Milestone: History → Future
Dear friends,
I believe this is a milestone that needs to be documented. Therefore, I am posting here – with my sincere thanks to Antonio, who has repeatedly shown us the simple path to using AI – a summary of a remarkable event:
“Curso virtual implementación IA en nuestras aplicaciones” (February 2025)
I am also re-posting the video from 2017, where Antonio gave us an early introduction to Artificial Intelligence, at a time when most of us could hardly imagine what it really meant.
https://mybergland.com/fwforum/meeting.mp4
Greetings to all and best regards,
Otto
1) General Summary of the Forum Article
This is the clean summary of the forum article “Curso virtual implementación IA en nuestras aplicaciones” (February 2025):
Background
Initiated by Leandro Arévalo (Colombia): because of the strong rise of AI, a virtual course (webinar) should be organized to learn how to integrate AI into Harbour/FiveWin applications.
Goal: hands-on introduction by Antonio Linares (FiveTech).
Organization
Format: Webinar, virtual via Skype.
Language: Spanish, with English translation if needed.
Timeframe: End of February 2025 (26–27 Feb).
Costs:
€50 – webinar only
€90 – webinar + FiveWin 32-bit with new AI classes
€100 – webinar + FiveWin 32- and 64-bit
Payment via PayPal to Antonio Linares, participation confirmed by payment.
Revenue was used to buy hardware for AI training.
Topics (Agenda)
Introduction to AI: description, types, models with Harbour/xHarbour.
New FiveWin classes:
TOpenAI (ChatGPT)
TDeepSeek (DeepSeek)
TOllama (local/private AI)
TGemini (Google Gemini)
AI Agents: building autonomous assistants inside Harbour/FWH.
SQL queries from natural language.
Tax and labor-law calculations.
Predictive analytics & forecasts.
Bookkeeping assistant (for entering invoices, categorization, error detection).
Fine-tuning models with own data (SFT, RL).
Local models: Phi4, Llama Vision, Qwen, DeepSeek.
Participants & Timeline
Aim: at least 20 participants.
Start 5pm Spanish time (other time zones listed).
Attendees from Spain, Latin America, USA, Turkey, Austria, etc.
Known names: Wilson Gamboa, Ralph del Castillo, Enrique Vertiz, Carlos Sincuir, Juan Navas, Lautaro Moreira, Otto Atzwanger and many more.
Over 40 paying participants.
Technical Setup
Install Ollama and download models (phi4, llama3.2-vision, deepseek-r1:14b).
~17 GB disk space required.
Recommended common path for FiveWin: C:\FW64IA to avoid path problems.
Execution
Webinar took place on 26 February 2025.
All topics were covered in one day.
Sessions were recorded and later published on YouTube.
Attendees received the newest FiveWin versions including the AI classes.
Conclusion
The webinar was a milestone:
Start of a new era “Harbour + AI”.
Many practical examples, strong participant engagement.
Discussions about opportunities, risks, and hardware requirements (e.g. NVIDIA GPUs).
👉 In short: This webinar was the first major organized introduction to AI for the Harbour/FiveWin community, with Antonio Linares leading. It demonstrated the integration of ChatGPT, DeepSeek, Ollama and other models into desktop apps, including AI agents, SQL generation, bookkeeping assistants and predictive analytics. It marked the beginning of a new direction for xHarbour/FiveWin development.
2) Expanded Topics
Here’s a more detailed breakdown of the webinar topics:
1. Introduction to AI in Harbour/xHarbour
Overview of current AI models (ChatGPT, DeepSeek, Gemini, Ollama).
Cloud vs local models → pros and cons (confidentiality, speed, hardware).
2. New FiveWin Classes
TOpenAI → ChatGPT integration (support dialogs, chat windows).
TDeepSeek → fast & affordable AI model access.
TOllama → run AI locally with Ollama (no cloud, keep private data).
TGemini → use Google Gemini (text, code, images).
3. AI Agents
How to make AI play an active role, not just Q&A.
Example: receive a client request, fetch DBF data, generate answer, send email.
4. SQL Queries from Natural Language
Example: “Clients from Lima with sales > 1000” → AI generates SQL query.
Benefit: non-technical users can query data.
5. Tax & Labor-Law Calculations
Examples: salary/leave calculations, tax deductions.
AI helps generate formulas instead of manually coding them.
6. Predictive Analytics & Forecasts
Compare budget goals vs. actual data.
Forecast sales, hotel occupancy, seasonal booking trends.
7. Bookkeeping Assistant
Demo: AI helps enter invoices.
Suggests account codes, VAT, detects duplicates.
Semi-automatic assistant to speed up accounting.
8. Fine-Tuning with Own Data
Training with company data via SFT or RL.
Goal: models specialized for internal domains (tourism, accounting).
9. Local Models (for confidential data)
Examples with Phi4, Llama Vision, Qwen, DeepSeek.
Installed and run via Ollama.
⚡ Key message: These integrations are ready now – developers can immediately build AI features into Harbour/FWH apps using the new classes.
3) Practical Demonstrations
These are the hands-on demos shown during the webinar:
1. Chat Window with AI
A simple FiveWin GUI window using TWebView2.
User types → request sent to AI (Gemini, ChatGPT, DeepSeek).
AI reply displayed in the same window.
2. Ollama (Local AI)
Install Ollama, download models (phi4, llama3.2-vision, deepseek-r1:14b).
Run from CMD:
ollama run phi4 --verbose
Advantage: works offline, safe for confidential data.
3. SQL from Natural Language
Demo:
Input: “Show me customers from Lima with sales > 1000”
Output:
SELECT * FROM clientes WHERE ciudad='Lima' AND ventas>1000
4. Bookkeeping Assistant
Enter invoices with AI support:
Reads invoice text, suggests account, VAT.
Alerts on duplicates.
5. Predictive Analytics
Compare actual vs. planned numbers.
Forecast trends (sales, hotel bookings, etc.).
6. Gemini Example Code
Antonio shared a FiveWin snippet:
7. Common Test Setup
Participants installed FiveWin in C:\FW64IA.
Ensured all could run Antonio’s live code samples without path issues.
8. Recording
The webinar was recorded.
Full video was later published on YouTube.
✅ Takeaway:
The webinar was highly practical, with working examples for:
AI chat windows
Natural-language SQL
Bookkeeping assistants
Ollama local AI
Forecasting & analytics
It demonstrated that Harbour/FiveWin + AI is already fully usable today.
Dear friends,
I believe this is a milestone that needs to be documented. Therefore, I am posting here – with my sincere thanks to Antonio, who has repeatedly shown us the simple path to using AI – a summary of a remarkable event:
“Curso virtual implementación IA en nuestras aplicaciones” (February 2025)
I am also re-posting the video from 2017, where Antonio gave us an early introduction to Artificial Intelligence, at a time when most of us could hardly imagine what it really meant.
https://mybergland.com/fwforum/meeting.mp4
Greetings to all and best regards,
Otto
1) General Summary of the Forum Article
This is the clean summary of the forum article “Curso virtual implementación IA en nuestras aplicaciones” (February 2025):
Background
Initiated by Leandro Arévalo (Colombia): because of the strong rise of AI, a virtual course (webinar) should be organized to learn how to integrate AI into Harbour/FiveWin applications.
Goal: hands-on introduction by Antonio Linares (FiveTech).
Organization
Format: Webinar, virtual via Skype.
Language: Spanish, with English translation if needed.
Timeframe: End of February 2025 (26–27 Feb).
Costs:
€50 – webinar only
€90 – webinar + FiveWin 32-bit with new AI classes
€100 – webinar + FiveWin 32- and 64-bit
Payment via PayPal to Antonio Linares, participation confirmed by payment.
Revenue was used to buy hardware for AI training.
Topics (Agenda)
Introduction to AI: description, types, models with Harbour/xHarbour.
New FiveWin classes:
TOpenAI (ChatGPT)
TDeepSeek (DeepSeek)
TOllama (local/private AI)
TGemini (Google Gemini)
AI Agents: building autonomous assistants inside Harbour/FWH.
SQL queries from natural language.
Tax and labor-law calculations.
Predictive analytics & forecasts.
Bookkeeping assistant (for entering invoices, categorization, error detection).
Fine-tuning models with own data (SFT, RL).
Local models: Phi4, Llama Vision, Qwen, DeepSeek.
Participants & Timeline
Aim: at least 20 participants.
Start 5pm Spanish time (other time zones listed).
Attendees from Spain, Latin America, USA, Turkey, Austria, etc.
Known names: Wilson Gamboa, Ralph del Castillo, Enrique Vertiz, Carlos Sincuir, Juan Navas, Lautaro Moreira, Otto Atzwanger and many more.
Over 40 paying participants.
Technical Setup
Install Ollama and download models (phi4, llama3.2-vision, deepseek-r1:14b).
~17 GB disk space required.
Recommended common path for FiveWin: C:\FW64IA to avoid path problems.
Execution
Webinar took place on 26 February 2025.
All topics were covered in one day.
Sessions were recorded and later published on YouTube.
Attendees received the newest FiveWin versions including the AI classes.
Conclusion
The webinar was a milestone:
Start of a new era “Harbour + AI”.
Many practical examples, strong participant engagement.
Discussions about opportunities, risks, and hardware requirements (e.g. NVIDIA GPUs).
2) Expanded Topics
Here’s a more detailed breakdown of the webinar topics:
1. Introduction to AI in Harbour/xHarbour
Overview of current AI models (ChatGPT, DeepSeek, Gemini, Ollama).
Cloud vs local models → pros and cons (confidentiality, speed, hardware).
2. New FiveWin Classes
TOpenAI → ChatGPT integration (support dialogs, chat windows).
TDeepSeek → fast & affordable AI model access.
TOllama → run AI locally with Ollama (no cloud, keep private data).
TGemini → use Google Gemini (text, code, images).
3. AI Agents
How to make AI play an active role, not just Q&A.
Example: receive a client request, fetch DBF data, generate answer, send email.
4. SQL Queries from Natural Language
Example: “Clients from Lima with sales > 1000” → AI generates SQL query.
Benefit: non-technical users can query data.
5. Tax & Labor-Law Calculations
Examples: salary/leave calculations, tax deductions.
AI helps generate formulas instead of manually coding them.
6. Predictive Analytics & Forecasts
Compare budget goals vs. actual data.
Forecast sales, hotel occupancy, seasonal booking trends.
7. Bookkeeping Assistant
Demo: AI helps enter invoices.
Suggests account codes, VAT, detects duplicates.
Semi-automatic assistant to speed up accounting.
8. Fine-Tuning with Own Data
Training with company data via SFT or RL.
Goal: models specialized for internal domains (tourism, accounting).
9. Local Models (for confidential data)
Examples with Phi4, Llama Vision, Qwen, DeepSeek.
Installed and run via Ollama.
3) Practical Demonstrations
These are the hands-on demos shown during the webinar:
1. Chat Window with AI
A simple FiveWin GUI window using TWebView2.
User types → request sent to AI (Gemini, ChatGPT, DeepSeek).
AI reply displayed in the same window.
2. Ollama (Local AI)
Install Ollama, download models (phi4, llama3.2-vision, deepseek-r1:14b).
Run from CMD:
ollama run phi4 --verbose
Advantage: works offline, safe for confidential data.
3. SQL from Natural Language
Demo:
Input: “Show me customers from Lima with sales > 1000”
Output:
SELECT * FROM clientes WHERE ciudad='Lima' AND ventas>1000
4. Bookkeeping Assistant
Enter invoices with AI support:
Reads invoice text, suggests account, VAT.
Alerts on duplicates.
5. Predictive Analytics
Compare actual vs. planned numbers.
Forecast trends (sales, hotel bookings, etc.).
6. Gemini Example Code
Antonio shared a FiveWin snippet:
#include "FiveWin.ch"
function Main()
local oWnd, oWebView
local oGemini := TGemini():New()
fw_SetUnicode( .F. )
DEFINE WINDOW oWnd TITLE "Chat AI" SIZE 650, 800
oWebView = TWebView2():New( oWnd )
oWebView:SetHtml( hb_memoRead( "chat.html" ) )
oWebView:bOnBind = { | aInfo, cAnswer | ;
oGemini:Send( aInfo[ 1 ] ), ;
cAnswer := "sendResponse('" + hb_Utf8ToStr( oGemini:getValue() ) + "')", ;
cAnswer := StrTran( cAnswer, Chr( 10 ), "" ), ;
oWebView:Eval( cAnswer ) }
ACTIVATE WINDOW oWnd CENTER ;
ON RESIZE oWebView:SetSize( nWidth, nHeight )
oGemini:End()
return nilParticipants installed FiveWin in C:\FW64IA.
Ensured all could run Antonio’s live code samples without path issues.
8. Recording
The webinar was recorded.
Full video was later published on YouTube.
The webinar was highly practical, with working examples for:
AI chat windows
Natural-language SQL
Bookkeeping assistants
Ollama local AI
Forecasting & analytics
It demonstrated that Harbour/FiveWin + AI is already fully usable today.
