botserver/docs/src/chapter-06/architecture.md

Architecture Overview

Auto Bootstrap Process

The Auto Bootstrap process is responsible for initializing and configuring the entire BotServer environment after installation. It ensures that all system components are installed, configured, and started automatically, and that bots are created from predefined templates.

1. Bootstrap Initialization

The process begins with the BootstrapManager, which is instantiated with an installation mode (Local or Container) and an optional tenant name. It initializes the PackageManager, which detects the operating system and sets up the base installation path (e.g., /opt/gbo or botserver-stack).

2. Component Registration and Installation

The PackageManager registers all system components such as:

• tables (PostgreSQL database)
• cache (Valkey/Redis)
• drive (MinIO object storage)
• llm (local LLM server)
• email, proxy, directory, alm, dns, meeting, table_editor, doc_editor, desktop, devtools, bot, system, vector_db, host

Each component has a ComponentConfig defining:

• Ports and dependencies
• Download URLs and binaries
• Pre/post-install commands
• Environment variables
• Execution commands

During bootstrap, required components (tables, drive, cache) are installed and started automatically.
For example:

• The tables component generates secure database credentials, writes them to .env, and applies SQL migrations to initialize the schema.
• The drive component creates secure credentials and stores them encrypted in the database.

3. Bot Configuration

After components are installed, the bootstrap process updates the bot configuration in the database.
The method update_bot_config() ensures each component’s configuration is linked to a bot record in the bot_configuration table.
If no bot exists, a new UUID is generated to associate configuration entries.

4. Template-Based Bot Creation

The method create_bots_from_templates() scans the templates/ directory for folders ending in .gbai (e.g., default.gbai, announcements.gbai).
Each .gbai folder represents a bot template.

For each template:

• The folder name is converted into a human-readable bot name (e.g., default.gbai → “Default”).
• If the bot doesn’t exist in the bots table, a new record is inserted with:
  • Default LLM provider (openai)
  • Default configuration ({"model": "gpt-4", "temperature": 0.7})
  • Context provider (database)
  • Active status (true)

This automatically creates bots from templates during bootstrap.

5. Template Upload to MinIO

After bots are created, the method upload_templates_to_minio() uploads all template files recursively to a MinIO bucket (S3-compatible storage).
This makes templates accessible for runtime bot operations and ensures persistence across environments.

6. Summary

The Auto Bootstrap process performs the following steps automatically:

1. Detects environment and installation mode.
2. Registers and installs required components.
3. Initializes the database and applies migrations.
4. Updates bot configuration records.
5. Creates bots from .gbai templates.
6. Uploads templates to MinIO for storage.

This process ensures that after installation, the system is fully operational with preconfigured bots derived from templates, ready to serve requests immediately.