Ask AI
Skip to main content

LLM Completion Request

Function: LLM Completion Request​

Send a prompt to an AI language model (LLM) and receive a generated response. This action allows you to leverage powerful AI capabilities for tasks like text generation, summarization, data extraction, and more, either as plain text or structured data.

Input​

  • User prompt (STRING, Required) A clear, concise instruction or question that you want the AI to respond to. This is the primary input for the AI.
  • User prompt placeholders (OBJECT, Optional) If your "User prompt" contains placeholders like \{\{NAME\}\}, you can define their values here. The AI will see the prompt with these placeholders replaced by their corresponding values.
    • Example: If your prompt is "Hello, {{NAME}}!", and you set NAME to "Alice", the AI will receive "Hello, Alice!".
  • System prompt (STRING, Optional) Initial instructions that guide the AI's behavior and persona throughout the conversation. This helps set the context and tone for the AI's responses.
    • Example: "You are a helpful assistant that provides concise answers."
  • System prompt placeholders (OBJECT, Optional) Similar to user prompt placeholders, these replace \{\{PLACEHOLDER\}\} values within your "System prompt".
  • Model (SELECT_ONE, Required) Choose the specific AI language model you want to use. Different models have varying capabilities, costs, and performance characteristics.
    • Available options include: gpt-5, gpt-5-mini, gpt-4o, gpt-4o-mini, claude-sonnet, claude-haiku, claude-opus, mistral-medium, mistral-large, llama-4, gemini-2.5-flash, and many others.
  • Files (ARRAY of FILE, Optional) A list of files (e.g., documents, images) that you want the AI to analyze or refer to when generating its response. This feature is only available for certain models.
    • Visibility: This input only appears if you select a model that supports file attachments (e.g., specific GPT, Claude, Llama, Mistral, or Gemini models).
  • Api token (PASSWORD, Optional) Provide your own API token for the selected AI model. If you leave this blank, the platform will use your available AI credits.

Output​

  • Result (VARIABLE) The variable where the AI's response will be stored. If a 'Response format' is specified, the output will be an OBJECT matching that structure. Otherwise, it will be a plain STRING.
  • Response format (DATA_FORMAT, Optional) Specify a predefined data structure (like a database table schema) that the AI should use to format its response. If provided, the AI will attempt to return its answer as a structured OBJECT conforming to this format. Leave this blank if you expect a plain text response.

Execution Flow​

Real-Life Examples​

Example 1: Generating a Marketing Slogan​

Scenario: You need a catchy slogan for a new eco-friendly coffee brand.

  • Inputs:
    • User prompt: "Generate 5 short, catchy marketing slogans for a new eco-friendly coffee brand called 'Green Bean Brew'. Focus on sustainability and great taste."
    • System prompt: "You are a creative marketing expert specializing in brand messaging."
    • Model: gpt-4o-mini
    • Result: CoffeeSlogans
  • Result: The variable CoffeeSlogans will contain a string with five suggested slogans, such as: "1. Green Bean Brew: Sip Sustainably, Taste Exceptionally. 2. Your Cup, Our Planet: Green Bean Brew. 3. Earth-Friendly Coffee, Unforgettable Flavor. 4. Green Bean Brew: Good for You, Good for Earth. 5. Taste the Future, Sustain the Planet with Green Bean Brew."

Example 2: Extracting Structured Data from Customer Reviews​

Scenario: You have customer reviews and want to automatically extract the product name, rating, and a summary of the feedback into a structured format for analysis.

  • Inputs:
    • User prompt: "Analyze the following customer review: 'I absolutely love the new 'SmartWatch Pro'! The battery life is incredible, lasting days on a single charge. However, the strap feels a bit cheap. Overall, a solid 4 out of 5 stars.' Extract the product name, star rating, and a brief summary of the feedback."
    • System prompt: "You are an expert in customer feedback analysis. Always output in JSON format."
    • Model: gpt-4o
    • Response format: A DATA_FORMAT named ReviewAnalysis with the following fields:
      • productName (STRING)
      • rating (NUMBER)
      • summary (STRING)
    • Result: AnalyzedReview
  • Result: The variable AnalyzedReview will contain an OBJECT structured as follows: 
    \{
      "productName": "SmartWatch Pro",
      "rating": 4,
      "summary": "Excellent battery life, but the strap quality could be improved."
    \}

Example 3: Summarizing a Document with Specific Instructions​

Scenario: You have a long project proposal document and need a concise summary focusing on the budget and timeline sections for a quick management review.

  • Inputs:
    • User prompt: "Summarize the key budget allocations and project timeline details from the attached document. Focus on the total budget, major cost categories, and critical milestones."
    • System prompt: "You are a project management assistant. Provide a clear and concise summary suitable for executives."
    • Model: claude-opus-4-0 (or another model that supports file input)
    • Files: Project_Proposal_Q4_2024.pdf (a PDF file containing the proposal)
    • Result: ProposalSummary
  • Result: The variable ProposalSummary will contain a string summarizing the budget and timeline information extracted from the Project_Proposal_Q4_2024.pdf document.