07.2 applications

Applications

An Application is a self-contained AI service within a BindAI project.

Applications group together agents, workflows, configuration, and execution logic for a specific business purpose.

A single project may contain multiple independent applications that share common resources while serving different users or use cases.

---

What is an Application?

An application is the primary runtime unit inside a project.

Conceptually:

Project

↓

Application

↓

Agents

↓

Responses

Applications define how users interact with your AI system.

---

Why Applications?

Large AI systems often solve multiple business problems.

For example:

• Customer Support
• Internal Knowledge Assistant
• Sales Automation
• Document Analysis
• Code Review

Each of these can be implemented as a separate application within the same project.

---

Application Architecture

A typical application contains:

Application

├── Agents

├── Workflows

├── Configuration

└── Runtime

Applications may also access shared project resources such as tools and knowledge.

---

Multiple Applications

A project can host multiple applications simultaneously.

Project

├── Support App

├── HR Assistant

├── Finance Assistant

└── Analytics App

Each application operates independently while sharing common infrastructure.

---

Agents Inside Applications

Applications own one or more agents.

Application

↓

Support Agent

↓

Language Model

Each agent performs a specific reasoning task within the application’s domain.

---

Shared Resources

Applications can reuse project-level resources.

Examples include:

• shared tools
• shared workflows
• shared knowledge bases
• shared memory providers

This encourages consistency and avoids duplication.

---

Running an Application

Applications provide a simple execution interface.

Conceptually:

result = project.run(
    application="support",
    agent="assistant",
    message="Hello",
)

The application locates the requested agent and executes it.

---

Streaming Responses

Applications also support streaming output.

User

↓

Application

↓

Agent

↓

Streaming Response

Streaming is useful for chat interfaces where responses are displayed as they are generated.

---

Workflow Integration

Applications may invoke workflows instead of individual agents.

User Request

↓

Application

↓

Workflow

↓

Agents

↓

Result

This enables complex orchestration while keeping the application interface simple.

---

Configuration

Each application can have its own configuration.

Examples include:

• default model
• available agents
• enabled workflows
• provider settings
• application metadata

Application configuration remains isolated from other applications.

---

Separation of Responsibilities

Applications should focus on a single domain.

Good examples:

Customer Support

Human Resources

Finance

Engineering

Avoid creating one application that attempts to solve every problem.

---

Application vs Project

These concepts serve different purposes.

Project	Application
Top-level container	Individual AI service
Contains multiple applications	Contains agents and workflows
Shared resources	Domain-specific behavior
Infrastructure	Runtime entry point

Projects organize.

Applications execute.

---

Typical Structure

An application may look like this:

support/

├── agents/

├── workflows/

├── prompts/

├── configuration/

└── resources/

The exact layout depends on the project’s requirements.

---

Best Practices

• Create one application per business domain.
• Keep application responsibilities focused.
• Reuse shared project resources.
• Register only the agents needed by that application.
• Keep configuration local to the application whenever possible.
• Prefer several focused applications over one monolithic application.

---

Summary

Applications are the runtime services inside a BindAI project.

They organize agents, workflows, and configuration into focused AI solutions while sharing common project infrastructure such as tools, knowledge, and scheduling.