@pheem49/mint 1.8.2 → 1.9.0

package/Cargo.toml

@@ -7,7 +7,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "1.8.2"
+version = "1.9.0"
 edition = "2024"
 license = "AGPL-3.0-only"
 

package/Release_Note.md

@@ -1,10 +1,32 @@
-# Release Notes - Mint Agent v1.8.2
+# Release Notes - Mint Agent v1.9.0
 
-We are excited to release **Mint Agent v1.8.2**! This release introduces a configurable Multi-Agent system for sequential collaboration, powerful local workspace file management features, real-time AI-generated workflow suggestions in the user's active language, mobile layout optimization, and general UI/UX polish.
+We are excited to release **Mint Agent v1.9.0**! This release introduces a configurable Multi-Agent system for sequential collaboration, powerful local workspace file management features, real-time AI-generated workflow suggestions in the user's active language, mobile layout optimization, and general UI/UX polish.
+
+---
+
+## 🔌 Custom Provider Support
+
+Users can now add any **OpenAI-compatible endpoint** as a fully configurable custom AI provider, giving complete control over which LLM backend Mint uses.
+
+### Backend (`config.rs`, `chat.rs`)
+- **New data model:** Added `CustomProvider`, `CustomProviderModel`, and `CustomProviderHeader` structs to `MintConfig`. Custom providers are persisted to `mint-config.json` under the `customProviders` key.
+- **Provider ID format:** Each custom provider is stored with a user-defined slug ID and is referenced in `ai_provider` as `"custom:<id>"` (e.g. `"custom:myprovider"`).
+- **Chat routing:** `send_chat()` and `stream_chat()` now dispatch `custom:*` providers through `call_custom_provider()` / `stream_custom_provider()`, which use the existing OpenAI-compatible HTTP code path.
+- **Multiple models:** Each provider stores a list of models; the active model is resolved from `customModelSelections` extra config (set by the UI model selector dropdown).
+- **Custom HTTP headers:** All configured headers are applied to every request sent to that provider.
+- **`available_providers()`** now returns `Vec<String>` (up from `Vec<&'static str>`) to accommodate dynamic custom provider IDs.
+- **`resolve_custom_provider()`** helper added to look up a `CustomProvider` by key (strips the `custom:` prefix automatically).
+
+### Frontend (`SettingsWindow.tsx`, `GeneralTab.tsx`, `AgentsTab.tsx`)
+- **Settings → General → Active Provider dropdown** now includes all configured custom providers dynamically.
+- **Model selector:** When a custom provider is selected as the active provider, a second dropdown appears to pick among that provider's configured models.
+- **Custom Providers section:** A new section at the bottom of General settings displays each custom provider as a card. Each card allows inline editing of: Provider ID (slug-validated), Display Name, Base URL, API Key (password), per-model rows (model ID + display name), and per-header rows (name + value). Providers can be added with `+ Add custom provider` and deleted inline.
+- **Agents tab:** Custom providers (with a configured base URL) now appear in the AI Provider dropdown when creating or editing an agent. The model selector shows that provider's configured models.
 
 ---
 
 ## 🔍 Web Search Source Cards
+
 - **Favicon Source Cards:** After a web search, compact clickable source cards now appear above the AI response bubble — each showing the website's favicon, domain name, and a tooltip with the page snippet. Clicking a card opens the source in the browser.
 - **Frontend-only change:** Parsed directly from existing `AgentProgress` `ToolEnd` events — no backend changes required.
 - The AI response also includes a plain-text `Sources:` section listing title and URL for each result used.
@@ -37,9 +59,13 @@ Improved the autocomplete experience in the terminal:
 - **Simplified Exit Options:** Removed `/quit` from the autocomplete suggestions and help output (while retaining hidden execution support) to clean up the interface.
 - **Real-Time Agent & Model Status:** The CLI live thinking status bar dynamically displays which agent and model is executing.
 
-### 📋 5. Fix Clipboard Permission Popup on Text Paste (Ctrl+V)
+### 📋 5. Fix Clipboard Permission Popup on Text Paste (Ctrl+V) & Desktop Image Paste Fallback
 - Fixed a bug where pasting text (or pressing `Ctrl + V`) would trigger a browser-level clipboard permission warning popup (asking to "Allow Paste").
 - Removed the redundant global `keydown` Ctrl+V listener and the async `navigator.clipboard.read()` fallback on text paste events. The application now correctly relies on standard, synchronous `event.clipboardData` values during paste events, ensuring seamless pasting of both images and text without prompting the user.
+- **Tauri Native Fallback:** Restored an asynchronous fallback using Tauri's native backend `readClipboardImage` command specifically for the desktop application. This solves the Linux/Windows WebView limitations where WebKitGTK and standard web APIs fail to expose file handles in standard paste events.
+- **Double Paste Throttling:** Implemented a 100ms throttle wrapper around the native clipboard read callback, preventing duplicate image attachments caused by concurrent window-level capture and element-level paste event propagation.
+
+
 
 ### 📦 6. Codebase Refactoring & Shared Utilities
 - **Shared UI & Progress Helpers:** Refactored duplicate helper functions and TypeScript interfaces from the desktop renderer (`src/renderer/src`) and web renderer (`src/renderer/src-web`) into the unified `shared` codebase.
@@ -68,8 +94,16 @@ Improved the autocomplete experience in the terminal:
 ### 🤖 11. AI Model Self-Awareness Context Injection
 - **Dynamic Context Injection**: Updated the core prompt compilation layer to automatically append an `[Active Environment Context]` block to the system instructions. This supplies the active AI model name and provider to the AI model itself, allowing the AI to successfully know its own model configurations in real-time when queried by the user.
 
+### 🎨 12. Material Icon Theme in Agent Activity Tracker
+- **Visual File & Folder Tracking**: The agent activity tracker (e.g. "Exploring file" progress log) now resolves target filenames/foldernames to display specific Material Icon Theme SVGs (e.g. JavaScript icon for `.js`, Rust icon for `.rs` / Cargo directories). This replaces the generic wireframe outline icons, resulting in a cleaner and more polished UI.
+
+### 🎙️ 13. Stricter Voice Reply Settings Control
+- **Strict Control Switch**: The global "Enable Voice Reply" toggle now acts as a strict master switch. When disabled, Mint will never speak responses out loud under any circumstances (including when using voice conversation/microphone mode).
+- **Voice Mode Loop Resumption**: When "Enable Voice Reply" is disabled but voice conversation mode is active, the microphone automatically resumes listening after the AI response is printed on screen, maintaining the continuous hands-free voice loop.
+
 ---
 
+
 ## 🛠️ Codebase Changes
 
 ### Tauri Backend (`src-tauri` & `crates/mint-core`)
@@ -111,6 +145,10 @@ Improved the autocomplete experience in the terminal:
 - Refactor duplicated types (`DiffHunk`, `FileChange`), helper functions (`numericSetting`, `errorMessage`, `readImage`, `readDocument`, `createTrimmedImagePreview`, `lightenColor`, `hexToRgb`, `applyThemeStyles`), and progress parsers (`parseFileChangesFromProgress`) from the desktop and web components to a central `src/renderer/shared` repository. Corrected a type checking issue by exporting `AgentProgress` and `InteractionMemory` from `shared/agentProgress.ts`.
 - Consolidate microphone interface to client-side Speech-to-Text (`SpeechRecognition`), extract STT states, silence-detection timers, and event listeners into a reusable React hook `useSpeechToText` in [speech.ts](file:///home/pheem49/vscode/Project/Mint-CLI/src/renderer/shared/utils/speech.ts), and clean up both desktop (`src/renderer/src/components/ChatPanel.tsx`) and web (`src/renderer/src-web/components/ChatPanel.tsx`) `ChatPanel` components.
 - Implement 24-picture pagination, a manual "Load More" button, and API/image fetch cache-busters in `PicturesLibrary.tsx` and `tauri.ts` across desktop and web directories.
+- Refactor file and folder icon mappings from `WorkspacePanel.tsx` into a central, reusable utility `src/renderer/shared/utils/fileIcons.ts`.
+- Update `AgentActivityTable.tsx` to resolve target filenames/foldernames and display their respective Material Icon Theme SVGs instead of default wireframe outlines.
+- Update Desktop and Web CSS stylesheets (`styles.css`) to support rendering and scaling Material Icon SVGs inside the agent activity tracker.
+- Update `ChatPanel.tsx` (Desktop and Web) to strictly check the global `enableVoiceReply` setting. If voice reply is disabled but voice conversation mode is active, the microphone listening loop is automatically scheduled to resume without reading the AI response out loud.
 
 ### CLI Agent (`crates/mint-cli`)
 - Redefine live status print lines (`plan_lines`, `tasks_lines`, `activities_lines`, `explored_lines`) to accept progress tick state and apply the `get_bullet` helper.
@@ -123,6 +161,13 @@ Improved the autocomplete experience in the terminal:
 - Fixed `AgentProgress::Thinking` pattern matching compiler error (E0027) in `crates/mint-cli/src/agent.rs` and render active agent and model names in the live thinking terminal status bar.
 - Implement horizontal scrolling viewport for the CLI prompt input box in `crates/mint-cli/src/main.rs`, slicing the visible text to fit the terminal screen and adjusting the cursor column offset accordingly, allowing arbitrary length instructions.
 - Save `provider_change` system interaction events to the database and render a beautiful styled horizontal divider in the console when switching models via `/models` inside `crates/mint-cli/src/main.rs`.
+- Implement dynamic skill invocation with `$` prefix in the CLI interactive chat.
+- Add realtime autocomplete and pagination dropdown list for local/workspace and global skills under the `$` trigger, resolving and rendering skill descriptions.
+- Add frontmatter YAML parser in `crates/mint-core/src/skills.rs` to fetch skill description values.
+- Support confirmation dialogue to display skill contents and confirm activation before running the Agent, allowing inline or prompted tasks.
+- Register `/skill add <path>` (and `/skill install <path>`) slash commands to automatically copy/install a skill file or folder into the global `~/.config/mint/mint-skills/` directory.
+
+
 
 
 ### Github CI/CD Workflows

package/crates/mint-cli/src/agent.rs

@@ -189,39 +189,47 @@ pub async fn run_code_agent_with_options(
             agent_name,
             model_name,
         } => {
-            if !options.fast_mode {
-                if let Ok(mut status) = progress_live_status.lock() {
-                    let label = if let (Some(a), Some(m)) = (agent_name, model_name) {
-                        format!("{} ({}) is thinking ({} • Esc to interrupt)", a, m, format_elapsed(Duration::from_secs(elapsed_secs)))
-                    } else {
-                        format!("Thinking ({} • Esc to interrupt)", format_elapsed(Duration::from_secs(elapsed_secs)))
-                    };
-                    status.thinking = Some(label);
-                    render_live_status(&mut status);
-                }
+            if !options.fast_mode
+                && let Ok(mut status) = progress_live_status.lock()
+            {
+                let label = if let (Some(a), Some(m)) = (agent_name, model_name) {
+                    format!(
+                        "{} ({}) is thinking ({} • Esc to interrupt)",
+                        a,
+                        m,
+                        format_elapsed(Duration::from_secs(elapsed_secs))
+                    )
+                } else {
+                    format!(
+                        "Thinking ({} • Esc to interrupt)",
+                        format_elapsed(Duration::from_secs(elapsed_secs))
+                    )
+                };
+                status.thinking = Some(label);
+                render_live_status(&mut status);
             }
         }
         AgentProgress::Thought { thought } => {
-            if !options.fast_mode {
-                if let Ok(mut status) = progress_live_status.lock() {
-                    commit_activity_snapshot(&mut status);
-                    print_timeline_note(&thought);
-                    status.thinking = None;
-                    render_live_status(&mut status);
-                }
+            if !options.fast_mode
+                && let Ok(mut status) = progress_live_status.lock()
+            {
+                commit_activity_snapshot(&mut status);
+                print_timeline_note(&thought);
+                status.thinking = None;
+                render_live_status(&mut status);
             }
         }
         AgentProgress::ToolStart { action, input } => {
             if !options.fast_mode {
-                if action == "create_plan" || action == "update_plan" {
-                    if let Some(steps) = extract_plan_steps(&input) {
-                        if let Ok(mut status) = progress_live_status.lock() {
-                            status.thinking = None;
-                            status.plan_steps = steps;
-                            render_live_status(&mut status);
-                        }
-                        return;
+                if (action == "create_plan" || action == "update_plan")
+                    && let Some(steps) = extract_plan_steps(&input)
+                {
+                    if let Ok(mut status) = progress_live_status.lock() {
+                        status.thinking = None;
+                        status.plan_steps = steps;
+                        render_live_status(&mut status);
                     }
+                    return;
                 }
                 if let Some(label) = explored_action_label(&action, &input) {
                     if let Ok(mut status) = progress_live_status.lock() {
@@ -313,12 +321,12 @@ pub async fn run_code_agent_with_options(
         } => {
             if !options.fast_mode {
                 if action == "create_plan" || action == "update_plan" {
-                    if let Some(steps) = extract_plan_steps(&input) {
-                        if let Ok(mut status) = progress_live_status.lock() {
-                            status.thinking = None;
-                            status.plan_steps = steps;
-                            render_live_status(&mut status);
-                        }
+                    if let Some(steps) = extract_plan_steps(&input)
+                        && let Ok(mut status) = progress_live_status.lock()
+                    {
+                        status.thinking = None;
+                        status.plan_steps = steps;
+                        render_live_status(&mut status);
                     }
                 } else if command_was_run(&result)
                     && let Some(commands) = ran_command_labels(&action, &input)
@@ -335,39 +343,38 @@ pub async fn run_code_agent_with_options(
             if action == "web_search"
                 && !result.starts_with("Web search error:")
                 && result != "No web search results found."
+                && let Ok(mut status) = progress_live_status.lock()
             {
-                if let Ok(mut status) = progress_live_status.lock() {
-                    let sources = parse_web_search_sources(&result);
-                    status.web_sources.extend(sources);
-                }
+                let sources = parse_web_search_sources(&result);
+                status.web_sources.extend(sources);
             }
         }
     };
 
     let chunk_live_status = Arc::clone(&live_status);
     let on_chunk = |summary: String| {
-        if !options.fast_mode {
-            if let Ok(mut status) = chunk_live_status.lock() {
-                status.thinking = None;
-                commit_activity_snapshot(&mut status);
-                clear_live_status(&mut status);
-            }
+        if !options.fast_mode
+            && let Ok(mut status) = chunk_live_status.lock()
+        {
+            status.thinking = None;
+            commit_activity_snapshot(&mut status);
+            clear_live_status(&mut status);
         }
         let formatted_summary = format_markdown_bold(&sanitize_latex(&summary));
         print!("\n  {MINT}Mint:{RESET} ");
         render_live_summary(&formatted_summary);
 
         // Print web search sources if any were collected
-        if let Ok(mut status) = chunk_live_status.lock() {
-            if !status.web_sources.is_empty() {
-                println!();
-                println!("  {DIM}Sources:{RESET}");
-                for (i, (title, url)) in status.web_sources.iter().enumerate() {
-                    println!("  {DIM}{}.{RESET} {BLUE}{}{RESET}", i + 1, title);
-                    println!("     {DIM}{}{RESET}", url);
-                }
-                status.web_sources.clear();
+        if let Ok(mut status) = chunk_live_status.lock()
+            && !status.web_sources.is_empty()
+        {
+            println!();
+            println!("  {DIM}Sources:{RESET}");
+            for (i, (title, url)) in status.web_sources.iter().enumerate() {
+                println!("  {DIM}{}.{RESET} {BLUE}{}{RESET}", i + 1, title);
+                println!("     {DIM}{}{RESET}", url);
             }
+            status.web_sources.clear();
         }
 
         println!();
@@ -396,14 +403,14 @@ pub async fn run_code_agent_with_options(
         }
     };
     agent_done.store(true, Ordering::Relaxed);
-    if !options.fast_mode {
-        if let Ok(mut status) = live_status.lock() {
-            status.thinking = None;
-            if res.is_err() {
-                commit_activity_snapshot(&mut status);
-            }
-            clear_live_status(&mut status);
+    if !options.fast_mode
+        && let Ok(mut status) = live_status.lock()
+    {
+        status.thinking = None;
+        if res.is_err() {
+            commit_activity_snapshot(&mut status);
         }
+        clear_live_status(&mut status);
     }
     let res = res.map_err(|e| anyhow!("{}", e))?;
 
@@ -436,19 +443,19 @@ fn parse_web_search_sources(result: &str) -> Vec<(String, String)> {
     for line in result.lines() {
         let trimmed = line.trim();
         // Match numbered title lines: "1. Title text"
-        if let Some(rest) = trimmed.split_once(". ") {
-            if rest.0.parse::<usize>().is_ok() {
-                current_title = Some(rest.1.trim().to_owned());
-                continue;
-            }
+        if let Some(rest) = trimmed.split_once(". ")
+            && rest.0.parse::<usize>().is_ok()
+        {
+            current_title = Some(rest.1.trim().to_owned());
+            continue;
         }
         // Match URL lines: "URL: https://..."
-        if let Some(url) = trimmed.strip_prefix("URL: ") {
-            if let Some(title) = current_title.take() {
-                let url = url.trim().to_owned();
-                if !url.is_empty() {
-                    sources.push((title, url));
-                }
+        if let Some(url) = trimmed.strip_prefix("URL: ")
+            && let Some(title) = current_title.take()
+        {
+            let url = url.trim().to_owned();
+            if !url.is_empty() {
+                sources.push((title, url));
             }
         }
     }
@@ -680,7 +687,7 @@ fn render_live_status(status: &mut LiveStatus) {
         let stripped = strip_ansi_escapes(line);
         let line_len = stripped.chars().filter(|&c| !is_thai_combining(c)).count();
         let physical_lines = if width > 0 {
-            (line_len + width - 1) / width
+            line_len.div_ceil(width)
         } else {
             1
         }
@@ -753,7 +760,11 @@ fn clear_live_status(status: &mut LiveStatus) {
 
 fn get_bullet(name: &str, is_thinking: bool, tick: usize) -> String {
     let char_str = if is_thinking {
-        if (tick / 4) % 2 == 0 { "●" } else { "○" }
+        if (tick / 4).is_multiple_of(2) {
+            "●"
+        } else {
+            "○"
+        }
     } else {
         "●"
     };
@@ -966,7 +977,7 @@ fn format_markdown_bold(text: &str) -> String {
                 } else {
                     if c == '-' || c == '*' || c == '+' {
                         let after = &line[idx + c.len_utf8()..];
-                        if after.chars().next().map_or(false, |c| c.is_whitespace()) {
+                        if after.chars().next().is_some_and(|c| c.is_whitespace()) {
                             is_list_item = true;
                             marker_char = c;
                         }
@@ -1043,27 +1054,27 @@ fn sanitize_latex(text: &str) -> String {
     let mut s = text.to_owned();
     for (pat, uni) in [
         // arrows
-        ("$\\rightarrow$",     "→"),
-        ("$\\leftarrow$",      "←"),
-        ("$\\Rightarrow$",     "⇒"),
-        ("$\\Leftarrow$",      "⇐"),
+        ("$\\rightarrow$", "→"),
+        ("$\\leftarrow$", "←"),
+        ("$\\Rightarrow$", "⇒"),
+        ("$\\Leftarrow$", "⇐"),
         ("$\\leftrightarrow$", "↔"),
         // comparison
-        ("$\\leq$",    "≤"),
-        ("$\\geq$",    "≥"),
-        ("$\\neq$",    "≠"),
+        ("$\\leq$", "≤"),
+        ("$\\geq$", "≥"),
+        ("$\\neq$", "≠"),
         ("$\\approx$", "≈"),
         // math
         ("$\\times$", "×"),
-        ("$\\div$",   "÷"),
-        ("$\\pm$",    "±"),
+        ("$\\div$", "÷"),
+        ("$\\pm$", "±"),
         ("$\\infty$", "∞"),
-        ("$\\cdot$",  "·"),
+        ("$\\cdot$", "·"),
         // sets
-        ("$\\in$",     "∈"),
+        ("$\\in$", "∈"),
         ("$\\subset$", "⊂"),
-        ("$\\cup$",    "∪"),
-        ("$\\cap$",    "∩"),
+        ("$\\cup$", "∪"),
+        ("$\\cap$", "∩"),
     ] {
         s = s.replace(pat, uni);
     }

package/crates/mint-cli/src/image.rs

@@ -27,21 +27,21 @@ pub fn read_clipboard_image() -> Result<Option<String>> {
         if let Ok(output) = std::process::Command::new("xclip")
             .args(["-selection", "clipboard", "-t", "image/png", "-o"])
             .output()
+            && output.status.success()
+            && !output.stdout.is_empty()
         {
-            if output.status.success() && !output.stdout.is_empty() {
-                let encoded = BASE64.encode(&output.stdout);
-                return Ok(Some(format!("data:image/png;base64,{encoded}")));
-            }
+            let encoded = BASE64.encode(&output.stdout);
+            return Ok(Some(format!("data:image/png;base64,{encoded}")));
         }
 
         if let Ok(output) = std::process::Command::new("wl-paste")
             .args(["--type", "image/png"])
             .output()
+            && output.status.success()
+            && !output.stdout.is_empty()
         {
-            if output.status.success() && !output.stdout.is_empty() {
-                let encoded = BASE64.encode(&output.stdout);
-                return Ok(Some(format!("data:image/png;base64,{encoded}")));
-            }
+            let encoded = BASE64.encode(&output.stdout);
+            return Ok(Some(format!("data:image/png;base64,{encoded}")));
         }
 
         Ok(None)