Workflow Templates

This workflow automates the process of analyzing a contract submitted via a web form. It extracts the text from an uploaded PDF, uses AI to identify potential red flags, and sends a summary report to a Telegram chat. Prerequisites Before you can use this workflow, you'll need a few things set up. 1. JotForm Form You need to create a form in JotForm with at least two specific fields: Email Address**: A standard field to collect the user's email. File Upload**: This field will be used to upload the contract or NDA. Make sure to configure it to allow .pdf files. 2. API Keys and IDs JotForm API Key**: You can generate this from your JotForm account settings under the "API" section. Gemini API Key**: You'll need an API key from Google AI Studio to use the Gemini model. Telegram Bot Token**: Create a new bot by talking to the @BotFather on Telegram. It will give you a unique token. Telegram Chat ID**: This is the ID of the user, group, or channel you want the bot to send messages to. You can get this by using a bot like @userinfobot. Node-by-Node Explanation Here is a breakdown of what each node in the workflow does, in the order they execute. 1. JotForm Trigger What it does**: This node kicks off the entire workflow. It actively listens for new submissions on the specific JotForm you select. How it works**: When someone fills out your form and hits "Submit," JotForm sends the submission data (including the email and a link to the uploaded file) to this node. 2. Grab Attachment Details (HTTP Request) What it does**: The initial data from JotForm doesn't contain a direct download link for the file. This node takes the submissionID from the trigger and makes a request to the JotForm API to get the full details of that submission. How it works**: It constructs a URL using the submissionID and your JotForm API key to fetch the submission data, which includes the proper download URL for the uploaded contract. 3. Grab the Attached Contract (HTTP Request) What it does**: Now that it has the direct download link, this node fetches the actual PDF file. How it works**: It uses the file URL obtained from the previous node to download the contract. The node is set to expect a "file" as the response, so it saves the PDF data in binary format for the next step. 4. Extract Text from PDF File What it does**: This node takes the binary PDF data from the previous step and extracts all the readable text from it. How it works**: It processes the PDF and outputs plain text, stripping away any formatting or images. This raw text is now ready to be analyzed by the AI. 5. AI Agent (with Google Gemini Chat Model) What it does**: This is the core analysis engine of the workflow. It takes the extracted text from the PDF and uses a powerful prompt to analyze it. The "Google Gemini Chat Model" node is connected as its "brain." How it works**: It sends the contract text to the Gemini model. The prompt instructs Gemini to act as an expert contract analyst. It specifically asks the AI to identify major red flags and hidden/unfair clauses. It also tells the AI to format the output as a clean report using Telegram's MarkdownV2 style and to keep the response under 1500 characters. 6. Send a text message (Telegram) What it does**: This is the final step. It takes the formatted analysis report generated by the AI Agent and sends it to your specified Telegram chat. How it works**: It connects to your Telegram bot using your Bot Token and sends the AI's output ($json.output) to the Chat ID you've provided. Because the AI was instructed to format the text in MarkdownV2, the message will appear well-structured in Telegram with bolding and bullet points.

Voice Assistant Interface with n8n and OpenAI This workflow creates a voice-activated AI assistant interface that runs directly in your browser. Users can click on a glowing orb to speak with the AI, which responds with voice using OpenAI's text-to-speech capabilities. Who is it for? This template is perfect for: Developers looking to add voice interfaces to their applications Customer service teams wanting to create voice-enabled support systems Content creators building interactive voice experiences Anyone interested in creating their own "Alexa-like" assistant How it works The workflow consists of two main parts: Frontend Interface: A beautiful animated orb that users click to activate voice recording Backend Processing: Receives the audio transcription, processes it through an AI agent with memory, and returns voice responses The system uses: Web Speech API for voice recognition (browser-based) OpenAI GPT-4o-mini for intelligent responses OpenAI Text-to-Speech for voice synthesis Session memory to maintain conversation context Setup requirements n8n instance (self-hosted or cloud) OpenAI API key with access to: GPT-4o-mini model Text-to-Speech API Modern web browser with Web Speech API support (Chrome, Edge, Safari) How to set up Import the workflow into your n8n instance Add your OpenAI credentials to both OpenAI nodes Copy the webhook URL from the "Audio Processing Endpoint" node Edit the "Voice Assistant UI" node and replace YOUR_WEBHOOK_URL_HERE with your webhook URL Access the "Voice Interface Endpoint" webhook URL in your browser Click the orb and start talking! How to customize the workflow Change the AI personality**: Edit the system message in the "Process User Query" node Modify the visual style**: Customize the CSS in the "Voice Assistant UI" node Add more capabilities**: Connect additional tools to the AI Agent Change the voice**: Select a different voice in the "Generate Voice Response" node Adjust memory**: Modify the context window length in the "Conversation Memory" node Demo Watch the template in action: https://youtu.be/0bMdJcRMnZY

How it works A complete AI-powered study assistant system that lets you chat naturally with your documents stored in Google Drive: The system has two connected workflows: 1. Document Indexing Pipeline (Sub-workflow): • Accepts Google Drive folder URLs • Automatically fetches all files from the folder • Converts documents to plain text • Generates 768-dimensional embeddings using Google Gemini • Stores everything in Supabase vector database for semantic search 2. Study Chat Agent (Main workflow): • Provides a conversational chat interface • Automatically detects and processes Google Drive links shared in chat • Searches your indexed documents using semantic similarity • Maintains conversation history across sessions • Includes calculator for math problems • Responds naturally using Google Gemini 2.5 Pro Use Cases: Students studying for exams, researchers managing papers, professionals building knowledge bases, anyone needing to query large document collections conversationally. Set up steps Prerequisites: • Google Drive OAuth2 credentials • Google Gemini API key (free tier available) • Supabase account with Postgres connection • ~15 minutes setup time Complete Setup: Part 1: Document Indexing Workflow Add Google Drive OAuth2 credentials to the Drive nodes Configure Supabase Postgres credentials in the SQL node Add Supabase API credentials to the Vector Store node Add Google Gemini API key to the Embeddings node Part 2: Study Agent Workflow Import the Study Agent workflow Verify the "Folder all file to vector" tool links to the indexing workflow Add Google Gemini API credentials to both Gemini nodes Configure Supabase API credentials in the Vector Store node Add Postgres credentials for Chat Memory Deploy and access the chat via webhook URL How to Use: Open the chat interface (webhook URL) Paste a Google Drive folder link in the chat Wait for indexing to complete (~1-2 minutes) Start asking questions about your documents The AI will search and answer from your materials Note: The indexing workflow runs automatically when you share Drive links in chat, or you can run it manually to pre-load documents. System Components: Main Agent:** Gemini 2.5 Pro with conversational AI Vector Search:** Supabase with pgvector (768-dim embeddings) Memory:** Postgres chat history (10-message context window) Tools:** Document retrieval, Drive indexing, calculator Embedding Model:** Google Gemini text-embedding-004

Template Description This description details the template's purpose, how it works, and its key features. You can copy and use it directly. Overview This is a powerful n8n "meta-workflow" that acts as a Supervisor. Through a simple Telegram bot, you can dynamically create, manage, and delete countless independent, AI-driven market monitoring agents (Watchdogs). This template is a perfect implementation of the "Workflowception" (workflow managing workflows) concept in n8n, showcasing how to achieve ultimate automation by leveraging the the n8n API. How It Works ? Telegram Bot Interface: Execute all operations by sending commands to your own Telegram Bot: /add SYMBOL INTERVAL PROMPT: Add a new monitoring task. /delete SYMBOL: Delete an existing monitoring task. /list: List all currently running monitoring tasks. /help: Get help information. Use Telegram Bot to control The watchdog workfolw created in the below Dynamic Workflow Management: Upon receiving an /add command, the Supervisor system reads a "Watchdog" template, fills in your provided parameters (like trading pair and time interval), and then automatically creates a brand new, independent workflow via the n8n API and activates it. Persistent Storage: All monitoring tasks are stored in a PostgreSQL database, ensuring your configurations are safe even if n8n restarts. The ID of each newly created workflow is also written back to the database to facilitate future deletion operations. AI-Powered Analysis: Each created "Watchdog" workflow runs on schedule. It fetches the latest candlestick chart by calling a self-hosted tradingview-snapshot service. This service, available at https://github.com/0xcathiefish/tradingview-snapshot, works by simulating a login to your account and then using TradingView's official snapshot feature to generate an unrestricted, high-quality chart image. An example of a generated snapshot can be seen here: https://s3.tradingview.com/snapshots/u/uvxylM1Z.png. To use this, you need to download the Docker image from the packages in the GitHub repository mentioned above, and run it as a container. The n8n workflow then communicates directly with this container via an HTTP API to request and receive the chart snapshot. After obtaining the image, the workflow calls a multimodal AI model (Gemini). It sends both the chart image and your custom text-based conditions (e.g., "breakout above previous high on high volume" or "break below 4-hour MA20") to the AI for analysis, enabling truly intelligent chart interpretation and alert triggering. Key Features Workflowception: A prime example of one workflow using an API to create, activate, and delete other workflows. Full Control via Telegram: Manage your monitoring bots from anywhere, anytime, without needing to log into the n8n interface. AI Visual Analysis: Move beyond simple price alerts. Let an AI "read" the charts for you to enable complex, pattern-based, and indicator-based intelligent alerts. Persistent & Extensible: Built on PostgreSQL for stability and reliability. You can easily add more custom commands.

Description This workflow vectorizes the TUSS (Terminologia Unificada da Saúde Suplementar) table by transforming medical procedures into vector embeddings ready for semantic search. It automates the import of TUSS data, performs text preprocessing, and uses Google Gemini to generate vector embeddings. The resulting vectors can be stored in a vector database, such as PostgreSQL with pgvector, enabling efficient semantic queries across healthcare data. What Problem Does This Solve? Searching for medical procedures using traditional keyword matching is often imprecise. This workflow enhances the search experience by enabling semantic similarity search, which can retrieve more relevant results based on the meaning of the query instead of exact word matches. How It Works Import TUSS data: Load medical procedure entries from the TUSS table. Preprocess text: Clean and prepare the text for embedding. Generate embeddings: Use Google Gemini to convert each procedure into a semantic vector. Store vectors: Save the output in a PostgreSQL database with the pgvector extension. Prerequisites An n8n instance (self-hosted). A PostgreSQL database with the pgvector extension enabled. Access to the Google Gemini API. TUSS data in a structured format (CSV, database, or API source). Customization Tips You can adapt the preprocessing logic to your own language or domain-specific terms. Swap Google Gemini with another embedding model, such as OpenAI or Cohere. Adjust the chunking logic to control the granularity of semantic representation. Setup Instructions Prepare a source (database or CSV) with TUSS data. You need at least two fields: CD_ITEM (Medical procedure code) DS_ITEM (Medical procedure description) Configure your Oracle or PostgreSQL database credentials in the Credentials section of n8n. Make sure your PostgreSQL database has pgVector installed. Replace the placeholder table and column names with your actual TUSS table. Connect your Google Gemini credentials (via OpenAI proxy or official connector). Run the workflow to vectorize all medical procedure descriptions.

BigQuery RAG with OpenAI Embeddings This workflow demonstrates how to use Retrieval-Augmented Generation (RAG) with BigQuery and OpenAI. By default, you cannot directly use OpenAI Cloud Models within BigQuery. Try it This template comes with access to a *public BigQuery table** that stores part of the n8n documentation (about nodes and triggers), allowing you to try the workflow right away: n8n-docs-rag.n8n_docs.n8n_docs_embeddings* ⚠️ *Important:* BigQuery uses the *requester pays model.* The table is small (~40 MB), and BigQuery provides *1 TB of free processing per month**. Running 3–4 queries for testing should remain within the free tier, unless your project has already consumed its quota. More info here: BigQuery Pricing* Why this workflow? Many organizations already use BigQuery to store enterprise data, and OpenAI for LLM use cases. When it comes to RAG, the common approach is to rely on dedicated vector databases such as Qdrant, Pinecone, Weaviate, or PostgreSQL with pgvector. Those are good choices, but in cases where an organization already uses and is familiar with BigQuery, it can be more efficient to leverage its built-in vector capabilities for RAG. Then comes the question of the LLM. If OpenAI is the chosen provider, teams are often frustrated that it is not directly compatible with BigQuery. This workflow solves that limitation. Prerequisites To use this workflow, you will need: A good understanding of BigQuery and its vector capabilities A BigQuery table containing documents and an embeddings column The embeddings column must be of type FLOAT and mode REPEATED (to store arrays) A data pipeline that generates embeddings with the OpenAI API and stores them in BigQuery This template comes with a public table that stores part of the n8n documentation (about nodes and triggers), so you can try it out: n8n-docs-rag.n8n_docs.n8n_docs_embeddings How it works The system consists of two workflows: Main workflow** → Hosts the AI Agent, which connects to a subworkflow for RAG Subworkflow** → Queries the BigQuery vector table. The retrieved documents are then used by the AI Agent to generate an answer for the user.

This n8n template demonstrates how to automate stock market technical analysis to detect key trading signals and send real-time alerts to Discord. It's built to monitor for the Golden Cross (a bullish signal) and the Death Cross (a bearish signal) using simple moving averages. Use cases are many: Automate your personal trading strategy, monitor a portfolio for significant trend changes, or provide automated analysis highlights for a trading community or client group. 💡 Good to know This template relies on the Alpha Vantage API, which has a free tier with usage limits (e.g., API calls per minute and per day). Be mindful of these limits, especially if monitoring many tickers. The data provided by free APIs may have a slight delay and is intended for informational and analysis purposes. Disclaimer**: This workflow is an informational tool and does not constitute financial advice. Always do your own research before making any investment decisions. ⚙️ How it works The workflow triggers automatically every weekday at 5 PM, after the typical market close. It fetches a list of user-defined stock tickers from the Set node. For each stock, it gets the latest daily price data from Alpha Vantage via an HTTP Request and stores the new data in a PostgreSQL database to maintain a history. The workflow then queries the database for the last 121 days of data for each stock. A Code node calculates two Simple Moving Averages (SMAs): a short-term (60-day) and a long-term (120-day) average for both today and the previous day. Using If nodes, it compares the SMAs to see if a Golden Cross (short-term crosses above long-term) or a Death Cross (short-term crosses below long-term) has just occurred. Finally, a formatted alert message is sent to a specified Discord channel via a webhook. 🚀 How to use Configure your credentials for PostgreSQL and select them in the two database nodes. Get a free Alpha Vantage API Key and add it to the "Fetch Daily History" node. For best practice, create a Header Auth credential for it. Paste your Discord Webhook URL into the final "HTTP Request" node. Update the list of stock symbols in the "Set - Ticker List" node to monitor the assets you care about. The workflow is set to run on a schedule, but you can press "Test workflow" to trigger it manually at any time. ✅ Requirements An Alpha Vantage account for an API key. A PostgreSQL database to store historical price data. A Discord account and a server where you can create a webhook. 🎨 Customising this workflow Easily change the moving average periods (e.g., from 60/120 to 50/200) by adjusting the SMA_SHORT and SMA_LONG variables in the "Compute 60/120 SMAs" Code node. Modify the alert messages in the "Set - Golden Cross Msg" and "Set - Death Cross Msg" nodes. Swap out Discord for another notification service like Slack or Telegram by replacing the final HTTP Request node.

🕵️‍♂️ Reddit Intelligence Monitor: AI-Powered Scraping with BrowserAct Automate your market research and competitor analysis with this powerful "Set and Forget" workflow. It monitors Reddit for specific keywords and competitor subreddits, uses BrowserAct for stealth scraping, analyzes the sentiment with AI, and delivers a daily intelligence digest to your Google Sheets. 💡 Key Features Powered by BrowserAct**: Leverages cloud browser automation to stealthily scrape Reddit data without getting blocked. Dual-Track Monitoring**: Simultaneously tracks "Brand Competitors" (Subreddits) and "Topic Keywords" (Search Results). AI Analysis**: Summarizes the top 3 trending posts into a single concise daily report, filtering out noise. Structured Archive**: Automatically cleans, formats, and archives intelligence with source links into Google Sheets. 🛠️ How it Works Config Read: Reads a list of monitoring targets from a Google Sheet. Route: Splits the task into two paths (Competitor vs. Keyword) based on input type. Scrape: BrowserAct navigates to the target Reddit pages and extracts the latest posts. Process: Custom Code nodes clean the data and merge top 3 posts into a single prompt. Analyze: AI Agent generates an executive summary for each topic. Archive: Final reports are appended to your "Report" Google Sheet. 📋 Setup Guide Google Sheets: Create a sheet with two tabs: Config: Columns keywords (for search terms) and competitor (for subreddit names). Report: Columns Date, Competitor/Keyword, Summary, Link. BrowserAct: Connect your BrowserAct credentials and ensure you have the Reddit scraping task template ready. AI Model: Configure the Google Gemini Chat Model (or swap for OpenAI). Schedule: Enable the Schedule Trigger for daily automated runs.

This workflow creates an automated Product Intelligence Engine that continuously collects signals from multiple product sources and generates structured PRD updates using AI. It ingests conversations, feedback, support tickets, analytics, and design comments, standardizes them, analyzes them with an AI PRD Agent, and automatically updates a Google Doc with structured PRD recommendations. Instead of manually reviewing Slack threads, Zoom calls, Jira comments, support tickets, and customer forms, this workflow centralizes everything into one intelligent PRD analysis system. High-Level Architecture - The workflow runs in 4 layers: 1. Signal Ingestion Layer Captures product signals from: • Slack (channel messages + app mentions) • Customer Form submissions • Zoom recordings (scheduled) • Jira comments (scheduled) • Zendesk tickets (scheduled) • Figma comments (file updates) • Platform analytics via webhook • (Extendable to Salesforce / HubSpot) 2. Standardization Layer Each source passes through a Format Node that: • Extracts relevant text • Normalizes metadata • Adds timestamps • Labels source type All inputs are converted into a unified "product signal" object. 3. Intelligence Layer (AI PRD Agent) All signals are merged into a single stream using a Merge node. The PRD Analysis Agent then: • Extracts feature requests • Detects scope changes • Identifies risks and constraints • Evaluates priority signals • Detects target user shifts • Generates structured PRD updates 4. PRD Governance Layer - output in a Google Doc The structured AI output is appended to a Google Doc, which is fully traceable. This creates a living PRD that continuously evolves based on real product signals. Required Credentials (And How To Add Them): You will need to configure the following credentials in n8n: 1. Slack Used for Slack Trigger. Steps: Create a Slack App at api.slack.com Enable: app_mentions:read channels:history chat:write (optional if you want replies) Install app to workspace Copy Bot OAuth Token In n8n → Create Slack API credential Paste token Reference - https://www.youtube.com/watch?v=qk5JH6ImK0I 2. Zoom (OAuth2) Used to fetch recordings. Steps: Create an OAuth App in Zoom Marketplace Add the Redirect URL from n8n Copy Client ID + Secret Add Zoom OAuth2 credential in n8n Connect account Reference - https://www.youtube.com/watch?v=BC6O_3LYgac 3. Google Docs (OAuth2) Used to update PRD document. Steps: Create Google Cloud Project Add Doc URl to n8n Replace the example Google Doc URL with your own PRD document. Reference - https://www.youtube.com/watch?v=iieEHvu93dc 4. Jira (Cloud) Steps: Generate API token from Atlassian Create Jira Software Cloud credential Enter: Email API token Domain Reference - https://www.youtube.com/watch?v=T4z7lzqSZDY 5. Zendesk Steps: Generate API token Add Zendesk credential Enter: Subdomain Email API token 6. Figma Steps: Generate a personal access token in Figma Add Figma credentials with the team ID Paste token 7. Platform Analytics Webhook Replace: <PLACEHOLDER_VALUE__your_analytics_api_endpoint> With your real analytics endpoint. You can: • Send Mixpanel exports • Send Amplitude exports • Or POST custom JSON What Makes This Powerful • Eliminates product signal silos • Creates AI-driven PRD governance • Ensures traceability of decisions • Enables continuous PRD evolution • Scales across teams

📘 Description This workflow turns raw product inputs into a complete, launch-ready AI-generated social media campaign package. It accepts product details via webhook, sanitizes messy fields, generates a strategic campaign blueprint, produces Instagram captions, creates discovery-optimized hashtags, generates photorealistic commercial images, computes optimal posting times, assembles all outputs into a unified JSON package, and finally delivers the entire campaign to Slack. Multiple AI agents work in sequence to generate structured outputs — each parsed and validated using strict JSON schemas. Images produced by DALL·E 3 are uploaded to Cloudinary for hosting. A post-processing module then merges captions, images, hashtags, and schedules into a final payload. A robust error handler ensures every failure is captured and sent to Slack with diagnostic information. This workflow replaces an entire marketing team’s creative production pipeline, producing consistent, multi-asset campaign kits in minutes. ⚙️ What This Workflow Does (Step-by-Step) 🟢 Receive Product Details via Webhook Captures incoming product data including name, description, benefits, audience, and brand voice. 🧹 Clean & Normalize Product Input Fields Sanitizes escaped characters, trims whitespace, and prepares stable fields for AI consumption. 🧠 Generate Campaign Blueprint Using AI Creates a full strategic blueprint in structured JSON: • Article summary • Insights • Tone and target audience mapping • Platform-specific post objects 🧠 LLM Engine + Structured Parser for Blueprint Ensures blueprint output is clean, validated JSON aligned with schema. ✍️ Generate Instagram Captions Using AI Produces five short, conversion-ready captions + CTAs, based on blueprint insights. 🧠 Caption LLM + Structured Parser Validates caption schema for downstream use. #️⃣ Generate Hashtag Set Using AI Creates 12–18 optimized hashtags using discovery strategy (broad, mid, niche). 🧠 Hashtag LLM + Parser Validates and ensures hashtags follow correct JSON structure. 🎨 Split Campaign Posts for Image Generation Breaks out each post’s image prompt for independent asset creation. 🖼️ Generate Social Media Image Using AI Uses DALL·E 3 to create ultra-realistic, 8K-style commercial visuals tailored to the campaign. ☁️ Upload Generated Image to Cloudinary Uploads rendered images and retrieves secure public URLs. 🕒 Generate Optimal Posting Schedule Using AI Recommends best posting time per platform (Asia/Kolkata timezone) + reasoning. 🧠 Schedule LLM + Parser Ensures a structured schedule schema with platform, time, and rationale. 🔀 Combine All Campaign Assets Merges: • Cloudinary image URLs • Captions + CTAs • Hashtag set • Posting schedule into one final dataset. 🧩 Prepare Final Campaign Package JSON Constructs production-ready unified JSON: images, captions, hashtags, schedule. 💬 Send Final Campaign Package to Slack Delivers formatted campaign output with: • Image URLs • Captions + CTAs • Hashtags • Posting times for immediate creative review. 🚨 Error Handler Trigger → Slack Alert Captures workflow failures and sends structured debugging info to Slack. 🧩 Prerequisites • OpenAI API (GPT-4o + DALL·E 3) • Cloudinary account (image hosting) • Slack bot token • Valid webhook endpoint • Clean product input JSON 💡 Key Benefits ✔ Full AI-generated multi-asset campaign in minutes ✔ Eliminates manual copywriting, design, and planning ✔ Ensures structured, reliable JSON at every stage ✔ Creates polished commercial visuals instantly ✔ Produces posting strategy tailored to audience behavior ✔ Unified campaign delivery straight to Slack 👥 Perfect For Consumer brands launching fast cycles Agencies needing rapid campaign generation Teams without in-house designers/copywriters Influencers or D2C founders wanting automated content production

HOW IT WORKS This workflow automates end-to-end data intelligence processing by ingesting structured data (CSV, JSON), enriching it through multiple AI analysis pathways, and generating actionable insights. Designed for business analysts, data scientists, and operations teams, it solves the problem of manual data enrichment and fragmented analysis by consolidating diverse AI models (GPT-4, LLM analysis, sentiment detection) into a unified pipeline. Data flows from source ingestion → enrichment/validation → branching into three specialized analysis paths (Competitive Intelligence, Sentiment Analysis, Market Insights) → aggregation → result storage (Google Sheets) and notifications (Slack, Gmail). Each path applies distinct AI models for comprehensive intelligence gathering. SETUP STEPS Configure OpenAI API key in credentials Set up Google Sheets connection with service account Add Slack webhook for notifications Connect Gmail for automated report distribution Configure NVIDIA API (if using specialized models) Map input data source (CSV upload or API endpoint) Test each branch independently before full deployment PREREQUISITES OpenAI API key, Google Sheets access, Slack workspace, Gmail account, basic n8n familiarity. USE CASES Market research automation, competitive intelligence monitoring, customer feedback analysis at scale CUSTOMIZATION Swap AI models (Claude, Gemini, Llama), add/remove analysis branches, modify output destinations BENEFITS Eliminates manual data processing (80% time savings), enables simultaneous multi-perspective analysis

📘 Description This workflow automates document understanding by accepting uploaded PDF or TXT files, extracting their text, generating a structured summary and question–answer set using GPT-4o, validating the AI output, and returning a clean JSON response to the requester. It also sends an internal Slack preview and logs malformed outputs for debugging. It performs intelligent file-type detection, handles binary text extraction, enforces strict JSON formatting from the AI model, and ensures that the final response is clean, structured, and ready for use in downstream systems. All errors—missing text, invalid JSON, or malformed AI output—are captured automatically in Google Sheets. The workflow is designed as a plug-and-play document-analysis engine that converts any uploaded document into meaningful insights instantly. ⚙️ What This Workflow Does (Step-by-Step) 📥 Receive Document Upload via Webhook Captures incoming files (PDF or TXT) posted to the webhook endpoint. 🔍 Check If Uploaded File Is PDF / TXT Detects file extension and routes it correctly for extraction: PDF → PDF extractor TXT → text extractor Other file types are ignored. 📝 Extract Text from Document Extracts readable text from PDF binaries Reads raw plain text from TXT files The extracted text becomes input for the AI analysis. 🤖 Generate Summary & Q&A Using AI Uses GPT-4o to produce: A 150–200 word summary Five structured Q&A pairs Output must strictly follow the specified JSON schema. 🧠 LLM Engine + Memory Context GPT-4o provides the reasoning engine Memory buffer maintains short context for stability Output parser ensures schema compliance ⚠️ Validate AI Output Before Processing Checks whether output is non-empty and correctly structured. Invalid → logged to Google Sheets. 📊 Log Invalid AI Output to Google Sheet Records failures for audit, debugging, and retraining. 🧹 Unwrap AI Output Object Removes unnecessary array wrappers and normalizes the result. 📤 Prepare Final Response Payload Ensures the workflow responds with a single clean JSON object. 🔁 Send Final Summary & Q&A Response to Webhook Returns the final structured JSON to the requesting system. 💬 Send Summary Preview to Slack Shares a short preview (first 300 characters) for internal visibility. 🧩 Prerequisites Webhook endpoint configured for uploads Azure OpenAI GPT-4o credentials Google Sheets OAuth connection Slack bot token 💡 Key Benefits ✔ Fully automated PDF/TXT understanding ✔ AI-powered summary + structured Q&A ✔ Strict JSON compliance for downstream systems ✔ Error-proof: logs all failures for investigation ✔ Slack visibility for quick internal review ✔ Works with minimal human involvement 👥 Perfect For Research teams Documentation workflows Customer-support intelligence Interview screening document parsing Internal knowledge extraction systems