Maker AI: The Ultimate AI Ecosystem for Delphi (Version 3.0)
Presented by Gustavo Enriquez
Core Problem
* AI integration is fragmented: Multiple APIs (OpenAI, Gemini, Claude, LLaMA, Grok) each with unique SDKs and syntax.
* Leads to high dependency and code obsolescence when models change.
* Delphi developers need stability and longevity, not constant rewrites.
Maker AI Solution
A complete AI ecosystem for Delphi, 100% native (no Python bridges), offering:
* Universal Connector for APIs:
* Single unified interface for all models (chat, video, audio).
* Central component: TAChatConnection (driver factory for AI).
* Includes optimized drivers for major models (OpenAI, Gemini, LLaMA).
* Extensible via TADBase class for future models.
* Dual Approach:
* High-level abstraction for most tasks.
* Direct access to specific drivers for advanced features (e.g., TAChatGemini).
Key Features
1. Automatic Fallback System (3 Levels):
* Native: Model supports feature.
* Internal: Switch to another model automatically.
* Programmatic: Developer-defined fallback via events.
2. File Handling:
* Intelligent orchestration for PDFs, images, etc.
* No extra components needed.
3. Prompt Management:
* TA Prompts: Versioned prompt libraries with dynamic parameters.
* Three-layer memory:
* Conversation history
* Persistent key-value memory
* System message for base instructions.
4. Function Calling (TA Functions):
* Expose Delphi functions to AI for real business logic integration.
* Example: `GetOrderStatus`, `GetStockLevel`.
5. External Integration:
* MCP (Model Context Protocol) support for connecting AI to external tools (Google Calendar, web browsing, file management).
* Works even with local models like LLaMA.
6. Enterprise Knowledge Management:
* RAG 2.0 Engine:
* Semantic search via TyraVector.
* Multi-provider embeddings.
* TyroGraph:
* Hybrid RAG + Knowledge Graph.
* Handles structured and unstructured data.
* Enables complex queries (e.g., shortest path between supplier and customer).
7. Agent Orchestration:
* TA Agent Manager: Design workflows with multiple AI agents collaborating.
8. Voice Interaction:
* TA Voice Monitor:
* Real-time speech-to-text.
* Keyword activation (like Alexa/Siri).
* Cross-platform.
9. UI Components:
* TAIChatBubble, TAIChatList, TAIChatInput for professional chat interfaces.
Hands-On Demo Highlights
* Basic Chat:
* Connected to local LLaMA model.
* Switched to universal connector for multiple providers (Claude, OpenAI, Gemini).
* Function Calling:
* Implemented `GetSystemDateTime` function.
* Advanced Features:
* Asynchronous responses.
* Image description and translation.
* JSON output and text block extraction.
* Web search with source links.
* Code interpreter (generated WAV tone, charts, Word templates).
* Multimedia Generation:
* Video generation (Gemini Veo 2 Plus).
* Image generation (Gemini 3, OpenAI).
* Voice generation and transcription.
* Enterprise Tools:
* RAG, Knowledge Graphs, MCP services.
* Integration with RAD Server/DataSnap.
Why It Matters
* Maker AI turns Delphi into a first-class AI platform:
* Unified architecture.
* Extensible and future-proof.
* Enables proactive, autonomous AI systems.
* Focus on business logic, not API complexity.
Closing Notes
* Maker AI Suite = Complete ecosystem for AI in Delphi.
* Includes demos and source code for immediate adoption.
* Community feedback: Highly positive (praised for innovation and presentation style).
See the rest of the sessions here.