Regex Slots - Cognigy Documentation
Regex Slots
Regex Slots are user-defined Slots, which apply regular expressions (regex) to match and extract specific patterns from user inputs. Unlike Lexicon Slots, which rely on Keyphrases, Regex Slots are ideal for structured data that follows a predictable format, such as:
- Phone numbers (
(123) 456-7890) - Email addresses (
user@example.com) - Product codes (
ABC-123-XYZ)
You can use this type of Regex Slots only as Slot Fillers.
Recommendations
- Regex Slots require careful crafting to avoid overly broad or restrictive patterns.
- Complex regular expressions can be error-prone; test them thoroughly to avoid missing matches or capturing incorrect data.
- Avoid conflicts with Lexicon Slots and System Slots by ensuring unique Slot names.
Working with Regex Slots
To start working with Regex Slots, you can use one of the following options:
- Apply to an Entire Flow
- Apply After a Specific Point in a Flow
To define a regex pattern that runs on all user inputs in the Flow, follow these steps:
- In the Flow editor, go to NLU > Slot Fillers.
- In the upper-left corner, click + New Slot Filler.
- In the New Slot Filler window, fill in the following fields:
- Name – enter a unique name for the Slot Filler. For example,
phone_number. - Type – select Regular Expression.
- Regular Expression – enter a regex pattern. Patterns must start and end with
/. Flags likegfor global matching aren’t supported. - Context Key – enter a Slot name to store the matched data in the Context object. For example,
phone_number.
- Name – enter a unique name for the Slot Filler. For example,
To define a regex pattern that runs after a specific user input in the Flow, follow these steps:
- In the Flow editor, add a Regex Slot Filler Node where you want to process the input, for example, after a Question Node.
- In the Node editor, configure the following fields:
- Regular Expression - enter a regex pattern without leading and trailing slashes, for example,
^\d{3}-\d{3}-\d{4}$. - Flags – enter flags you want to apply to the regular expression, for example,
ifor case-insensitive,mfor multiline. Note that thegflag should always be specified. - Slot Name – enter the Slot name to store the matched data in the
slots.<slot-name>Input object. For example,phone_number.
- Regular Expression - enter a regex pattern without leading and trailing slashes, for example,
Once a Regex Slot is filled, the extracted data can be used in the conversation logic, for example, for the following purposes:
- Validating user input, for example, ensuring a valid email format.
- Triggering specific actions, for example, saving a phone number to the Contact Profile.
- Combining with Question Nodes to prompt users to re-enter missing or incorrectly formatted information.
Example
These examples show how regex patterns fill Slots in the Context object.
Single Match
User Input: My phone number is 555-123-4567
Regex Pattern: /\d{3}-\d{3}-\d{4}/
Slot Name: phone_number
Result: The context.phone_number Slot is filled with 555-123-4567. The AI Agent responds Thanks for providing your phone number: {{context.slots.phone_number}}!
Multiple Matches
When the user input contains multiple values that match the pattern, only the first match is stored.
User Input: The booking codes are 123456 and 789123
Regex Pattern: /\b\d{6}\b/
Slot Name: booking_codes
Result: The context.booking_codes Slot is filled with 123456.
Use Cases
- Extracting a booking code:
/^[A-Z]{3}-\d{4}$/to captureXYZ-1234. - Capturing email addresses with the
@companydomain:/^[a-zA-Z0-9._%+-]+@company\.com$/. - Identifying dates in specific formats:
/^\d{2}\/?\d{2}\/?\d{4}$/for04/22/2025.