AI Copilot

AuraBot is AuraBoot's built-in AI assistant. It understands your application's metadata -- models, fields, records, and workflows -- and can answer natural language questions, generate reports, execute commands, and automate multi-step tasks.

Supported LLM Providers

AuraBot integrates with multiple LLM providers. Configure one or more in Settings > AI Configuration:

Provider	Models	Best For
OpenAI	GPT-4o, GPT-4o-mini	General-purpose queries, code generation
Anthropic	Claude 3.5 Sonnet, Claude 3 Opus	Complex reasoning, long-context analysis
Azure OpenAI	GPT-4, GPT-3.5	Enterprise environments with Azure compliance
Google	Gemini Pro, Gemini Flash	Multi-modal and fast inference
DeepSeek	DeepSeek V3, DeepSeek Coder	Cost-effective coding and analysis
Zhipu AI	GLM-4	Chinese language optimization
Ollama	Llama 3, Mistral, Qwen	Self-hosted, privacy-first deployments
Custom	Any OpenAI-compatible API	Self-hosted or specialized models

Natural Language Queries

Users can ask questions about their data in plain language:

"How many leads were created this month?"

AuraBot translates this into a database query using the metadata registry:

1. Identifies the crm_lead model
2. Generates a SQL query with the appropriate date filter
3. Returns the result in a formatted response

More examples:

• "Show me all overdue tasks assigned to John"
• "What is the total contract value for Q1?"
• "List the top 5 customers by revenue"

AuraBot automatically respects data permissions -- users only see results they are authorized to access.

Agent Control Plane

The Agent Control Plane (ACP) enables AuraBot to perform guarded multi-step operations:

How It Works

1. User Request -- "Create a new project called Q2 Marketing with 3 tasks"
2. Planning -- AuraBot decomposes the request into discrete steps
3. Execution -- Each step calls the appropriate command through the standard pipeline
4. Verification -- Results are validated before proceeding to the next step
5. Summary -- AuraBot reports what was created with links to the records

Supported Agent Actions

Action	Description
Query Data	Search and filter records across models
Execute Commands	Create, update, delete records via DSL commands
Generate Reports	Build ad-hoc reports with charts and tables
Analyze Trends	Compare metrics across time periods
Explain Workflows	Describe BPM process status and next steps
Suggest Actions	Recommend next steps based on record state

Agent Tool Contracts

AuraBot should call tools through explicit contracts, not by guessing internal APIs. A tool contract describes purpose, input schema, output shape, risk, confirmation policy, and side effects. This is what lets AuraBot choose between "query open tasks", "create draft task", and "transition task to approved" without treating all operations as generic updates.

Contract field	Why it matters
purpose	Helps the model choose the right tool
whenToUse	Narrows the valid business scenario
whenNotToUse	Prevents common misuse
inputContract	Gives the model strict argument structure
confirmationPolicy	Controls whether the user must approve
sideEffects	Makes write operations explicit

Plugin authors should add clear hints to commands and named queries that are intended for agent use.

Safety Guardrails

Agent actions are subject to the same permission system as manual operations:

• Commands execute through the full 20+ stage pipeline (auth, validation, audit)
• Destructive operations (delete, state transitions) require explicit user confirmation
• All agent actions are logged in the audit trail with source: "ai_agent"

Chat Interface

AuraBot is accessible from:

• Global Chat -- Click the AI icon in the top navigation bar
• Record Context -- Open AuraBot from any record detail page for context-aware assistance
• Inline Assist -- In form fields, use the AI icon to auto-fill or suggest values

Conversation Memory

AuraBot maintains conversation context within a session. Follow-up questions work naturally:

User: "Show me leads from last week" AuraBot: [displays 12 leads] User: "Which of those are from the website channel?" AuraBot: [filters to 5 leads with source=website]

Configuration

Basic Setup

1. Navigate to Settings > AI Configuration
2. Select a provider and enter the API key
3. Choose the default model
4. Set token limits and rate limiting

Advanced Options

Setting	Description	Default
maxTokensPerRequest	Maximum tokens per LLM call	4096
temperature	Creativity vs. determinism (0-1)	0.3
enableAgentMode	Allow guarded multi-step actions	true
confirmDestructive	Require confirmation for delete/transition	true
auditAllQueries	Log all AI interactions	true

Next Steps

• Agent System — Architecture behind AuraBot and agent execution
• Agent Workflows — Configure tools, providers, confirmations, and traces
• REST API — API endpoints for programmatic access
• BPM Workflows — Automate business processes
• Permissions — AI actions respect the permission model