3. Script Language > AI - Artificial Intelligence Commands > !Calling Any API Samples > !LM-Studio

Overview

This script demonstrates how to process an image and use the LM Studio API to analyze and describe it using an AI model.

The script:

1.- Loads an input image

2.- Resizes it to 256x256 pixels

3.- Encodes it to BASE64

4.- Sends it to LM Studio's API

5.- Returns an AI

6.- generated description

About LM Studio

LM Studio is a free, open-source desktop application that allows users to discover, download, and run large language models (LLMs) locally on Windows, macOS, or Linux. It provides a user-friendly GUI to interact with models from Hugging Face, ensuring data privacy by keeping everything offline.

The Model: Gemma-3-27b-it and gemma-3-27b-it@q8_0

The script uses 'gemma-3-27b-it@q8_0' as the AI model. Gemma is a family of lightweight, open-source LLMs developed by Google, designed for research and efficient local deployment. The '3' indicates the third generation, '27b' refers to 27 billion parameters, and 'it' denotes an instruction-tuned variant optimized for tasks like image analysis. '@q8_0' specifies 8-bit quantization, reducing memory usage while maintaining performance, ideal for local execution on consumer hardware.

Script Code uploading an Image to have it analyzed by the AI

' Variable Definitions:

$$IMX=?path\T2.jpg

$$IMK=?path\T5.jpg

$$MOD=gemma-3-27b-it@q8_0

' Input image path, output image path, and AI model specification

' Image Processing:

ANA.Load|1|$$IMX

ANA.Resize To|1|2|256|256

ANA.Save|2|$$IMK

'ANA.Show|2!

' Loads input image, resizes to 256x256, saves processed image (display commented out)

' API Setup:

$$URB=http://localhost:1234/

$$REA=v1/chat/completions

VAV.$$URL=$$URB$$REA

HTP.InitHdr|-|-

HTP.AddHdr|Content-Type|application/json

HTP.AddHdr|Authorization|Bearer -

' Sets up base URL, endpoint, and HTTP headers

' AI Configuration:

$$SYS=You are a helpful assistant that can analyze images and provide click coordinates.

$$USR=Describe the picture.

' System and user prompts for AI

' Image Encoding:

CFF.$$IMK|$$IMZ

AIC.Encode to BASE64|$$IMZ|$$B64

$$PRE=data:image/jpeg;base64,

VAV.$$IMG=$$PRE$$B64

' Encodes image to BASE64 with MIME prefix

' JSON Preparation:

' Escape the System Prompt

AIC.ESC|$$SYS|$$ESS

' Escape strings for JSON using AIC.ESC

AIC.ESC|$$USR|$$ESU

AIC.ESC|$$IMG|$$ESI

VAV.$$CNT=$$ESU $$ESI

VAV.$$JSO={"model": "$$MOD", "messages": [{"role": "system", "content": "$$ESS"},

VAV.$$JSO+ {"role": "user", "content": "$$CNT"}], "max_tokens": 100, "temperature": 0.7, "stream": false}

' Prepares JSON payload with escaped strings

' HTTP Request:

$$CMD=POST

HTP.Setdebug|0

' Must be large because the Model may need >10 Minutes to generate an answer

' Sends POST request with 3600s timeout

HTP.SetTO|3600000

HTP.Request|$$CMD|$$URL|$$JSO|0|0|0|0|$$RET

' Response Handling:

' Parses response, shows description, and cleans up

PSJ.Parse|-|$$DOC

$$PAT=choices[0].message.content

PSJ.GetStr|$$DOC|$$PAT|$$TEX

MBX.$$TEX

PSJ.GetLastResultW|$$TEA

CLB.Set Unitext|$$TEA

PSJ.FreeAll

ENR.

Script Code 2:

$$IMK=?path\T5.jpg

$$MOD=gemma-3-27b-it@q8_0

STW.ct|Chrome_WidgetWin_1|X - Brave

'SWP.67,60|1714,1840|T

ANA.Snap|1

ANA.Save|1|$$IMK

'ANA.Load|1|$$IMX

'ANA.Resize To|1|2|256|256

'ANA.Save|2|$$IMK

$$URB=http://localhost:1234/

$$REA=v1/chat/completions

VAV.$$URL=$$URB$$REA

HTP.InitHdr|-|-

HTP.AddHdr|Content-Type|application/json

HTP.AddHdr|Authorization|Bearer -

$$SYS=You are a helpful assistant that can analyze and describe

$$SYS+images and provide click coordinates.

$$USR=Describe the picture.

CFF.$$IMK|$$IMZ

AIC.Encode to BASE64|$$IMZ|$$B64

$$PRE=data:image/jpeg;base64,

VAV.$$IMG=$$PRE$$B64

' Escape strings for JSON using AIC.ESC

AIC.ESC|$$USR|$$ESU

AIC.ESC|$$SYS|$$ESS

AIC.ESC|$$IMG|$$ESI

VAV.$$CNT=$$ESU $$ESI

VAR.$$JSO={"model": "$$MOD", "messages":

VAR.$$JSO+[{"role": "system", "content": "$$ESS"},

VAR.$$JSO+ {"role": "user", "content": "$$CNT"}],

VAR.$$JSO+ "max_tokens": 91000, "temperature": 0.7,

VAR.$$JSO+ "stream": false}

'LEN.$$JSO|$$JSL

$$CMD=POST

HTP.Setdebug|0

HTP.SetTO|3600000

HTP.Request|$$CMD|$$URL|$$JSO|0|0|0|0|$$RET

PSJ.Parse|-|$$DOC

$$PAT=choices[0].message.content

PSJ.GetStr|$$DOC|$$PAT|$$TEX

MBX.$$TEX

PSJ.GetLastResultW|$$TEA

CLB.Set Unitext|$$TEA

PSJ.FreeAll

ENR.

Requirements and Downloads

Requirements:- LM Studio running on localhost:1234- Input image at specified path- Gemma-3-27b-it model available

Download LM Studio from these official sources (as of March 21, 2025):

- Official Website: https://lmstudio.ai (Windows, macOS, Linux versions)

- TechSpot: https://www.techspot.com/downloads/7798-lm-studio.html (Latest version download)

System Requirements:- 16GB+ RAM recommended- 6GB+ VRAM for GPU support (NVIDIA/AMD)- AVX2 processor support- macOS 13.6+ for Apple Silicon (M1/M2/M3)