Skip to content
Agentic Workflowdocs
v0.8.2Install free

get started

Summarise a page with AI

Add a chat model, on your device, with an API key or with Ollama, and get three bullets for whatever page you're reading.

Lesson 3 of 5· 8 min · Level 1 Works on-device
You will
  • Read a whole page with Get All Text
  • Connect a chat model to a Basic LLM Chain
  • Write a prompt that uses the page text
  • Show the result
Before you start

AI in a workflow takes two nodes: a chain that holds the prompt, and a model plugged into the chain’s Model slot. The model can run on your device (free, private, offline after the download), through a provider with your API key, or through Ollama on your computer.

workflow · summarize-page-llm-chain.awf 4 steps

Summarise this page in 3 bullets: a workflow with 4 steps, starting with When Started and ending with Display Markdown.

  1. When Started(Trigger — open its doc page)
  2. Get All Text(In-page action — open its doc page)
  3. Basic LLM Chain(AI — open its doc page) + Web LLM
  4. Display Markdown(In-page action — open its doc page)
Open in AWFlow
What you'll build: read the page, ask the model for three bullets, show them.

Steps

Section titled “Steps”

Step 1:Start a workflow that reads the page

Create a new Blank canvas workflow named Summarise this page, as in lesson 2. Add the When Started trigger, then click its + and add Get All Text. Set Output field name to text.

Step 1, close-up of the app: Blank canvas, outlined.Step 1, close-up of the app: Blank canvas, outlined.Full screen: the whole screenshot for step 1
Blank canvas
The Create a workflow dialog, step 1 of 3 (Source, Choose, Details): Start from a template (recommended) at the top, then Blank canvas, which is selected, From the marketplace and Import, with Cancel and Continue buttons.The Create a workflow dialog, step 1 of 3 (Source, Choose, Details): Start from a template (recommended) at the top, then Blank canvas, which is selected, From the marketplace and Import, with Cancel and Continue buttons.

Check: When Started → Get All Text is on a new canvas.

Step 2:Add a Basic LLM Chain

Click the + on the right of Get All Text, type LLM and pick Basic LLM Chain. Its Model slot sits under the node, with its own +.

Step 2, close-up of the app: Basic LLM Chain, outlined.Step 2, close-up of the app: Basic LLM Chain, outlined.Full screen: the whole screenshot for step 2
Basic LLM Chain
The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.

Check: a Basic LLM Chain is linked to Get All Text, with an empty Model slot underneath.

Step 3:Connect a model

Click the + under Model and pick one of these:

Pick Web LLM. In Model, open the list and choose a model under Available to download. Llama 3.2 1B Instruct is the smallest and fastest to download; a 3B model writes better summaries. It downloads the first time it runs, then works offline and nothing leaves your device. See Local AI to install models ahead of time.

In Firefox, some graphics setups can’t run Web LLM; the model then shows Not supported here. Use Transformers Chat instead.

Step 3, close-up of the app: Model slot, with a model node under it, outlined.Step 3, close-up of the app: Model slot, with a model node under it, outlined.Full screen: the whole screenshot for step 3
Model slot, with a model node under it
The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.

Check: a model node hangs under the chain's Model slot and shows the model you picked.

Step 4:Write the prompt

Open Basic LLM Chain and paste this into Prompt:

Prompt
Summarise this page in 3 short bullets for a busy reader.


{{ $input.text }}

{{ $input.text }} is replaced by the page text from Get All Text when the workflow runs.

Step 4, close-up of the app: The Prompt field, outlined.Step 4, close-up of the app: The Prompt field, outlined.Full screen: the whole screenshot for step 4
The Prompt field
The settings panel of the Basic LLM Chain step in the Summarise this page example, with the Prompt field asking for 3 short bullets and ending with {{ $input.text }}.The settings panel of the Basic LLM Chain step in the Summarise this page example, with the Prompt field asking for 3 short bullets and ending with {{ $input.text }}.

Check: the Prompt field shows your text, with {{ $input.text }} at the end.

Step 5:Show the result

Click the + on the right of Basic LLM Chain and add Display Markdown. Set Title to Page summary and Content to {{ $input.response }}: the chain puts the model’s answer in a field called response.

Step 5, close-up of the app: Display Markdown shows the result, outlined.Step 5, close-up of the app: Display Markdown shows the result, outlined.Full screen: the whole screenshot for step 5
Display Markdown shows the result
The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.

Check: Display Markdown is linked after the chain, with Content set to {{ $input.response }}.

Step 6:Run it on an article

Switch to the tab with the article, then back to the panel, and click Test Workflow. The first on-device run waits for the model download (progress shows in the node’s run log); later runs take a few seconds. Click Save when it works.

Step 6, close-up of the app: Test Workflow, outlined.Step 6, close-up of the app: Test Workflow, outlined.Full screen: the whole screenshot for step 6
Test Workflow
The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.The workflow editor with the Summarise this page in 3 bullets example: a sticky note, then When Started, Get All Text, Basic LLM Chain and Display Markdown in a row. A Web LLM model node hangs under the chain’s Model slot. The chain has a warning badge and the Ready to run? button counts one setting still to fill.

Check: Basic LLM Chain → Output → Actual shows a response with three bullets about the article.

The nodes in this lesson: Get All Text, Basic LLM Chain, Display Markdown. How models plug into nodes: Model dependencies.

Ask Aria