Add token counter to the chat

app/MindWork AI Studio/Assistants/I18N/allTexts.lua

@@ -1618,6 +1618,9 @@ UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3403290862"] = "The selec
 -- Select a provider first
 UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3654197869"] = "Select a provider first"
 
+-- Estimated amount of tokens:
+UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T377990776"] = "Estimated amount of tokens:"
+
 -- Start new chat in workspace '{0}'
 UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3928697643"] = "Start new chat in workspace '{0}'"
 
@@ -5095,6 +5098,9 @@ UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1019424746"] = "Startup log file
 -- Browse AI Studio's source code on GitHub — we welcome your contributions.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1107156991"] = "Browse AI Studio's source code on GitHub — we welcome your contributions."
 
+-- The Tokenizer library serves as the base framework for integrating the DeepSeek tokenizer.
+UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1132433749"] = "The Tokenizer library serves as the base framework for integrating the DeepSeek tokenizer."
+
 -- ID mismatch: the plugin ID differs from the enterprise configuration ID.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1137744461"] = "ID mismatch: the plugin ID differs from the enterprise configuration ID."
 
@@ -5329,6 +5335,9 @@ UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T566998575"] = "This is a library
 -- Used .NET SDK
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T585329785"] = "Used .NET SDK"
 
+-- We use the DeepSeek Tokenizer to estimate the number of tokens an input will generate.
+UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T591393704"] = "We use the DeepSeek Tokenizer to estimate the number of tokens an input will generate."
+
 -- This library is used to manage sidecar processes and to ensure that stale or zombie sidecars are detected and terminated.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T633932150"] = "This library is used to manage sidecar processes and to ensure that stale or zombie sidecars are detected and terminated."
 

app/MindWork AI Studio/Components/ChatComponent.razor

@@ -34,7 +34,7 @@
     </ChildContent>
     <FooterContent>
         <MudElement Style="flex: 0 0 auto;">
-            <MudTextField
+            <UserPromptComponent
                 T="string"
                 @ref="@this.inputField"
                 @bind-Text="@this.userInput"
@@ -50,8 +50,11 @@
                 Disabled="@this.IsInputForbidden()"
                 Immediate="@true"
                 OnKeyUp="@this.InputKeyEvent"
+                WhenTextChangedAsync="@(_ =>this.CalculateTokenCount())"
                 UserAttributes="@USER_INPUT_ATTRIBUTES"
                 Class="@this.UserInputClass"
+                DebounceTime="TimeSpan.FromSeconds(1)" 
+                HelperText="@this.TokenCountMessage"
                 Style="@this.UserInputStyle"/>
         </MudElement>
         <MudToolBar WrapContent="true" Gutters="@false" Class="border border-solid rounded" Style="border-color: lightgrey;">
@@ -124,7 +127,6 @@
                     <MudIconButton Icon="@Icons.Material.Filled.Error" Color="Color.Error"/>
                 </MudTooltip>
             }
-            <MudIconButton />
         </MudToolBar>
     </FooterContent>
 </InnerScrolling>

app/MindWork AI Studio/Components/ChatComponent.razor.cs

@@ -3,6 +3,7 @@
 using AIStudio.Provider;
 using AIStudio.Settings;
 using AIStudio.Settings.DataModel;
+using AIStudio.Tools.Services;
 
 using Microsoft.AspNetCore.Components;
 using Microsoft.AspNetCore.Components.Web;
@@ -36,6 +37,9 @@ public partial class ChatComponent : MSGComponentBase, IAsyncDisposable
 
     [Inject]
     private IDialogService DialogService { get; init; } = null!;
+
+    [Inject] 
+    private RustService RustService { get; init; } = null!;
 
     private const Placement TOOLBAR_TOOLTIP_PLACEMENT = Placement.Top;
     private static readonly Dictionary<string, object?> USER_INPUT_ATTRIBUTES = new();
@@ -58,10 +62,12 @@ public partial class ChatComponent : MSGComponentBase, IAsyncDisposable
     private Guid currentWorkspaceId = Guid.Empty;
     private CancellationTokenSource? cancellationTokenSource;
     private HashSet<FileAttachment> chatDocumentPaths = [];
+    private string tokenCount = "0";
+    private string TokenCountMessage => $"{this.T("Estimated amount of tokens:")} {this.tokenCount}";
 
     // Unfortunately, we need the input field reference to blur the focus away. Without
     // this, we cannot clear the input field.
-    private MudTextField<string> inputField = null!;
+    private UserPromptComponent<string> inputField = null!;
 
     #region Overrides of ComponentBase
 
@@ -405,6 +411,9 @@ private async Task InputKeyEvent(KeyboardEventArgs keyEvent)
         // Was a modifier key pressed as well?
         var isModifier = keyEvent.AltKey || keyEvent.CtrlKey || keyEvent.MetaKey || keyEvent.ShiftKey;
 
+        if (isEnter)
+            await this.CalculateTokenCount();
+
         // Depending on the user's settings, might react to shortcuts:
         switch (this.SettingsManager.ConfigurationData.Chat.ShortcutSendBehavior)
         {
@@ -523,6 +532,7 @@ private async Task SendMessage(bool reuseLastUserPrompt = false)
         this.userInput = string.Empty;
         this.chatDocumentPaths.Clear();
         await this.inputField.BlurAsync();
+        this.tokenCount = "0";
 
         // Enable the stream state for the chat component:
         this.isStreaming = true;
@@ -901,6 +911,20 @@ private Task EditLastBlock(IContent block)
         return Task.CompletedTask;
     }
 
+    private async Task CalculateTokenCount()
+    {
+        if (this.inputField.Value is null)
+        {
+            this.tokenCount = "0";
+            return;
+        }
+        var response = await this.RustService.GetTokenCount(this.inputField.Value);
+        if (response is null)
+            return;
+        this.tokenCount = response.TokenCount.ToString();
+        this.StateHasChanged();
+    }
+
     #region Overrides of MSGComponentBase
 
     protected override async Task ProcessIncomingMessage<T>(ComponentBase? sendingComponent, Event triggeredEvent, T? data) where T : default

app/MindWork AI Studio/Components/UserPromptComponent.cs

@@ -0,0 +1,68 @@
+using Microsoft.AspNetCore.Components;
+using Timer = System.Timers.Timer;
+
+namespace AIStudio.Components;
+
+/// <summary>
+/// Debounced multi-line text input built on <see cref="MudTextField{T}"/>.
+/// Keeps the base API while adding a debounce timer.
+/// Callers can override any property as usual.
+/// </summary>
+public class UserPromptComponent<T> : MudTextField<T>
+{
+    [Parameter]
+    public TimeSpan DebounceTime { get; set; } = TimeSpan.FromMilliseconds(800);
+
+    [Parameter]
+    public Func<string, Task> WhenTextChangedAsync { get; set; } = _ => Task.CompletedTask;
+
+    private readonly Timer debounceTimer = new();
+    private string text = string.Empty;
+    private string lastParameterText = string.Empty;
+    private string lastNotifiedText = string.Empty;
+    private bool isInitialized;
+
+    protected override async Task OnInitializedAsync()
+    {
+        this.text = this.Text ?? string.Empty;
+        this.lastParameterText = this.text;
+        this.lastNotifiedText = this.text;
+        this.debounceTimer.AutoReset = false;
+        this.debounceTimer.Interval = this.DebounceTime.TotalMilliseconds;
+        this.debounceTimer.Elapsed += (_, _) =>
+        {
+            this.debounceTimer.Stop();
+            if (this.text == this.lastNotifiedText)
+                return;
+
+            this.lastNotifiedText = this.text;
+            this.InvokeAsync(async () => await this.TextChanged.InvokeAsync(this.text));
+            this.InvokeAsync(async () => await this.WhenTextChangedAsync(this.text));
+        };
+
+        this.isInitialized = true;
+        await base.OnInitializedAsync();
+    }
+
+    protected override async Task OnParametersSetAsync()
+    {
+        // Ensure the timer uses the latest debouncing interval:
+        if (!this.isInitialized)
+            return;
+
+        if(Math.Abs(this.debounceTimer.Interval - this.DebounceTime.TotalMilliseconds) > 1)
+            this.debounceTimer.Interval = this.DebounceTime.TotalMilliseconds;
+
+        // Only sync when the parent's parameter actually changed since the last change:
+        if (this.Text != this.lastParameterText)
+        {
+            this.text = this.Text ?? string.Empty;
+            this.lastParameterText = this.text;
+        }
+
+        this.debounceTimer.Stop();
+        this.debounceTimer.Start();
+
+        await base.OnParametersSetAsync();
+    }
+}

app/MindWork AI Studio/Pages/Information.razor

@@ -291,6 +291,8 @@
                     <ThirdPartyComponent Name="sysinfo" Developer="Guillaume Gomez & Open Source Community" LicenseName="MIT" LicenseUrl="https://github.com/GuillaumeGomez/sysinfo/blob/main/LICENSE" RepositoryUrl="https://github.com/GuillaumeGomez/sysinfo" UseCase="@T("This library is used to manage sidecar processes and to ensure that stale or zombie sidecars are detected and terminated.")"/>
                     <ThirdPartyComponent Name="tempfile" Developer="Steven Allen, Ashley Mannix & Open Source Community" LicenseName="MIT" LicenseUrl="https://github.com/Stebalien/tempfile/blob/master/LICENSE-MIT" RepositoryUrl="https://github.com/Stebalien/tempfile" UseCase="@T("This library is used to create temporary folders for saving the certificate and private key for communication with Qdrant.")"/>
                     <ThirdPartyComponent Name="Lua-CSharp" Developer="Yusuke Nakada & Open Source Community" LicenseName="MIT" LicenseUrl="https://github.com/nuskey8/Lua-CSharp/blob/main/LICENSE" RepositoryUrl="https://github.com/nuskey8/Lua-CSharp" UseCase="@T("We use Lua as the language for plugins. Lua-CSharp lets Lua scripts communicate with AI Studio and vice versa. Thank you, Yusuke Nakada, for this great library.")" />
+                    <ThirdPartyComponent Name="DeepSeek-V3.2 Tokenizer" Developer="DeepSeek-AI" LicenseName="MIT" LicenseUrl="https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/mit.md" RepositoryUrl="https://huggingface.co/deepseek-ai/DeepSeek-V3.2/tree/main" UseCase="@T("We use the DeepSeek Tokenizer to estimate the number of tokens an input will generate.")" />
+                    <ThirdPartyComponent Name="Tokenizer" Developer="Anthony Moi, Nicolas Patry, Pierric Cistac, Arthur Zucker & Open Source Community" LicenseName="Apache-2.0" LicenseUrl="https://github.com/huggingface/tokenizers/blob/main/LICENSE" RepositoryUrl="https://github.com/huggingface/tokenizers" UseCase="@T("The Tokenizer library serves as the base framework for integrating the DeepSeek tokenizer.")" />
                     <ThirdPartyComponent Name="HtmlAgilityPack" Developer="ZZZ Projects & Open Source Community" LicenseName="MIT" LicenseUrl="https://github.com/zzzprojects/html-agility-pack/blob/master/LICENSE" RepositoryUrl="https://github.com/zzzprojects/html-agility-pack" UseCase="@T("We use the HtmlAgilityPack to extract content from the web. This is necessary, e.g., when you provide a URL as input for an assistant.")"/>
                     <ThirdPartyComponent Name="ReverseMarkdown" Developer="Babu Annamalai & Open Source Community" LicenseName="MIT" LicenseUrl="https://github.com/mysticmind/reversemarkdown-net/blob/master/LICENSE" RepositoryUrl="https://github.com/mysticmind/reversemarkdown-net" UseCase="@T("This library is used to convert HTML to Markdown. This is necessary, e.g., when you provide a URL as input for an assistant.")"/>
                     <ThirdPartyComponent Name="wikEd diff" Developer="Cacycle & Open Source Community" LicenseName="None (public domain)" LicenseUrl="https://en.wikipedia.org/wiki/User:Cacycle/diff#License" RepositoryUrl="https://en.wikipedia.org/wiki/User:Cacycle/diff" UseCase="@T("This library is used to display the differences between two texts. This is necessary, e.g., for the grammar and spelling assistant.")"/>

app/MindWork AI Studio/Plugins/languages/de-de-43065dbc-78d0-45b7-92be-f14c2926e2dc/plugin.lua

@@ -1620,6 +1620,9 @@ UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3403290862"] = "Der ausge
 -- Select a provider first
 UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3654197869"] = "Wähle zuerst einen Anbieter aus"
 
+-- Estimated amount of tokens:
+UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T377990776"] = "Geschätzte Anzahl an Tokens:"
+
 -- Start new chat in workspace "{0}"
 UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3928697643"] = "Neuen Chat im Arbeitsbereich \"{0}\" starten"
 
@@ -5097,6 +5100,9 @@ UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1019424746"] = "Startprotokollda
 -- Browse AI Studio's source code on GitHub — we welcome your contributions.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1107156991"] = "Sehen Sie sich den Quellcode von AI Studio auf GitHub an – wir freuen uns über ihre Beiträge."
 
+-- The Tokenizer library serves as the base framework for integrating the DeepSeek tokenizer.
+UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1132433749"] = "Die Tokenizer‑Bibliothek dient als Basis‑Framework für die Integration des DeepSeek‑Tokenizers."
+
 -- ID mismatch: the plugin ID differs from the enterprise configuration ID.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1137744461"] = "ID-Konflikt: Die Plugin-ID stimmt nicht mit der ID der Unternehmenskonfiguration überein."
 
@@ -5331,6 +5337,9 @@ UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T566998575"] = "Dies ist eine Bib
 -- Used .NET SDK
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T585329785"] = "Verwendetes .NET SDK"
 
+-- We use the DeepSeek Tokenizer to estimate the number of tokens an input will generate.
+UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T591393704"] = "Wir verwenden den DeepSeek‑Tokenizer, um die Token‑Anzahl einer Eingabe zu schätzen."
+
 -- This library is used to manage sidecar processes and to ensure that stale or zombie sidecars are detected and terminated.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T633932150"] = "Diese Bibliothek wird verwendet, um Sidecar-Prozesse zu verwalten und sicherzustellen, dass veraltete oder Zombie-Sidecars erkannt und beendet werden."
 

app/MindWork AI Studio/Plugins/languages/en-us-97dfb1ba-50c4-4440-8dfa-6575daf543c8/plugin.lua

@@ -1620,6 +1620,9 @@ UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3403290862"] = "The selec
 -- Select a provider first
 UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3654197869"] = "Select a provider first"
 
+-- Estimated amount of tokens:
+UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T377990776"] = "Estimated amount of tokens:"
+
 -- Start new chat in workspace "{0}"
 UI_TEXT_CONTENT["AISTUDIO::COMPONENTS::CHATCOMPONENT::T3928697643"] = "Start new chat in workspace \"{0}\""
 
@@ -5097,6 +5100,9 @@ UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1019424746"] = "Startup log file
 -- Browse AI Studio's source code on GitHub — we welcome your contributions.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1107156991"] = "Browse AI Studio's source code on GitHub — we welcome your contributions."
 
+-- The Tokenizer library serves as the base framework for integrating the DeepSeek tokenizer.
+UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1132433749"] = "The Tokenizer library serves as the base framework for integrating the DeepSeek tokenizer."
+
 -- ID mismatch: the plugin ID differs from the enterprise configuration ID.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T1137744461"] = "ID mismatch: the plugin ID differs from the enterprise configuration ID."
 
@@ -5331,6 +5337,9 @@ UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T566998575"] = "This is a library
 -- Used .NET SDK
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T585329785"] = "Used .NET SDK"
 
+-- We use the DeepSeek Tokenizer to estimate the number of tokens an input will generate.
+UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T591393704"] = "We use the DeepSeek Tokenizer to estimate the number of tokens an input will generate."
+
 -- This library is used to manage sidecar processes and to ensure that stale or zombie sidecars are detected and terminated.
 UI_TEXT_CONTENT["AISTUDIO::PAGES::INFORMATION::T633932150"] = "This library is used to manage sidecar processes and to ensure that stale or zombie sidecars are detected and terminated."
 

app/MindWork AI Studio/Tools/Rust/TokenCountInfo.cs

@@ -0,0 +1,6 @@
+namespace AIStudio.Tools.Rust;
+
+public sealed class TokenCountInfo
+{
+    public int TokenCount { get; set; }
+}

app/MindWork AI Studio/Tools/Services/RustService.Tokenizer.cs

@@ -0,0 +1,27 @@
+using AIStudio.Tools.Rust;
+
+namespace AIStudio.Tools.Services;
+
+public sealed partial class RustService
+{
+    public async Task<TokenCountInfo?> GetTokenCount(string text)
+    {
+        try
+        {
+            var cts = new CancellationTokenSource(TimeSpan.FromSeconds(5));
+            var payload = new { text };
+            var response = await this.http.PostAsJsonAsync("/system/tokenizer/count", payload, this.jsonRustSerializerOptions, cts.Token);
+            response.EnsureSuccessStatusCode();
+            return await response.Content.ReadFromJsonAsync<TokenCountInfo>(this.jsonRustSerializerOptions, cancellationToken: cts.Token);
+        }
+        catch (Exception e)
+        {
+            if(this.logger is not null)
+                this.logger.LogError(e, "Error while getting token count from Rust service.");
+            else
+                Console.WriteLine($"Error while getting token count from Rust service: '{e}'.");
+
+            return null;
+        }
+    }
+}

runtime/Cargo.toml

@@ -42,6 +42,7 @@ pptx-to-md = "0.4.0"
 tempfile = "3.8"
 strum_macros = "0.27"
 sysinfo = "0.38.0"
+tokenizers = "0.22.2"
 
 # Fixes security vulnerability downstream, where the upstream is not fixed yet:
 time = "0.3.47" # -> Rocket

runtime/src/lib.rs

@@ -17,4 +17,5 @@ pub mod qdrant;
 pub mod certificate_factory;
 pub mod runtime_api_token;
 pub mod stale_process_cleanup;
-mod sidecar_types;
+mod sidecar_types;
+pub mod tokenizer;

runtime/src/main.rs

@@ -11,7 +11,7 @@ use mindwork_ai_studio::environment::is_dev;
 use mindwork_ai_studio::log::init_logging;
 use mindwork_ai_studio::metadata::MetaData;
 use mindwork_ai_studio::runtime_api::start_runtime_api;
-
+use mindwork_ai_studio::tokenizer::{init_tokenizer};
 
 #[tokio::main]
 async fn main() {
@@ -43,8 +43,12 @@ async fn main() {
         info!("Running in production mode.");
     }
 
+    if let Err(e) = init_tokenizer() {
+        warn!(Source = "Tokenizer"; "Error during the initialisation of the tokenizer: {}", e);
+    }
+
     generate_runtime_certificate();
     start_runtime_api();
 
     start_tauri();
-}
+}

runtime/src/runtime_api.rs

@@ -89,6 +89,7 @@ pub fn start_runtime_api() {
                 crate::file_data::extract_data,
                 crate::log::get_log_paths,
                 crate::log::log_event,
+                crate::tokenizer::tokenizer_count,
                 crate::app_window::register_shortcut,
                 crate::app_window::validate_shortcut,
                 crate::app_window::suspend_shortcuts,