Generate Custom Text Content with IBM Granite 3.3 8B Instruct AI

Generate Custom Text Content with IBM Granite 3.3 8B Instruct AI

This workflow connects to Replicate’s API and uses the ibm-granite/granite-3.3-8b-instruct model to generate text.

✅ 🔵 SECTION 1: Trigger & Setup

⚙️ Nodes

1️⃣ On clicking 'execute'

What it does:* Starts the workflow manually when you hit Execute. Why it’s useful:* Perfect for testing text generation on-demand.

2️⃣ Set API Key

What it does:* Stores your Replicate API key* securely. Why it’s useful:** You don’t hardcode credentials into HTTP nodes — just set them once here. Beginner tip:** Replace YOUR_REPLICATE_API_KEY with your actual API key.

💡 Beginner Benefit

✅ No coding needed to handle authentication. ✅ You can reuse the same setup for other Replicate models.

✅ 🤖 SECTION 2: Model Request & Polling

⚙️ Nodes

3️⃣ Create Prediction (HTTP Request)

What it does:* Sends a POST request* to Replicate’s API to start a text generation job. Parameters include:** temperature, max_tokens, top_k, top_p. Why it’s useful:** Controls how creative or focused the AI text output will be.

4️⃣ Extract Prediction ID (Code)

What it does:* Pulls the prediction ID* and builds a URL for checking status. Why it’s useful:** Replicate jobs run asynchronously, so you need the ID to track progress.

5️⃣ Wait

What it does:* Pauses for 2 seconds* before checking the prediction again. Why it’s useful:** Prevents spamming the API with too many requests.

6️⃣ Check Prediction Status (HTTP Request)

What it does:* Polls the Replicate API for the current status* (e.g., starting, processing, succeeded). Why it’s useful:** Lets you loop until the AI finishes generating text.

7️⃣ Check If Complete (IF Condition)

What it does:* If the status is *succeeded, it goes to “Process Result.” Otherwise, it loops back to Wait and retries. Why it’s useful: Creates an automated polling loop without writing complex code.

💡 Beginner Benefit

✅ No need to manually refresh or check job status. ✅ Workflow keeps retrying until text is ready. ✅ Smart looping built-in with Wait + If Condition.

✅ 🟢 SECTION 3: Process & Output

⚙️ Nodes

8️⃣ Process Result (Code)

What it does:* Collects the final AI output*, status, metrics, and timestamps. Adds info like:**

✅ output → Generated text ✅ model → ibm-granite/granite-3.3-8b-instruct ✅ metrics → Performance data Why it’s useful:** Gives you a neat, structured JSON result that’s easy to send to Sheets, Notion, or any app.

💡 Beginner Benefit

✅ Ready-to-use text output. ✅ Easy integration with any database or CRM. ✅ Transparent metrics (when it started, when it finished, etc.).

✅✅✅ ✨ FULL FLOW OVERVIEW

| Section | What happens | | ------------------------------ | ---------------------------------------------------------------------------- | | ⚡ Trigger & Setup | Start workflow + set Replicate API key. | | 🤖 Model Request & Polling | Send request → get Prediction ID → loop until job completes. | | 🟢 Process & Output | Extract clean AI-generated text + metadata for storage or further workflows. |

📌 How You Benefit Overall

✅ No coding needed — just configure your API key. ✅ Reliable polling — the workflow waits until results are ready. ✅ Flexible — you can extend output to Google Sheets, Slack, Notion, or email. ✅ Beginner-friendly — clean separation of input, process, and output.

✨ With this workflow, you’ve turned Replicate’s IBM Granite LLM into a no-code text generator — running entirely inside n8n! ✨