Quick Start

This guide walks you through the essential steps to get a working AI application on Bivariant. By the end, you will have an installed app with a configured agent that you can interact with through the platform or embed on a website.

Prerequisites

• A Bivariant account. Sign up at bivariant.com if you don't have one.
• Access to an organization. New accounts are provisioned with a default organization and space.

Step 1: Navigate your workspace

After signing in, you land on your organization dashboard. The key areas in the interface:

• Apps — installed applications in the current space
• Blueprints — application templates you can build or edit
• Collections — structured data tables
• Conversations — messaging threads between users and agents

Your organization comes with a default space, which acts as an isolated environment for your projects. You can create additional spaces to separate concerns (for example, staging and production).

Step 2: Install an app

Apps are installed from blueprints — either ones you build yourself, or ones available from the marketplace.

1. Go to Apps in the sidebar.
2. Click Browse Marketplace.
3. Find a pre-built app (for example, "Customer Support Assistant") and click Install.

Installation takes a few seconds. During this process, the platform:

• Creates an app instance scoped to your organization
• Provisions service accounts for the app and its agents
• Sets up the data structures defined in the blueprint (collections, schemas)

Once installed, the app and its agents are ready to use.

Step 3: Configure an agent

Open your installed app and navigate to the Agents tab.

Select the agent to configure:

• Name and avatar — the identity shown to users in conversations
• Description — context for the AI model about the agent's role and behavior
• Knowledge base — documents and content the agent uses for retrieval-augmented generation (RAG)
• Default flow — the automation workflow triggered when a conversation starts

Add a knowledge source

1. In the agent settings, go to Knowledge Base.
2. Click Add Document and upload a file (PDF, text, or URL).
3. The platform processes the document, splits it into chunks, generates embeddings, and indexes it for semantic search.

The agent can now reference this content when answering questions.

Step 4: Create a flow

Flows define the step-by-step logic an agent follows during a conversation.

1. Go to Flows in the sidebar or within the app.
2. Click Create Flow.
3. Name your flow (for example, "Handle Support Inquiry").

The flow builder opens with a visual canvas. A basic flow includes:

• A Start step — entry point of the flow
• An Assistant step — calls the AI model with the conversation context and knowledge base
• Branching steps — route logic based on the assistant's output or user input
• An End step — completes the flow

Connect steps by dragging edges between them. Each step can be configured with:

• Input — data passed to the step (can reference outputs from previous steps)
• Timeout — maximum execution time
• Error handling — behavior on failure (retry, continue, or stop)

Save and assign this flow as the agent's default flow.

Step 5: Test the agent

Go to Conversations and start a new conversation:

1. Click New Conversation.
2. Add your agent as a participant.
3. Send a message.

The agent responds using the configured flow. It retrieves relevant content from the knowledge base, generates a response through the AI model, and sends it back in the conversation thread.

Step 6: Deploy as a web widget

To make the agent available to end users on your website, use the Presence Widget.

1. In your app, navigate to Channels.
2. Create a new web channel and note the Channel ID.
3. Add the following script to your website:

<script>
  var bv_channelId = "YOUR_CHANNEL_ID";
  var bv_language = "fr"; // "fr" or "en"
  (function () {
    var s = document.createElement("script");
    s.async = true;
    s.src = "https://cdn.bivariant.com/presence.js";
    var parentNode = document.head || document.body;
    parentNode.appendChild(s);
  })();
</script>

Replace YOUR_CHANNEL_ID with your actual channel ID. The widget appears on your website and connects visitors directly to your agent.

See Presence Widget for detailed configuration options.

Next steps

• Core Concepts — understand the platform's architecture and data model
• Agents — learn how to configure agents, knowledge bases, and AI behavior
• Flows — build more complex automation workflows
• Integrations — connect external APIs and services
• Collections — work with structured data

Advanced Features

• Build Custom Integrations: Connect to your CRM, helpdesk, or billing systems
• Create Complex Flows: Build sophisticated automation workflows
• Develop Custom Blueprints: Package your solutions for reuse

Enterprise Deployment

• User Management: Add team members and configure permissions
• Production Deployment: Scale for production workloads
• Security Configuration: Implement enterprise security policies

Getting Help

Need assistance? Here are your resources:

• Community Forum: Connect with other developers
• Documentation: Comprehensive guides and references
• Support Center: Direct help from our team
• Video Tutorials: Step-by-step visual guides

Welcome to the future of AI application development. You're now ready to build sophisticated AI solutions that deliver real business value!