Docs
Search K

Appearance

AI Web Chat ​

The AI Web Chat node allows users to submit a prompt and optionally a context window to an AI web chat service (e.g., OpenAI) and receive a response. This response can either be interpreted as structured code (e.g., JSON) or plain text, and the result is stored in a named variable for downstream use.

🧠 Purpose ​

Use this node to:

  • Interact with an AI service via a simple input prompt.
  • Provide additional context to guide the AI’s response.
  • Store results from AI responses in a reusable workflow variable.
  • Optionally, extract structured code snippets from the response.

⚙️ Node Configuration ​

This node is configured via a modal form in the Workflow UI. Users must fill out the following fields:

Field LabelTypeRequiredDescription
User InputText✅The primary user message or question to send to the AI service.
Context WindowText✅Optional context to provide additional instruction or history to the AI.
Request Code OnlyBoolean✅If set to true, the node attempts to extract a code block or object from the AI's response.
Save to Variable NameText✅The name of the variable in which to store the result from the AI service.

💡 Context Window is pre-processed by appending system information such as country and timestamp before being sent to the AI service.

🧪 Example Use Case ​

Let’s say a user wants to generate a sample JSON schema for a user profile form. They would:

  • Input: "Create a JSON schema for a user profile with name, email, and age"
  • Context Window: "Return only the object without explanation"
  • Request Code Only: true
  • Save to Variable Name: userSchema

This will store the parsed JSON object (if successful) into userSchema.

🧰 Runtime Behavior ​

  1. Validation: Ensures all required inputs are present.
  2. Service Resolution: Looks up credentials for the ServiceAccount-OpenAI.
  3. AI Request: Constructs the final prompt using system metadata and context.
  4. Response Handling:
    • If Request Code Only is true, attempts to deserialize the result into a JSON object.
    • Fallback: Saves raw string output.
  5. Storage: Stores the final result in the specified variable name.

📦 Output ​

  • Output Variable: The result from the AI is saved to the variable provided in the Save to Variable Name field.
  • The result may be:
    • A raw string (e.g., a paragraph or code snippet).
    • A parsed JSON object if deserialization is successful and Request Code Only = true.

⚠️ Error Handling ​

  • If the AI service fails to return a response, the node fails with status "Fail" and a message (AIAgent) - AI Service Failed.
  • If the OpenAI service account cannot be found, the task will also fail.
  • Exceptions during deserialization are caught, and the raw output is used instead.

🔐 Permissions & Service Accounts ​

This node depends on a configured ServiceAccount-OpenAI. This must be defined and accessible within your automation environment.

📘 Notes ​

  • The AI service is invoked via AISingleReequest() with system and user context prepended.
  • The timestamp and country are dynamically added at runtime to enhance contextual relevance.

✅ Success Criteria ​

  • The AI returns a response.
  • If Request Code Only is true, the response is successfully deserialized (or falls back gracefully).
  • Output is stored under the specified variable name.
  • taskSuccess = true, statusReturn = "Completed"

Tentech 2024