Agents

AI Agents are intelligent assistants that can interact with your data, answer questions, perform tasks, and provide insights. They leverage large language models (LLMs) combined with tools and vector stores to provide context-aware, intelligent responses.

What are AI Agents?

AI Agents enable you to:

• Answer Questions: Answer questions about your data using natural language
• Perform Tasks: Execute tasks using AI tools (query data, generate reports, etc.)
• Provide Context: Use vector stores for RAG (Retrieval-Augmented Generation)
• Interact Naturally: Communicate using natural language conversations
• Access Your Data: Query and analyze your Objects, workflows, and reports
• Generate Insights: Provide insights and recommendations based on your data

Key Concepts

Agent Types

Agents can be categorized by type:

• User Agents: Custom agents created by users for specific purposes
• System Agents: Pre-built system agents available for use
• Template Agents: Template agents that can be used as starting points
• Internal Agents: Internal system agents (not exposed to users)

AI Providers

Agents use AI providers for language models:

• OpenAI: OpenAI models (GPT-4, GPT-3.5, etc.)
• Ollama: Local Ollama models
• VertexAI: Google Vertex AI models (Gemini, etc.)

Chat Models

Each provider offers different chat models:

• Model Selection: Choose appropriate model for your use case
• Capabilities: Different models have different capabilities
• Performance: Models vary in performance and cost

Vector Integration

Agents can use vectors for RAG:

• Context Retrieval: Retrieve relevant context from vector stores
• Semantic Search: Use semantic search to find relevant information
• Multiple Vectors: Configure multiple vectors for different context types
• Filtered Search: Filter vector searches to specific data sources

AI Tools

Agents can use AI tools to perform actions:

• Query Tools: Execute database queries on Objects
• Search Tools: Search Objects, workflows, and reports
• Report Tools: Generate and execute reports
• Custom Tools: Additional tools for specific functionality

Chat Memory

Agents can maintain conversation memory:

• Context Awareness: Remember previous conversation context
• Conversation History: Maintain conversation history
• Contextual Responses: Provide responses based on conversation history

How Agents Work

Agent Configuration

Agents are configured with:

1. Basic Information: Name, description, system prompt
2. AI Model: Provider and chat model selection
3. Vector Configuration: Vector stores for RAG
4. AI Tools: Tools the agent can use
5. Settings: Chat memory, chart rendering, etc.

Agent Execution

When an agent receives a query:

1. Query Processing: Agent processes the user's query
2. Context Retrieval: Retrieves relevant context from vectors (if configured)
3. Tool Selection: Determines which tools to use (if needed)
4. Tool Execution: Executes tools to gather information or perform actions
5. Response Generation: Generates response using LLM with context and tool results
6. Response Delivery: Returns response to user

RAG Process

When vectors are configured:

1. Query Analysis: Agent analyzes user query
2. Vector Search: Searches configured vectors for relevant context
3. Context Augmentation: Adds retrieved context to prompt
4. Response Generation: Generates response with context

Tool Usage

When tools are configured:

1. Tool Selection: Agent determines which tools to use
2. Tool Invocation: Calls tools with appropriate parameters
3. Result Processing: Processes tool results
4. Response Integration: Integrates tool results into response

Agent Features

Natural Language Interaction

• Conversational Interface: Chat-based interface for natural interaction
• Context Understanding: Understands context and conversation history
• Clarification: Asks clarifying questions when needed
• Multi-turn Conversations: Supports multi-turn conversations

Data Access

• Object Queries: Query Object data using natural language
• Workflow Information: Access workflow definitions and information
• Report Generation: Generate and execute reports
• Data Analysis: Analyze data and provide insights

Vector Integration

• RAG Support: Use vectors for Retrieval-Augmented Generation
• Semantic Search: Semantic search across your data
• Context-Aware: Context-aware responses using your data
• Multi-Vector Support: Configure multiple vectors

Tool Integration

• Query Execution: Execute database queries
• Data Retrieval: Retrieve data from Objects
• Report Generation: Generate reports dynamically
• Task Automation: Automate tasks using tools

Customization

• System Prompts: Customize agent behavior with system prompts
• Prompt Library: Create prompt library entries for common tasks
• Tool Selection: Choose which tools the agent can use
• Vector Configuration: Configure vectors for specific use cases

Using Agents

Agents can be:

• Created: Create custom agents for specific purposes
• Configured: Configure agents with vectors, tools, and settings
• Used: Interact with agents through chat interface
• Managed: View, edit, and manage agents
• Templated: Create agents from templates

Getting Started

1. Creating Agents: Learn how to create and configure agents
2. Using Agents: Understand how to interact with agents
3. Configuring Vectors for Agents: Learn how to configure vectors for RAG
4. Configure Tools: Select AI tools for your agents
5. Test Agents: Test agents with sample queries

Related Introduction

• Vectors Introduction - Understand vectors used for RAG with agents
• Objects Introduction - Learn about Objects that agents can query
• Workflows Introduction - Understand workflows that agents can access
• Reports Introduction - Learn about reports that agents can generate