DelphiTools WebPages

MakerAI 3.0 provides the architecture to manage the entire lifecycle of complex AI applications, acting as the operating system for intelligence within your software.

- ?? **Orchestrate Autonomous Agents:** Create powerful agents that manage state, utilize advanced **Semantic Memory**, and execute complex workflows defined by the internal graph engine.
- ?? **RAG Engine 2.0 (Knowledge & Context):** Connect AI to your private data with the robust Retrieval-Augmented Generation (RAG) V2.0 engine, supporting massive indexing, multi-model environments, and hybrid embeddings.
- ??? **Enjoy 100% Native Performance:** MakerAI is pure Delphi. This ensures optimal performance, seamless integration, and compilation without heavy external dependencies across all supported Delphi versions (**10.3 to 12 Athens**).
- ?? **Universal LLM Connectivity (AiConnection V3):** The unified connector, `TAiConnection`, features enhanced connectivity and stability for seamless integration with OpenAI, Gemini, Claude, Ollama, Groq, and many more.
- ??? **Enterprise-Grade Architecture:** Ideal foundation for integrating AI into mission-critical systems like **ERP, CRM, LMS, and industrial solutions**.

---

## ? Key Features (Version 3.0 Modules)

MakerAI 3.0 introduces next-generation features focused on complexity and reliability.

### ?? RAG (Retrieval-Augmented Generation) Engine 2.0

The completely overhauled RAG engine allows the language models to use your private knowledge with unprecedented precision.

- **Massive Indexing:** Efficiently index millions of documents for enterprise-scale applications.
- **Hybrid Embeddings:** Utilize a combination of sparse and dense vectors for superior search relevance.
- **Multi-Model Support:** Seamlessly integrate RAG across different LLM backends (OpenAI, Gemini, Ollama, etc.).

### ?? Advanced Semantic Memory

- **Unified Memory Model:** New semantic memory layer deeply integrated with the RAG engine and the Workflow Graphs, allowing agents to form personalized, long-term context that transcends single conversations.
- **Short and Long-Term Recall:** Equip agents with the ability to recall specific past interactions and generalize long-term learnings.

### ?? Workflow Engine (Graphs) & Agents

- **Visual and Code-Based Orchestration:** Define complex logical flows by connecting nodes and edges. Agents follow this defined logic, utilizing memory and tools to achieve goals.

### ?? MakerAI Tools System (Function Calling)

- **Unified Tool Management:** A standardized and improved system for defining and exposing native Delphi functions (Tools) to the language model, enabling the AI to interact with your application's core logic (e.g., querying databases, sending commands).

### ??? Voice Monitoring & Progressive Transcription

- **Activation Commands:** Integrated support for "wake-word" detection.
- **Live Transcription:** Progressive audio transcription, ideal for real-time applications and low-latency interaction.

### ?? Professional Visual Components (ChatUI)

- **Chat UI:** Professional-grade, easy-to-use visual components for quickly building multimedia-integrated chat interfaces (text, images, audio, video).

### ?? MCP (Model Context Protocol) Support

- **Native MCP Client & Server:** Connect to or expose your Delphi application functions as standard MCP tools, ensuring seamless interoperability with other platforms.

### ?? Universal LLM Connector (AiConnection V3)

- **Enhanced Connectivity:** Improved performance and stability for accessing all major providers (OpenAI, Gemini, Claude, Groq, Ollama, Mistral).
- **Native Multimodal Support:** Process and generate text, images, and audio seamlessly.

---

## ??? MakerAI 3.0 Core Modules & Components

The suite is composed of non-visual and visual components designed for maximum modularity and ease of integration into the Delphi environment.

### ?? Universal Connectivity & LLMs (AiConnection V3)

| Component | Description |
|:---|:---|
| `TAIChatConnections` | Unified universal connector for managing messages, contexts, and sessions with any LLM provider. |
| `TAIChatOpenAi` | Driver for OpenAI (GPT-3.5, GPT-4, DALL-E). |
| `TAIChatClaude` | Driver for Anthropic Claude. |
| `TAIChatGemini` | Driver for Google Gemini. |
| `TAIChatGrok` | Driver for xAI Grok. |
| `TAIChatMistral` | Driver for Mistral AI. |
| `TAIChatOllama` | Driver for local and open-source models via Ollama. |
| `TAIChatDeepSeek` | Driver for DeepSeek AI. |
| `TAIChatGroq` | Driver for Groq (high-speed processing). |

### ?? Knowledge & Memory Engines (RAG)

| Component | Description |
|:---|:---|
| `RAGVector` | Vector and embedding-based Retrieval Augmented Generation (RAG) system. Ideal for massive indexing of unstructured documents. |
| `RAGGraph` | Graph-based RAG system for context retrieval based on graph database structure and relationships. |
| `GraphDB` | Component for administering and querying the graph-based knowledge database (for relational knowledge). |

### ?? Autonomous Agents & Workflow Orchestration

| Component | Description |
|:---|:---|
| `TAiAgentManager` | The central orchestration engine that manages the lifecycle and execution of workflows and agents. |
| `TAiAgent` | Defines an autonomous unit capable of managing state, memory, and executing complex tasks within the workflow. |
| `TAiAgentNode` | Represents a specific step or task within a workflow. |
| `TAiAgentLink` | Defines the logical connections and transitions between workflow nodes (tasks). |
| **Benefits** | Eliminates multitasking complexity, simplifies debugging, and ensures workflow extensibility. |

### ?? Tools & Protocol (Function Calling / MCP)

| Component | Description |
|:---|:---|
| `TAIFunctions` | Enables and manages the *Function Calling* system for the LLM to interact with native Delphi functions. |
| **MCP Client** | Client implementation within `TAIFunctions` to consume functions from other MCP servers. |
| `MCPServer` | Dedicated server to expose Delphi application functions as standardized tools via HTTP or StdIO. |

### ?? User Interface & Multimedia

| Component | Description |
|:---|:---|
| `VoiceMonitor` | Engine for real-time voice processing, including wake-word detection and progressive transcription. |
| `ChatInput` | Visual component for text and multimedia input in chat interfaces. |
| `ChatList` | Visual component acting as the main container to display conversation history. |
| `Bubbles` | Specialized visual components for rendering chat messages (text, images, videos) with a professional format. |

---