AI Assistant

The AI Assistant is a built-in chat panel within HarbourBuilder that connects to Ollama for local AI-powered coding help. No API keys, no cloud services, no data leaving your machine — just a local model providing instant assistance while you code.

Supported Backends

Model Selector

The AI Assistant includes a model selector dropdown that lists all available models detected from your running Ollama or LM Studio instance. Available options typically include:

Chat Interface

The AI Assistant panel provides a familiar chat experience:

• Chat history — scrollable conversation view with your messages and AI responses.
• Input field — type your question or paste code for analysis.
• Send button — submit your message to the model.
• Streaming responses — watch the response appear token by token.
• Code blocks — generated code is displayed in formatted, syntax-highlighted blocks with a copy button.
• Model selector — switch models without restarting.
• Clear chat — reset the conversation history.

Example Prompts

Here are some effective prompts to try with the AI Assistant:

Using TOllama in Your Code

The same TOllama component that powers the AI Assistant is available in your own applications. Here's how to use it:

Basic Usage

#include "hbbuilder.ch"

function Main()

   local oOllama, cResponse

   DEFINE OLLAMA oOllama ;
      HOST "localhost" ;
      PORT 11434 ;
      MODEL "llama3"

   cResponse := oOllama:Chat( "What is Harbour programming language?" )

   ? cResponse

return nil

Streaming Response

static function StreamExample( oOllama, oMemo )

   local cPrompt := "Write a Harbour function to calculate fibonacci"

   oMemo:Append( "Generating...\n\n" )

   // Stream the response token by token into a Memo
   oOllama:ChatStream( cPrompt, ;
      { |cToken| oMemo:Append( cToken ) } )

return nil

Setting a System Prompt

oOllama:SetSystem( "You are a Harbour/xBase expert. ;
   Always provide complete, working code examples. ;
   Explain your reasoning clearly." )

cResponse := oOllama:Chat( "How do I read a file line by line?" )

Using LM Studio

LM Studio provides a compatible API. Simply point TOllama at the LM Studio server:

oOllama := TOllama():New()
oOllama:cHost := "localhost"
oOllama:nPort := 1234  // LM Studio default port
oOllama:cModel := "local-model"

Privacy Benefits

• No data exfiltration — your source code, business logic, and API keys never leave your machine.
• No API keys to manage — local models require no authentication or billing setup.
• Works offline — once models are downloaded, no internet connection is required.
• No rate limits — unlimited queries with no usage caps or per-token costs.
• Enterprise-safe — compliant with policies that prohibit sending code to external AI services.

Configuration

Open the AI Assistant panel from the menu Tools > AI Assistant or use the toolbar button. The settings gear icon lets you configure: