# Set Up your AI Agent ## **Full Process (Step-by-Step)** *** Creating an AI Agent is like building your own smart digital assistant — one that performs a specific task exactly the way you want it to. In this section, we’ll walk through the full process together, using a real-life example to show how it works. Let’s build an AI Agent called **“Essay Evaluator”** — an assistant that can grade essays using rubrics and generate helpful feedback. *** ### Step 1: Open the AI Agents Dashboard * From the left-hand menu, click on **AI Agents**. * This takes you to the main screen where you can: * Browse and reuse ready-made agents * See agents you’ve already created * Start building your own agent
*** ### Step 2: Choose How You Want to Build At the top of the screen, you’ll see this question: > **“What do you want to build?”** Here, you’ll describe in plain language what you want your AI Agent to do. For example: > *“I want to build an AI Agent that can read and evaluate student essays based on five criteria: content relevance, structure, grammar, critical thinking, and clarity. It should give a total score and individual feedback on each section.”*
Once you enter this, you have two options: #### ➤ Option 1: **Generate with AI** * Click **Generate** to let the platform automatically fill in all the fields for you. * The system will generate: * Agent name * Agent description * Welcome message * Tool logic prompt * You can review and edit everything before saving. #### ➤ Option 2: **Build Manually** * Click **Build Manually** if you want full control. * You’ll manually enter each setup field step-by-step: | Field | What to Enter | | --------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Agent Name** | A simple name like “Essay Evaluator” or “Grammar Coach” | | **Agent Description** | A short sentence explaining what the agent does. Example: *“An assistant that evaluates essays based on specific scoring criteria and gives helpful feedback.”* | | **Welcome Message** | This is the first message users will see. Example: *“Hello and welcome! I’m the Essay Evaluator. 📝🎓 Please upload or paste your essay to get started.”* | | **Tool Icon** | Optional image or icon to personalize the agent’s look |
✅ **Tip:** If you prefer customization and already have a clear idea, use the **Build Manually** option. If you need help getting started, **Generate** is a great head start. *** ### Step 3: Define the Tool Logic (Agent Instructions) Next, you’ll write a **Tool Prompt** — a detailed instruction that tells your AI Agent exactly what it should do. #### Example Tool Logic for “Essay Evaluator”: > *“Essay Evaluator greets the user and asks them to upload or paste their essay. It then evaluates the essay based on five criteria: (1) Content Relevance, (2) Structure & Organization, (3) Grammar & Language, (4) Originality & Critical Thinking, and (5) Clarity. For each area, the agent assigns a score, provides a brief comment, and then shares a total score out of 100 along with a summary paragraph of feedback.”* ✅ **Tip:** Be as clear and specific as possible — the more structured your instructions, the better your agent will perform. ### Step 4: Adjust Advanced Settings (Optional but Powerful) Once you've defined what your AI Agent should do, you have the option to **fine-tune how it behaves** by adjusting advanced AI settings. These controls help you shape the *style*, *depth*, and *creativity* of your agent’s responses. Here are the settings you’ll see: #### **AI Engine Options** This lets you choose the type of AI model that powers your Agent. Think of it as choosing the engine for your car — it affects how powerful and capable your Agent is.
You have two options: #### Option 1: Use an Existing AI Avatar as the Engine If you’ve already created an AI Avatar on the platform (like a tutor, coach, or expert), you can select that avatar as the core engine. This means your Agent will inherit the voice, tone, style, and personality from your Avatar. * Best for when you want your AI Agent to reflect your existing assistant’s approach. * Your Agent will still perform the specific task you assign, but with the Avatar’s voice and behavior. #### Option 2: Use a Predefined AI Model If you don’t want to use an Avatar, you can choose a general-purpose AI model: * **GPT-3.5**: Efficient and basic, suitable for simple tasks. * **GPT-4**: Advanced and more accurate for reasoning and detailed logic. * **GPT-4o**: Optimized version of GPT-4, faster and better at multi-step thinking and language nuance. > ✅ **Recommendation**: Use **GPT-4o** if you want the best mix of performance and depth — especially for evaluation, analysis, or tutoring use cases. ***
### **Impact of Temperature on Output** #### **Low Temperature (e.g. 0.1 – 0.3)** * **Output:** Safe, repetitive, very logical. * **Tone:** Formal, predictable. * **Use Case:** Factual answers, technical documentation, or summaries. **Example Prompt:** “Tell me a story about a dog.”\ **Low temp output:** > “Once upon a time, there was a dog. He lived in a house. He liked to play fetch.” *Very boring, but safe.* *** #### **High Temperature (e.g. 0.8 – 1.0)** * **Output:** Creative, diverse, surprising. * **Tone:** Casual, sometimes wacky or inconsistent. * **Use Case:** Brainstorming ideas, writing poems or fiction, jokes. **Same prompt:** > “Once upon a time, a sky-blue dog named Zibble rode a skateboard through the clouds, searching for the lost bone of Atlantis.” *Wild, imaginative, sometimes silly.* *** ### 2. **Impact of Top-p on Output** #### **High Top-p (e.g. 0.9 – 1.0)** * **Output:** Diverse and open-ended, includes many possible choices. * **Effect:** Allows occasional creative flair without being too random. * **Use Case:** Balanced creativity with coherence — e.g., marketing copy or storytelling. *** #### **Low Top-p (e.g. 0.3 – 0.6)** * **Output:** Tightly focused, limited vocabulary. * **Effect:** Reduces risk of strange or incorrect answers — but can sound robotic or repetitive. * **Use Case:** Short answers, precise instructions, customer support. *** ### **Combining the Two: How They Interact** | Temperature | Top-p | Output Style | | ------------ | ---------- | ------------------------------------------------------------------------ | | Low (0.2) | High (1.0) | Precise, logical, but a little boring | | High (1.0) | High (1.0) | Wildly creative, possibly chaotic | | Medium (0.7) | Low (0.5) | Focused but fresh | | High (0.9) | Low (0.3) | Less diversity, but still picks odd words — often quirky or inconsistent | ### TL;DR Summary | Setting | Effect on Output | | -------------- | --------------------------------- | | **Low Temp** | More predictable, repetitive | | **High Temp** | More surprising, creative, varied | | **Low Top-p** | Narrow choices, more focused | | **High Top-p** | Wider range, more expressive | *** ### Step 5: Enable Capabilities (What Can Your Agent Do?) Your AI Agent can be enhanced with optional capabilities. These are *add-ons* that expand what your agent is able to handle. You’ll see three options: *** #### **1. Internet Access** Allows your agent to **search the web in real-time** to fetch up-to-date information. Use it for: * Research-based agents * Product comparison tools * Current event-related tasks > ⚠️ **Not needed** for tasks like essay grading, math solving, or document-based feedback. *** #### **2. Math Environment** Activates a special math engine so your AI Agent can: * Do advanced calculations * Solve equations step-by-step * Show math solutions clearly > ✅ **Best for:** Math tutors, test prep agents (SAT, GRE, etc.), homework checkers. *** #### **3. Knowledge Base** Allows your Agent to **refer to PDFs or documents** that you upload. This makes your Agent more accurate, detailed, and personalized to your specific way of working. We’ll explore this further in Step 6 (below), but here’s a quick preview: > ✅ **Great for:** Agents that grade, explain, coach, review, or answer in-depth questions using your material. *** ### Step 6: Add a Knowledge Base (To Train Your Agent with Your Own Material) A **Knowledge Base** is a collection of learning materials (usually PDFs) that your Agent can study and refer to when answering questions. This makes your Agent more powerful and context-aware. It's like giving your assistant a binder full of instructions, examples, and guidelines. *** #### When Should You Use a Knowledge Base? Use it if: * You want your Agent to follow a specific grading rubric * You have teaching guides, SOPs, FAQs, or resource manuals * You want the Agent to speak in your voice or follow your method *** #### How to Add One If you checked the **Knowledge Base** box in Step 5, you’ll now see the following options:
**Option 1: Select Existing Knowledge Base** If you’ve already uploaded materials before, simply choose from the dropdown list. *** **Option 2: Create a New Knowledge Base** Click **“Create New Knowledge Base”** and fill out the form:
| Field | What to Enter | | ---------------- | -------------------------------------------------------------------------------------------------------------------------------- | | **Name** | Give it a clear title, like “Essay Evaluation Knowledge Base” | | **Description** | Describe what the documents contain. For example: *“Includes essay rubrics, sample scoring guides, and evaluator instructions.”* | | **Tags** | Add searchable keywords like: essay, rubric, SAT, IELTS | | **Upload Files** | Upload PDFs (drag and drop or browse from your device). Max 10 files, total size up to 50 MB. | **Examples of Uploadable Documents (according to our example)**: * `essay-rubric.pdf` * `grading-criteria-handbook.pdf` * `sample-high-score-essay.pdf` Once you hit **Create**, your Knowledge Base will be attached to your Agent and ready to use. > **Now your AI Agent knows your way of doing things.** We are adding an example of knowledge base creation for our essay evaluator tool.
*** #### What Happens Next? Once added, your AI Agent will be able to: * Reference these documents while answering questions * Stay consistent with your standards and formats * Explain its logic using the material you provided ### Complete Setup Summary | Step | What You Did | | ---------- | --------------------------------------------------------------------------- | | **Step 1** | Opened the AI Agents dashboard | | **Step 2** | Chose whether to use “Generate with AI” or “Build Manually” | | **Step 3** | Entered core details: name, description, welcome message, and image | | **Step 4** | Selected the AI engine — either an existing AI Avatar or a predefined model | | **Step 5** | Enabled capabilities like internet access, math solving, or knowledge base | | **Step 6** | Attached or created a Knowledge Base with your own training materials | | **Step 7** | Saved and launched your Agent for live use | Your AI Agent is now ready — fully customized, purpose-built, and powered by your instructions and resources. Need more support? Our support team is available 24/7 through: * Email: --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.myvega.ai/ai-agents/set-up-your-ai-agent.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.