Instruct

Instruct Module – Configuration Guide

The Instruct module in Great Wave AI Studio enables detailed setup of AI agents tailored to specific operational needs. Follow the steps below to configure your agent according to your desired behavior and interaction mode.

Step 1: Select the Agent Mode

Use-Context Mode

• Agent relies only on selected Index (from a Collection)

• Does not use the LLM’s general knowledge base

• Requires:

  • Index selection

  • LLM selection (used to interpret context only)

Context-And-LLM Mode

• Agent uses a combination of context (from the Index) and the LLM’s internal knowledge

• Balances curated data with AI’s generative capabilities

• Requires:

  • Index selection

  • LLM selection

LLM-Only Mode

• Agent ignores context and relies solely on its pre-trained knowledge

• Ideal for open-domain tasks without specific data sources

• Requires:

  • LLM selection only (Index is disabled)

API-Agent Mode

• Agent communicates with external APIs in real time to provide data-driven responses

• Does not use context or LLM knowledge for response generation

• Requires:

  • API endpoint and parameter configuration

  • General instructions for how the API should be used

Step 2: Select Index (Only for Context Modes)

• Choose an Index (from a Collection) for the agent to reference

• This becomes the primary source of truth for grounded responses

Step 3: Select the Language Model (LLM)

• Use the LLM dropdown to choose a model such as GPT-3.5 or GPT-4o

• LLM is used to interpret either the context, generate standalone responses, or construct API calls depending on the mode

Step 4: Configure General Instructions

• Use the General Instructions box to define:

  • Agent’s overall role and responsibilities

  • Domain-specific rules

  • Desired tone and style of response

  • Instructional logic (e.g., step-by-step reasoning)

Step 5: Advanced Options

Prepending Text

• Adds a static prefix to every user input before processing

• Useful for guiding the agent to a specific context or scenario

Labeling Context

• Customize how the context is referred to in responses (e.g., training data, user notes, company policy)

Context Utilization Choice

• Choose how context should be used:

  • Strict grounding: Agent uses only the Index

  • Hybrid mode: Agent uses context plus LLM’s general knowledge

API Configuration Interface (For API-Agent Mode Only)

Use this panel to define the structure and content of your API interactions

HTTP Method and Endpoint

• Select the HTTP method: GET, POST, PUT, or DELETE

• Enter the target API URL in the input field

Headers Section

• Click + Add Header to enter required HTTP headers as key-value pairs

Load Params Dropdown

• Expand this to input query parameters or body data needed for the API call

Params Section

• Key: Parameter name (e.g., userId, limit) – required

• Type: Choose data type (String, Number, Boolean)

• Content Type: Define as either Predefined or Dynamic

• Value: Input the actual value – required

Add Params

• Add as many parameters as required by your API

Final Step: Save and Test

• Click Refresh Agent to apply your changes

• Test agent responses across multiple scenarios

• Revisit and refine context, LLM choice, API parameters, and instructions as needed

• Use the "Open Live" button on the right side to distribute.