Skip to main content

Assistant AI

Overview​

The Assistants & Copilots API enables the creation of AI-driven agents that can interact in a conversational or task-oriented manner. These AI agents can be used for customer support chatbots, onboarding guides (interactive tutorials), or workflow automation helpers that understand commands. Outter’s Assistant AI Service covers these capabilities, allowing developers to integrate complex agent-like behavior via simple API calls.

Our assistant models are fine-tuned per use case, meaning they combine the power of general-purpose large language models (LLMs) with domain-specific knowledge. This approach allows each assistant to retrieve and synthesize the most relevant information from structured datasets, such as event databases for ticketing chatbots or curricula for education-focused assistants. By enriching responses with real-time, context-aware data, our assistants provide more accurate and actionable information.

Endpoints​

Outter supports various assistant endpoints. Typically, these are conversational in nature:

  • POST /api/v2/ai/assistants/chat – for a general-purpose chat assistant or support bot.
  • POST /api/v2/ai/assistants/{agent} – for specific types of agents (e.g., {agent} = onboarding for an onboarding copilot, or {agent} = support for a customer support agent).

Ticketing Chatbot: "What to Do in Barcelona Today?"​

A request from a ticketing platform's client application to provide recommendations for activities in Barcelona.

Request
POST /api/v2/ai/assistants/chat
Content-Type: application/json
Authorization: X-API-Key YOUR_API_KEY

{
  "user_id": "u7ef95ca7",
  "session_id": "chat-session-fe385582",
  "messages": [
    {"role": "user", "content": "What are some fun things to do in Barcelona today?"}
  ]
}
Response
{
  "reply": "Here are some exciting things to do in Barcelona today: \n1️⃣ Explore the **Sagrada Familia**, Gaudí’s architectural masterpiece. [Book Tickets](https://ticketblaster.eu/event/sagrada-familia-1245) \n2️⃣ Stroll down **La Rambla** and enjoy street performances. \n3️⃣ Visit **Park Güell** for stunning views. [Reserve Entry](https://ticketblaster.eu/event/park-guell-5623) \n4️⃣ Experience **local tapas** at Mercado de La Boqueria. \n5️⃣ Catch a **Flamenco show** for an authentic cultural experience. [Get Seats](https://ticketblaster.eu/event/flamenco-night-7834)",
  "session_id": "chat-session-fe385582",
  "agent": "ticketing_bot",
  "request_id": "req-chat-28d28e02",
  "model": "olivaw-ta2",
  "usage": {
    "tokens_used": 118,
    "timestamp": "2025-03-13T14:49:00Z"
  }
}

Onboarding Copilot: "Helping Teachers Create Engaging Online Courses"​

An onboarding assistant designed to guide teachers step-by-step in structuring an online course. The assistant retrieves curricula, best practices, and engagement strategies to help instructors craft an effective learning experience.

Request
POST /api/v2/ai/assistants/onboarding
Content-Type: application/json
Authorization: X-API-Key YOUR_API_KEY

{
  "user_id": "u9d27c63a",
  "session_id": "onboarding-session-786c08b0",
  "step": 1,
  "messages": [
    {"role": "system", "content": "You are an expert in online course creation. Guide the teacher through a structured onboarding process, one step at a time."},
    {"role": "user", "content": "How do I get started creating my course?"}
  ]
}
Response
{
  "step": 1,
  "reply": "Let's begin! First, define your **course topic and learning objectives**. \n1️⃣ Choose a **clear, engaging topic** that aligns with your expertise and your students’ needs.\n2️⃣ Define 3-5 **learning objectives** that describe what students will achieve by the end of the course. \n3️⃣ Identify your **target audience**: Who are your ideal students, and what is their skill level?\n\nOnce you've written this down, let me know, and we'll move to the next step!",
  "next_step": 2,
  "session_id": "onboarding-session-786c08b0",
  "agent": "teacher_onboarding_bot",
  "request_id": "req-onboard-83c912fa",
  "model": "olivaw-ea1",
  "usage": {
    "tokens_used": 152,
    "timestamp": "2025-03-13T14:50:00Z"
  }
}

Step-by-Step Interaction​

The onboarding assistant guides the teacher through structured steps, pulling in best practices, case studies, and examples to enrich each phase:

  1. Define Course Topic & Objectives (Step 1)
  2. Outline Course Modules & Lessons (Step 2)
  3. Create Interactive Content & Assignments (Step 3)
  4. Set Up Engagement Strategies (Quizzes, Discussion Forums) (Step 4)
  5. Finalize Course & Launch Plan (Step 5)

Each interaction prompts the user to confirm completion before moving to the next step, ensuring a logical and structured approach to course creation.

Best Practices​

  • Maintain Context: Always supply enough context in the messages history for the assistant to provide a relevant answer. Outter’s assistant service may maintain some state, but sending the recent conversation helps ensure continuity.
  • Handle Multi-turn Conversations: Design your client to handle back-and-forth interactions. The API will not automatically loop; the client must send each user message and display each assistant reply.
  • Monitor Behavior: AI assistants might sometimes produce irrelevant or incorrect answers. Monitor initial deployments and use Outter’s tools to refine the assistant (by tweaking prompts or providing example Q&A in the system context).
  • Agent Types: Outter offers different pre-configured agents (like a default support bot vs. an onboarding bot). Choose the appropriate endpoint or agent profile for your use case for best results, as they could be optimized for those scenarios.