Others

​

Required data

• Name
  Type: string
  Name of the global state variable. Example: customer_name
• Value
  Type: string
  Value to save into the state variable. Supports both static values and dynamic variables. Example: John Smith
  Example: @name

​

Notes

• Stores data globally for the duration of the flow.
• Can be accessed in other nodes, tools, and actions.
• If the variable already exists, its value will be overwritten.
• Use meaningful names to keep flows readable.
• Use @variable_name to pass dynamic values
• If the same state variable already exists, the value will be overwritten

​

When to use

• Save user input for later steps
• Store intermediate data between actions
• Pass values into tools or API calls

​

Best practices

• Use clear variable names (user_name, phone_number)
• Avoid overwriting important values unintentionally
• Store only necessary data for later use

​

Description

​

Async mode

• Async
  Type: boolean If enabled, the action runs in the background and the bot continues the conversation without waiting for the result.

​

Output

6

​

Required data

• Product
  Type: string Product name or short description used to identify the correct commodity code. Example: Wooden dining table

​

Optional parameters

• Language
  Type: string Language used for the search and response. Options: en, ro, ru, xx
  Default: en

​

Required data

• Function
  Type: code Custom function written in Python that processes input data and returns structured output. Example:
  def _foo(tool_params: dict, interaction_data: dict):
      # do something here
      return {"result": "ok"}
  

​

Optional parameters

• TTL
  Type: number Time-to-live for cached results (in seconds). Example: 30

​

Async mode

• Async
  Type: boolean If enabled, the action runs in the background and the bot continues the conversation without waiting for the result.

​

Output

{
  "key": "value"
}

​

Notes

• The function must return a JSON-serializable object.
• Use tool_params to access input parameters.
• Use interaction_data for conversation context.
• Define response schemas if needed for structured outputs.

​

Required data

• Column Name
  Type: string
  Name of the column used for filtering.
  Example: status
• CSV Category Name
  Type: string
  Name of the CSV dataset (category).
  Example: orders
• Filter Value
  Type: string
  Value used to filter rows.
  Example: completed

​

Optional parameters

• Comparison Operator
  Type: string Defines how values are compared. Options:
  • Equal
  • Not Equal
  • Greater Than
  • Less Than
  • Greater Than or Equal
  • Less Than or Equal
  • Contains
  • Does Not Contain
  • Starts With
  • Ends With
  Default: Equal

​

Async mode

• Async
  Type: boolean Runs in background without waiting for result.

​

Output

[
  {
    "column": "value"
  }
]

​

Notes

• Column names must match CSV exactly
• Operators depend on data type (text vs number)
• Use Contains / Starts With for text filtering
• Use numeric operators for numbers

​

Use cases

• Answer questions based on internal knowledge
• Search through documentation or FAQs
• Provide contextual responses from stored data
• Combine retrieval with AI-generated output

​

Required data

• Vector store ID
  Type: string
  Identifier of the vector database used for retrieval.
• Instruction
  Type: string
  Defines how the model should behave and format the response.
• Query
  Type: string
  Search input used to retrieve relevant data.

​

Optional data

• Model name
  Type: string
  Model used for response generation. Example: gpt-5 nano
• Reasoning level
  Type: string
  Controls depth of reasoning. Example: None, Low, Medium, High
• Web search
  Type: boolean
  Enables fallback to web data if needed.

​

Output

• Returns a generated response based on retrieved data
• Combines vector search results with model reasoning

​

Notes

• Quality depends on the vector store content
• Better queries improve retrieval accuracy
• Higher reasoning increases latency but improves results
• Use clear instructions to control tone and structure

​

Use cases

• Pass calculated or processed data to the model
• Override or enrich the assistant response
• Inject dynamic values into the conversation
• Control final output of a flow

​

Required data

• Custom value
  Type: string
  Value returned to the model. Example: Your appointment is confirmed for tomorrow at 3 PM

​

Optional data

• Async
  Type: boolean
  Runs the action in the background without waiting for the result.

​

Output

• Returns the defined value to the LLM
• Used by the assistant to generate or modify the response

​

Notes

• Value should be clear and ready for direct use
• Avoid unnecessary formatting or extra text
• Use when you need full control over what the model receives

​

Use cases

• Redirect user to another flow
• Split logic between different workflows
• Handle fallback or escalation scenarios
• Reuse existing flows

​

Required data

• Set current flow
  Type: string
  Name of the target flow or node. Example: booking_flow

​

Optional data

• Make as transition
  Type: boolean
  Treats the action as a transition between nodes.
• Incognito call
  Type: boolean
  Executes the flow without affecting visible conversation state.
• Only if everything was successful
  Type: boolean
  Executes only if previous actions completed successfully.
• Async
  Type: boolean
  Runs the action in the background without waiting for completion.

​

Output

• Redirects the conversation to the specified flow
• Updates the current execution context

​

Notes

• Target flow must exist in the system
• Use clear and consistent naming for flows
• Avoid circular flow transitions

​

Use cases

• Provide quick reply options
• Guide users through predefined flows
• Reduce typing effort
• Improve conversion in structured scenarios

​

Required data

• List of hints
  Type: string
  List of options displayed to the user, one per line. Example:
  Book an appointment
  Check availability
  Contact support
  

​

Optional data

• Async
  Type: boolean
  Runs the action in the background without waiting for completion.

​

Output

• Displays clickable hints in the widget
• User selection is returned as input to the conversation

​

Notes

• One hint per line
• Keep hints short and clear
• Limit the number of options to avoid overload
• Ensure options match available flow paths

​

Use cases

• Generate dynamic quick replies
• Suggest next steps based on user input
• Adapt hints to conversation context
• Improve engagement without hardcoded options

​

Required data

• Custom prompt
  Type: string
  Instruction for the LLM to generate hints. Example:
  Generate 3 short options a user might choose next for booking a service.
  Keep them concise and action-oriented.
  

​

Optional data

• Async
  Type: boolean
  Runs the action in the background without waiting for completion.

​

Output

• Displays generated hints in the widget
• Each hint is returned as selectable user input

​

Notes

• Keep prompts clear and specific
• Limit number of generated hints (e.g., 3–5)
• Ensure hints are short and easy to understand
• Avoid vague or overly long suggestions

​

Use cases

• Generate conversation summaries
• Extract key facts and decisions
• Prepare reports for team review
• Send summaries via email or integrations

​

Required data

• Instruction
  Type: string
  Defines how the summary should be structured and what to include. Example:
  Summarize the conversation. Include:
  - Main request
  - Key details
  - Final outcome
  

​

Optional data

• Cut conversation
  Type: boolean
  Clears or resets the conversation after summary is generated.
• Run immediately
  Type: boolean
  Executes the action instantly when triggered.
• Only if everything was successful
  Type: boolean
  Runs only if previous actions completed successfully.
• Async
  Type: boolean
  Runs the action in the background without waiting for completion.

​

Output

• Returns a structured summary of the conversation
• Can be used in further actions (email, CRM, logs, etc.)

​

Notes

• Keep instructions clear and structured
• Avoid overly long prompts
• Output format depends on the instruction
• Useful for automation and reporting workflows

​

Use cases

• Add delay between messages
• Simulate human-like response timing
• Wait before triggering next action
• Control pacing in workflows

​

Required data

• Timeout
  Type: number
  Number of seconds the assistant should wait before continuing. Example: 5

​

Optional data

• Async
  Type: boolean
  Runs the action in the background without blocking the conversation.

​

Output

• Delays execution of the next step in the flow
• Continues automatically after the specified time

​

Notes

• Value is in seconds
• Use short delays to avoid poor user experience
• Async mode allows conversation to continue without waiting