Information Extractor

Pull named pieces of information (a name, a date, an amount) out of free text with a chat model.

Try it

工作流 · selection-to-todoist.awf 4 个步骤

Selection to Todoist Task：包含 4 个步骤的工作流，从 Right-click a selection 开始，到 Create the Todoist task 结束。

1. 1 Right-click a selection（Context Menu — 打开其文档页面） Context Menu
2. 2 Get selected text（页面内操作 — 打开其文档页面） 页面内操作
3. 3 Draft a task title（Information Extractor — 打开其文档页面） Information Extractor + Local AI model
4. 4 Create the Todoist task（Todoist — 打开其文档页面） Todoist

在 AWFlow 中打开 复制 .awf

What it does

Information Extractor sends Text to the chat model connected to its Model slot and asks it to extract the fields you list in What to extract. The result comes back in one field, extracted.

Without an output parser, extracted is the model’s reply as text. Connect a Structured Output Parser to the Output Parser slot to get a real object whose fields you can use in later nodes.

When to use it

• Read an invoice or order email and pull out the customer, date and total.
• Turn a support message into fields (product, problem, urgency) before saving it to a sheet.
• Collect contact details from a page you are visiting.

To rewrite or answer freely, use Basic LLM Chain. To pick one label for the text, use Text Classifier.

Inputs and settings

设置项	类型	作用
Text 必填（必填）	Long text	The text to extract information from.
What to extract 必填（必填）	Long text	Describe the fields to extract, or connect an Output Parser.
Options	Group	The options for the node.
Options › System message	Long text	The system message to use.

Fill Text with an expression such as {{ $input.text }}.

What to extract is required even when an output parser is connected. List each field and its format, for example: customer name, invoice date (YYYY-MM-DD), total amount as a number, currency code.

Use System message under Options for extra rules, such as Leave a field empty when it is not in the text.

Outputs

• Output

Each output item has these fields:

Field	类型	What it holds
extracted	Any	The data the model extracted

Shape of an output item

{
  "extracted": null
}

With a Structured Output Parser connected, read single fields downstream with expressions such as {{ $input.extracted.total }}. Without one, extracted is text, so field access does not work.

Dependencies and credentials

Slot	Accepts	Connections
Model 必填（必填）	chat model	One node.
Output Parser	output parser	One node.

• Model: a chat model such as Chat OpenAI or Chat Anthropic. The model node holds the API key; this node has no credential of its own.
• Output Parser: a Structured Output Parser whose JSON Output Example matches the fields you want, or an Auto-fixing Output Parser to repair replies that do not parse.

Example workflow

1. Get Selected Text reads an order email you highlighted, with Output field name text.
2. Information Extractor with Text {{ $input.text }} and What to extract customer name, order number, total. Connect an Chat OpenAI model and a Structured Output Parser with JSON Output Example { "customer": "Jane Doe", "orderNumber": "A-1042", "total": 59.9 }.
3. Display Object shows extracted, or a Google Sheets node appends {{ $input.extracted.customer }} and {{ $input.extracted.total }} as a row.

Common issues

• {{ $input.extracted.field }} is empty. No output parser is connected, so extracted is plain text. Connect a Structured Output Parser.
• The parser fails on the reply. The model answered with extra text around the data. Add an Auto-fixing Output Parser, or use a stronger model.
• “Dependency model is required”. Connect a chat model to the Model slot.
• “Model not found”. The connected model node is not set up: open it and choose a model and its credential.
• Fields are guessed when missing from the text. Say in System message to leave missing fields empty.

Recipes with this node

Turn highlighted text into a Todoist task 入门 · 6 分钟 需要连接

Related nodes

• Structured Output Parser: define the shape of extracted.
• Basic LLM Chain: free-form prompts.
• Text Classifier: route text by category.
• Sentiment Analysis: route text by tone.