glebkudr · antonshalin76 · Jan 21, 2026
diff --git a/.gitignore b/.gitignore
@@ -62,6 +62,8 @@ logs/
 *.bak
 *.tmp
 Show-DirectoryTree.ps1
+shotgun.desktop
+run-shotgun.sh
 
 # If you have specific large files or data that shouldn't be in git
 # data/

diff --git a/README.md b/README.md
@@ -2,82 +2,77 @@
 
 ![Shotgun App Banner](https://github.com/user-attachments/assets/6dd15389-4ad9-493a-a0e7-9813eb143e38)
 
-**Tired of Cursor cutting off context, missing your files, and spitting out empty responses?**
-
-**Shotgun** is the bridge between your local codebase and the world's most powerful LLMs.
-It doesn't just copy files; it **intelligently packages your project context** and can **execute prompts directly** against OpenAI (GPT-4o/GPT-5), Google Gemini, or OpenRouter.
-
-> **Stop copy-pasting 50 files manually.**
-> 1. Select your repo.
-> 2. Let AI pick the relevant files (Auto-Context).
-> 3. Blast the payload directly to the model or copy it for use in Cursor/Windsurf.
+Shotgun is a desktop app for preparing project context and working with LLMs.  
+It collects selected files into a structured context, can auto‑select relevant files, and can send prompts directly to OpenAI, Gemini, or OpenRouter.
 
 ---
 
-## 1. What Shotgun Does
-Shotgun is a desktop power-tool that **explodes your project into a structured payload** designed for AI reasoning.
-
-It has evolved from a simple "context dumper" into a full-fledged **LLM Client for Codebases**:
-*   **Smart Selection:** Uses AI ("Auto-Context") to analyze your task and automatically select only the relevant files from your tree.
-*   **Direct Execution:** Configurable API integration with **OpenAI**, **Gemini**, and **OpenRouter**.
-*   **Prompt Engineering:** Built-in templates for different roles (Developer, Architect, Bug Hunter).
-*   **History & Audit:** Keeps a full log of every prompt sent and response received.
+## What Shotgun Does
+Shotgun turns your repository into an LLM‑friendly context:
+- builds a file tree and collects file contents;
+- respects `.gitignore` and custom rules;
+- shows size and token counts per file/folder;
+- outputs XML‑like blocks `<file path="...">...</file>`.
 
 ---
 
-## 2. Key Features
+## Key Features
 
-### 🧠 AI-Powered Context
-*   **Auto-Context:** Don't know which files are needed for a bug fix? Type your task, and Shotgun uses an LLM to scan your tree and select the relevant files for you.
-*   **Repo Scan:** supplement context retrieval with a `shotgun_reposcan.md` summary of your architecture to give the LLM high-level awareness before diving into code.
+### 🧠 Auto Context
+- **Auto‑Context**: the LLM selects relevant files based on your task.
+- **Repo Scan**: optional `shotgun_reposcan.md` for high‑level architecture context.
 
-### ⚡ Workflow Speed
-*   **Fast Tree Scan:** Go + Wails backend scans thousands of files in milliseconds.
-*   **Interactive Tree:** Manually toggle files/folders or use `.gitignore` and custom rule sets to filter noise.
-*   **One-Click Blast:** Generate a massive context payload instantly.
+### ⚡ Large‑project performance
+- **Multiple project roots**: add several root folders in one workspace.
+- **Project tree** with manual selection.
+- **Symlink traversal** with cycle protection.
+- **Offline token counting** (`o200k_base`) without API calls.
+- **Context size limit** via `settings.json` (`maxContextBytes`, default 50MB).
+- **Auto context generation** with debounce + queue (no constant restarts).
 
-### 🔌 Direct Integrations
-*   **OpenAI:** Support for GPT-4o and experimental support for **GPT-5** family models.
-*   **Google Gemini:** Native integration for Gemini 2.5/3 Pro & Flash.
-*   **OpenRouter:** Access hundreds of LLM's via a unified API.
+### 🔌 Integrations
+Supported providers:
+- OpenAI
+- Google Gemini
+- OpenRouter
 
-### 🛠 Developer Experience
-*   **Prompt Templates:** Switch modes easily (e.g., "Find Bug" vs "Refactor" vs "Write Docs").
-*   **History Tracking:** Never lose a generated patch. Browse past prompts, responses, and raw API payloads.
-*   **Privacy Focused:** Your code goes only to the API provider you choose. No intermediate servers.
+You can pick a model and run requests directly from the app.
 
 ---
 
-## 3. The Workflow
-
-Shotgun guides you through a 3-step process:
+## Workflow
 
 ### Step 1: Prepare Context
-*   **Select Project:** Open your local repository.
-*   **Filter:** Use the checkbox tree, `.gitignore`, or the **Auto-Context** button to define the scope.
-*   **Repo Scan:** Edit or load the high-level repository summary for better AI grounding.
-*   **Result:** A structured XML-like dump of your selected codebase.
+1. Add one or more project roots.
+2. Select files manually or use Auto‑Context.
+3. Context is generated automatically with progress.
+4. **Generated Project Context** shows an exact token count.
 
 ### Step 2: Compose & Execute
-*   **Define Task:** Describe what you need (e.g., "Refactor the auth middleware to use JWT").
-*   **Select Template:** Choose a persona (Dev, Architect, QA).
-*   **Execute:** Click **"Execute Prompt"** to send it to the configured LLM API immediately, OR copy the full payload to your clipboard for use in external tools like ChatGPT or Cursor.
+1. Describe your task.
+2. Choose a prompt template.
+3. Execute the request or copy the prompt.
 
-### Step 3: History & Apply
-*   **Review:** View the AI's response alongside your original prompt.
-*   **Diffs:** The AI output is optimized for `diff` generation.
-*   **Audit:** Inspect raw API calls for debugging or token usage analysis.
+### Step 3: History
+History of requests, responses, and call parameters.
 
 ---
 
-## 4. Installation
+## UI
+- **Resizable sidebar** (drag to resize).
+- **Resizable bottom console**.
 
-### Prerequisites
-*   **Go ≥ 1.20**
-*   **Node.js LTS**
-*   **Wails CLI:** `go install github.com/wailsapp/wails/v2/cmd/wails@latest`
+---
 
-### Clone & Build
+## Installation & Run
+
+### Requirements
+- **Go ≥ 1.20**
+- **Node.js LTS**
+- **Wails CLI**  
+  `go install github.com/wailsapp/wails/v2/cmd/wails@latest`
+
+### Build & Run
 ```bash
 git clone https://github.com/glebkudr/shotgun_code
 cd shotgun_code
@@ -87,32 +82,42 @@ cd frontend
 npm install
 cd ..
 
-# Run in Development Mode (Hot Reload)
+# Dev mode (hot reload)
 wails dev
 
-# Build Production Binary
+# Production build
 wails build
 ```
-*Binaries will be located in `build/bin/`.*
+Binaries are placed in `build/bin/`.
 
 ---
 
-## 5. Configuration
+## Configuration
 
-### LLM Setup
-Click the **Settings** (gear icon) in the app to configure providers:
-1.  **Provider:** Select OpenAI, Gemini, or OpenRouter.
-2.  **API Key:** Paste your key (stored locally).
-3.  **Model:** Select your preferred model (e.g., `gpt-4o`, `gemini-2.5-pro`, `claude-3.5-sonnet`).
+### Settings file
+Uses the XDG path `shotgun-code/settings.json`.  
+On Linux this is typically:
+```
+~/.config/shotgun-code/settings.json
+```
 
-### Custom Rules
-You can define global excludes (like `node_modules`, `dist`, `.git`) and custom prompt instructions that are appended to every request.
+Example:
+```json
+{
+  "maxContextBytes": 50000000,
+  "customIgnoreRules": "...",
+  "customPromptRules": "..."
+}
+```
 
----
+### Key options
+- **`maxContextBytes`** — context size limit (default 50MB).
+- **`customIgnoreRules`** — file exclusion rules.
+- **`customPromptRules`** — global prompt rules.
 
-## 6. Output Format
+---
 
-Shotgun generates context optimized for LLM parsing:
+## Output Format
 
 ```xml
 <file path="backend/main.go">
@@ -127,30 +132,22 @@ package main
 </file>
 ```
 
-This format allows models to understand file boundaries perfectly, enabling accurate multi-file refactoring suggestions.
-
 ---
 
-## 7. ⚖️ License & Usage
-
-My name is Gleb Curly, and I am an indie developer making software for a living.
-
-Shotgun is developed and maintained by **Curly's Technology Tmi**. 
-
-This project uses a **Community License** model:
-
-### 1. Free for Small Teams & Non-Commercial Use
-You can use Shotgun for free (including modification and internal use) if:
-- Your company/team generates **less than $1M USD** in annual revenue.
-- You do **not** use the code to build a competing public product.
+## License
 
-### 2. Commercial License (Enterprise)
-If your annual revenue exceeds **$1M USD**, you are required to purchase a commercial license with a pretty reasonable price.
+Shotgun is developed and maintained by **Curly's Technology Tmi**.
 
-Please contact me at **glebkudr@gmail.com** for pricing.
+### 1. Free for small teams and non‑commercial use
+You may use Shotgun for free if:
+- your annual revenue is **below $1M**;
+- you do **not** use the code to build a competing product.
 
-See [LICENSE.md](LICENSE.md) for the full legal text.
+### 2. Commercial license
+If your revenue exceeds $1M, a commercial license is required.  
+Contact: **glebkudr@gmail.com**  
+See `LICENSE.md` for details.
 
 ---
 
-*Shotgun – Load, Aim, Blast your code into the future.*
+*Shotgun — Load, Aim, Blast your code into the future.*
diff --git a/README_ru.md b/README_ru.md
@@ -0,0 +1,153 @@
+# Shotgun App
+
+![Shotgun App Banner](https://github.com/user-attachments/assets/6dd15389-4ad9-493a-a0e7-9813eb143e38)
+
+Shotgun — десктопное приложение для подготовки контекста проекта и работы с LLM.  
+Оно собирает выбранные файлы в структурированный контекст, умеет автоматически выбирать релевантные файлы, а также отправлять готовый промпт прямо в OpenAI, Gemini или OpenRouter.
+
+---
+
+## Что делает Shotgun
+Shotgun превращает ваш репозиторий в удобный для LLM контекст:
+- собирает дерево файлов и их содержимое;
+- уважает `.gitignore` и кастомные правила;
+- показывает размер и токены по каждому файлу/папке;
+- формирует контекст в XML‑подобном формате `<file path="...">...</file>`.
+
+---
+
+## Основные возможности
+
+### 🧠 Автоматический контекст
+- **Auto‑Context**: LLM сам выбирает файлы по описанию задачи.
+- **Repo Scan**: подключаемый `shotgun_reposcan.md` с архитектурным обзором.
+
+### ⚡ Быстрая работа с большими проектами
+- **Несколько корневых папок**: можно добавить несколько корней проекта в один workspace.
+- **Дерево проектов** с ручным выбором файлов и папок.
+- **Симлинки** обходятся рекурсивно с защитой от циклов.
+- **Офлайн‑подсчёт токенов** (кодировка `o200k_base`) без обращения к API.
+- **Ограничение размера контекста** через `settings.json` (`maxContextBytes`, дефолт 50MB).
+- **Авто‑генерация контекста** с дебаунсом и очередью (без постоянных перезапусков).
+
+### 🔌 Интеграции
+Поддерживаются провайдеры:
+- OpenAI
+- Google Gemini
+- OpenRouter
+
+Можно выбрать модель и выполнять запрос прямо из приложения.
+
+---
+
+## Процесс работы
+
+### Шаг 1: Prepare Context
+1. Добавьте один или несколько корней проекта.
+2. Выберите файлы вручную или через Auto‑Context.
+3. Контекст генерируется автоматически с прогресс‑баром.
+4. В блоке **Generated Project Context** показывается точное число токенов.
+
+### Шаг 2: Compose & Execute
+1. Опишите задачу.
+2. Выберите шаблон промпта.
+3. Выполните запрос или скопируйте промпт.
+
+### Шаг 3: History
+История запросов, ответов и параметров вызова.
+
+---
+
+## Интерфейс
+- **Сайдбар** регулируется по ширине перетаскиванием.
+- **Нижняя консоль** регулируется по высоте.
+
+---
+
+## Установка и запуск
+
+### Требования
+- **Go ≥ 1.20**
+- **Node.js LTS**
+- **Wails CLI**  
+  `go install github.com/wailsapp/wails/v2/cmd/wails@latest`
+
+### Сборка и запуск
+```bash
+git clone https://github.com/glebkudr/shotgun_code
+cd shotgun_code
+
+# Установить зависимости фронтенда
+cd frontend
+npm install
+cd ..
+
+# Режим разработки (hot reload)
+wails dev
+
+# Продакшен‑сборка
+wails build
+```
+Бинарники появляются в `build/bin/`.
+
+---
+
+## Конфигурация
+
+### Файл настроек
+Используется XDG‑путь `shotgun-code/settings.json`.  
+На Linux это обычно:
+```
+~/.config/shotgun-code/settings.json
+```
+
+Пример:
+```json
+{
+  "maxContextBytes": 50000000,
+  "customIgnoreRules": "...",
+  "customPromptRules": "..."
+}
+```
+
+### Важные параметры
+- **`maxContextBytes`** — лимит размера контекста (по умолчанию 50MB).
+- **`customIgnoreRules`** — правила исключения файлов.
+- **`customPromptRules`** — глобальные правила для промптов.
+
+---
+
+## Формат вывода
+
+```xml
+<file path="backend/main.go">
+package main
+...
+</file>
+
+<file path="frontend/src/App.vue">
+<template>
+...
+</template>
+</file>
+```
+
+---
+
+## Лицензия
+
+Shotgun разрабатывается и поддерживается **Curly's Technology Tmi**.
+
+### 1. Бесплатно для небольших команд и некоммерческого использования
+Можно использовать бесплатно, если:
+- годовой оборот компании **меньше $1M**;
+- вы не используете код для создания конкурирующего продукта.
+
+### 2. Коммерческая лицензия
+Если оборот выше $1M — требуется коммерческая лицензия.  
+Контакт: **glebkudr@gmail.com**  
+Подробности в `LICENSE.md`.
+
+---
+
+*Shotgun — Load, Aim, Blast your code into the future.*