Non-analytical questions or pipeline errors are intelligently routed to receive standard conversational replies in Teams. ## Setup 1. CRITICAL SECURITY: Restrict your Google Cloud Service Account IAM roles to BigQuery Data Viewer and BigQuery Job User ONLY. Do not use a highly privileged account. 2. Configure your Microsoft Teams trigger and posting credentials, specifying the target team and channel for the bot to operate in. 3. Configure your Google BigQuery credentials to allow reading metadata and executing queries. 4. Configure your Anthropic credentials for the three Claude LangChain nodes. 5. Update the "Configure Variables" node with your specific GCP Project ID and BigQuery Dataset ID. ## Requirements - A Google Cloud project with BigQuery enabled and existing structured datasets. - An Anthropic API key with access to the Claude 3.5 Sonnet model. - Microsoft Teams administrator access to authorize the n8n integration. ## Customization - Distributed Environments: If you are running n8n with multiple worker nodes (Queue Mode), you must swap the $getWorkflowStaticData() schema cache logic to read/write from a dedicated Key-Value store (like Redis or Supabase) to ensure state synchronization. - UI Layout: Modify the Create Teams Adaptive Card node to add corporate branding, custom colors, or alter the FactSet data preview structure. ## Additional info This architecture features built-in pipeline protections, including strict 40,000-character payload limits to prevent LLM context window blowouts, and fallback handling for "zero-row" query results to ensure users are never left without a response.

View

Need Custom Automation?

Get help designing a custom n8n workflow that connects your stack and fits your process.

Build a PDF Q&A system with LlamaIndex, OpenAI embeddings & Pinecone vector DB

Workflow preview

Important notice

1. Workflow Overview

Best for

Tools used

Source and attribution

1.1 Workflow description

Parse, Normalize, Extract, and Store PDF Content for RAG in Pinecone

What it does

Who’s it for

Requirements

How to use and customize

1.2 Logical Blocks

2. Block-by-Block Analysis

Block 1 - Google Drive Trigger

Block 2 - Download file

Block 3 - Default Data Loader

Block 4 - Sticky Note

Block 5 - Sticky Note1

Block 6 - Wait

Block 7 - If

Block 8 - Wait2

Block 9 - Sticky Note2

Block 10 - Sticky Note3

Block 11 - Upload to Llama Cloud

Block 12 - Check Parsing Status

Block 13 - Extract Markdown from Llama Cloud

Block 14 - Normalize Text

Block 15 - Chunk Text

Block 16 - Generate Embeddings

Block 17 - Store in Pinecone

Block 18 - Sticky Note4

3. Summary Table

4. Reproducing the Workflow from Scratch

1. Download the workflow JSON

2. Import the template into n8n

3. Configure credentials and variables

4. Test with sample data

5. Activate and monitor

5. General Notes & Resources

Frequently asked questions