How to use OLLAMA and host from PC (OpenGPT FREE ALT.)

Resources Community Tutorials
#1

Setting Up Ollama 3.2 for Roblox Studio

Step 1: Install Ollama

Download and install Ollama 3.2 from the official website:
Ollama

Step 2: Create a Shortcut for Ollama Server

  1. After installation, create a new .ms shortcut.
  2. Right-click the shortcut, select Properties, and set the Target field to:
    C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exe -NoExit -Command "$env:OLLAMA_HOST='0.0.0.0'; ollama serve"
  3. Apply the changes and close the properties window.

Step 3: Configure Router Port Forwarding

  1. Open CMD and type:
    ipconfig
  2. Find your Default Gateway and IPv4 Address.
  3. Open a web browser and enter your gateway IP to access router settings.
  4. Log in and navigate to the Port Forwarding section.
  5. Forward port 11434 for your IPv4 Address.

Step 4: Install Cloudflare Warp

  1. Download and install Cloudflare Warp:
    Download WARP
  2. Enable 1.1.1.1 settings in Warp.
  3. In advanced settings, create a exclude split tunnel for your PUBLIC IPv4 Address (you can find it here What Is My IP Address - See Your Public Address - IPv4 & IPv6)`.

Step 5: Set Up Duck DNS

  1. Go to Duck DNS and create an account.
  2. Create a new domain.
  3. Set your PUBLIC IPv4 Address and Ollama Port as the IP to tunnel.
  4. Now, your custom domain will map to your actual IP.

Step 6: Configure Roblox Studio

  1. Open Roblox Studio and create a new game.
  2. Enable HTTP Requests in game settings.
  3. Add the following Lua script to enable interactions with Ollama AI:
local url = "http://[your-duck-domain-name].duckdns.org:11434/api/generate"

local HttpService = game:GetService("HttpService")
local Players = game:GetService('Players')
local replicatedStorage = game:GetService('ReplicatedStorage')
local llama = replicatedStorage.Llama

local chatService = game:GetService('Chat')
local ts = game:GetService('TextService')
local runService = game:GetService('RunService')

local function httpPost(plr, msg)
    local requestData = { 
        model = "llama3.2",
        system = "ChatBot",
        prompt = plr.DisplayName .. ": " .. msg,
        stream = false,
        parameters = {
            max_tokens = math.random(200, 400),
            temperature = 0.7,
            format = "json"
        },
    }
    
    local jsonData = HttpService:JSONEncode(requestData)
    local success, rawResponse = pcall(function()
        return HttpService:PostAsync(url, jsonData, Enum.HttpContentType.ApplicationJson)
    end)
    
    if rawResponse then
       local decodedResponse = HttpService:JSONDecode('[['..rawResponse..']]')
		local decodedResult = HttpService:JSONDecode(decodedResponse[1][1]["response"])
		print(decodedResult)
		local textResponse = decodedResult["response"]
		--print(decodedResult["response"])
		
		-- Remove any leftover asterisks (*)
		textResponse = textResponse:gsub("%}", "")
		
		-- Replace bold (**text**)
		textResponse = textResponse:gsub("%*(.-)%*", "<b>%1</b>")

		-- Replace italic (/text/)
		textResponse = textResponse:gsub("//(.-)//", "<i>%1</i>")

		-- Remove any leftover asterisks (*)
		textResponse = textResponse:gsub("%*", "")

		textResponse = textResponse:gsub("%\\", "")
		--endline
		textResponse = textResponse:gsub("%/n", "<br />")
		
		if textResponse ~= "" then
			local filteredData = ts:FilterStringAsync(systemName.Value .. ": " .. textResponse, plr.UserId, Enum.TextFilterContext.PublicChat) 

			for ii, vv in pairs(Players:GetChildren()) do
				llama:FireClient(vv,filteredData:GetNonChatStringForUserAsync(vv.UserId),systemColor.Value)
			end

			chatService:Chat(systemObject.Value, filteredData:GetNonChatStringForBroadcastAsync(), Enum.ChatColor[systemChatColor.Value])
		end
		
		return decodedResult
    else
        print("HTTP Request failed")
    end
end

Players.PlayerAdded:Connect(function(plr)
    plr.Chatted:Connect(function(message)
        httpPost(plr, message)
    end)
end)

PlayerScript

local replicatedStorage = game:GetService("ReplicatedStorage")
local llama = replicatedStorage.Llama
local ChatService = game:GetService("TextChatService")

local plr = game:GetService('Players').LocalPlayer

llama.OnClientEvent:Connect(function(responseData, color)
	
	local chatChannel = ChatService.TextChannels.RBXGeneral
	if chatChannel then
		local newCol = "<font color='rgb("..tostring(math.floor(color.R*255))..","..tostring(math.floor(color.G*255))..","..tostring(math.floor(color.B*255))..")'>"
		-- Use rich text to format the message in red
		local nameColon = string.find(responseData, ":")
		local Message = newCol .. string.sub(responseData, 1, nameColon)  .. "</font>" .. string.sub(responseData, nameColon+1, -1)
		decodedData = chatChannel:DisplaySystemMessage(Message)
	end
end)

VIDEO

Now your Roblox game can interact with Ollama 3.2 AI!

2 Likes