Académie OrvixCode

OrvixCode is a Windows desktop development environment that connects AI chat with real projects, files, editor workflows, and a built-in terminal.

You can open or create projects, analyze and edit code with AI, run multi-file project tasks, and choose between supported local or cloud models.

Core capabilities

• Planning and continuation for long-code and large-project tasks.
• Multi-file project generation with dependency-aware manifest validation.
• Syntax-highlighted code editor, Problems area, AI Fix, and inline AI editing.
• Built-in terminal with AI-assisted terminal error analysis.
• Automatic installation and retry for missing Python packages using the same interpreter.
• Explain, edit, and fix selected code.
• Customizable keyboard shortcuts, response stopping, and Markdown formatting while streaming.
• Local Ollama models, cloud providers, Combo Mode, and Visual Studio Code integration.

Download the current Windows installer from the official OrvixCode product page.

https://orvixcode.orvixgames.com/download.php?lang=fr-fr

1. Download the installerOpen the OrvixCode product page and download the current Windows build.
2. Run SetupIf Windows shows a security prompt, verify the source before continuing.
3. Choose the install locationKeep the default path or select another supported location.
4. Enable VS Code integrationIf Setup offers to install or update the extension, keep the option enabled if you use VS Code.
5. Finish installationLaunch OrvixCode from the desktop shortcut or Start menu.

On first launch, OrvixCode may show setup preferences such as application language, workspace location, and basic behavior.

After setup, the main workspace lets you add a project, start a new chat, or configure AI providers and local models.

Left navigation

The left side gives quick access to projects, chats, tools, and main application areas. It can be collapsed when you need more editor space.

Tabs

Keep multiple chats or workspaces open at the same time. Closing a tab does not necessarily delete the saved chat.

AI chat area

Manage your prompt, active model, and project context here. Responses can also show affected files and the state of proposed edits.

OrvixCode projects are based on local folders. When a folder is selected as the workspace, AI can use task-relevant project context.

New or existing project

• Create a workspace with New Project.
• Attach an existing source-code folder directly.
• Use + New project from the project selector below the message box.

Automatic project creation for AI tasks

If you ask OrvixCode to generate a project and no real project is selected, it can create a project folder and add it to Projects. Duplicate names receive suffixes such as Project 2 or Project 3.

File planning for large projects

Large tasks are split into subsystems and files instead of being compressed into a few oversized files. Imports and dependency contracts are checked while generation proceeds; the manifest can expand when a missing local module is discovered.

Use AI chat to ask project questions, analyze errors, generate code, edit files, or request architecture suggestions.

File and selected-code context

Use the + menu in the composer to attach files. Attached text/code files are explicitly exposed to the AI as available context, and selected-code explanation requests are kept separate from project-generation intent.

Stop a response

While AI is responding, the send control changes to ■ Stop. Stopping cancels continuation and large-project generation steps and prevents late output from being appended.

Streaming view

Markdown and code blocks are formatted while the response streams instead of only after completion. Long user messages preserve line breaks and code blocks.

Prompt example

Review the authentication system in this project.
Find possible failures during session refresh.
If needed, edit the relevant files and explain the changes.

Editing from Ask

Ask is not limited to explanations. When your prompt clearly requests a code change, OrvixCode can prepare an actionable edit package.

AI edits can be more than text in a chat message. OrvixCode can turn a response into a change package that can be reviewed and applied to project files.

Creating new files

Agent tasks can create new files when the requested feature or refactor requires them.

The 1.2.0 code editor keeps the existing editing surface while adding syntax highlighting and error/warning layers.

Syntax and Problems

• Syntax highlighting for common languages.
• Parse/syntax errors can appear in red; warnings and some missing-module cases in yellow.
• Problematic code ranges can be underlined directly in the editor.

AI editing

• ★ Fix with AI: Fixes the open file from the Problems area without auto-saving.
• Edit with AI…: Sends only the selected text range plus your instruction and splices the replacement back into that range.
• Explain selected code with OrvixCode: Sends the selection as explanation context.

During inline AI editing, the selected range can temporarily blur and show ★ AI editing…. AI editor changes also participate in Ctrl+Z / Ctrl+Y history.

When running Python files, OrvixCode prefers the project .venv or venv interpreter when available.

Automatic missing-package install

If execution raises ModuleNotFoundError and the missing name is not a local project module, OrvixCode can try pip with the same interpreter and then retry the file.

Some import names map to package names, including cv2 → opencv-python, PIL → Pillow, yaml → PyYAML, and sklearn → scikit-learn.

Fix with AI

If the run fails, ★ Fix with AI can send the error, active file, and relevant context to the AI.

Use the built-in terminal like a normal command terminal.

• cd changes persist for the terminal session.
• clear / cls clears the output.
• Command history is stored for the session; use Up/Down to revisit commands.

Fix terminal errors with AI

When a command fails, ★ Fix with AI can send the last command, error output, and the open file when relevant.

OrvixCode tracks the state of AI edit packages.

• Diff / Preview: compare the proposed or applied change.
• Apply: write the prepared change to the file.
• Applied: keep a stable state for a change that was already written.
• Undo: revert the latest supported AI change.

OrvixCode provides a Visual Studio Code extension that opens an OrvixCode panel in the Activity Bar and brings the desktop app's cloud and local AI workflows into the editor.

Main capabilities

• AI chat panel with Ask, Edit, and Agent modes
• Cloud and local Ollama model selection
• Active-file and selected-code context
• Context-menu actions for selected code
• User-approved file edits, diff preview, and undo
• Model connection and error status
• Optional chat synchronization with the desktop app

Quick AI actions

The extension quick actions now go beyond explanation and basic fixes:

• Explain code
• Fix errors
• Refactor
• Review project
• Generate tests
• Code review
• Generate documentation
• Suggest performance improvements

Install OrvixCode from the Visual Studio Code Extensions view. Regular users do not need to manage VSIX packages manually.

1. Open ExtensionsUse the Extensions icon or Ctrl+Shift+X.
2. Search for OrvixCodeUse the Marketplace search box.
3. Install the extensionVerify the publisher shown for OrvixCode and choose Install.
4. Open the panelSelect the OrvixCode icon from the Activity Bar.
5. Connect the desktop appMake sure OrvixCode desktop is installed for supported AI workflows.

Select a code block in the editor and use the OrvixCode commands from the context menu.

Agent mode is designed for tasks that are not limited to one file. It can coordinate work across the open workspace.

Good Agent tasks

• Add a new feature
• Fix a bug that spans multiple files
• Create a module or supporting files
• Refactor project structure
• Coordinate HTML, CSS, and JavaScript changes

Supported chat and edit activity from the VS Code extension can be synchronized into the OrvixCode desktop history.

Synced sessions may appear with a VS Code · … prefix, and edit events can be associated with the corresponding chat.

Viona can be used as the default or primary model inside OrvixCode. Available model names and options can vary by version and configuration.

The VS Code panel can display the active model together with its connection state.

OrvixCode can run supported models locally through Ollama. Local model requests are sent to the Ollama service on your computer and do not require a separate cloud AI API key.

Installed models can appear under Local / Ollama in the model selector and can be used in compatible Ask, Edit, Agent, and Combo workflows.

Choosing a model

Use smaller coding models for limited-memory systems and larger models when you need more reasoning capacity and your hardware can support the memory requirements.

The local model center is available under Settings → Local Models. It manages the Ollama runtime, hardware checks, free disk space, installed models, and troubleshooting tools.

1. Open Local ModelsCheck whether the Ollama runtime is available.
2. Prepare OllamaIf it is not installed, use the supported setup flow and verify the local service.
3. Check compatibilityReview RAM, VRAM, and free-disk recommendations before downloading.
4. Download a modelUse pause, resume, or cancel controls while the model is being downloaded.
5. Select the modelAfter installation it should appear in the main model selector.

Local model tools

• Start the Ollama service
• Retry GPU mode after a safe CPU fallback
• Open runtime logs
• Open the model storage folder
• Run compatibility checks

OrvixCode can centralize Viona and supported external AI providers in one provider screen. The model list can be refreshed based on the models available to your account or API access.

Combo mode lets you assign different models to different stages of a coding task. For example, one model can plan, another can write code, and a third can review the result.

OrvixCode prepares context according to the task instead of sending every project file on every request. Selected code, the active file, relevant project files, and chat history can be prioritized differently.

Adaptive context

When a request is too large for a model or local memory budget, OrvixCode can reduce context and retry with a smaller working set.

OrvixCode detects requests that clearly require large output even when the prompt is short. This is not limited to Python; long-code tasks are supported across HTML/CSS/JavaScript/TypeScript, Go, Rust, PHP, Ruby, Java/Kotlin, C/C++/C#, Swift, Dart, Lua, Scala, Elixir, F#, Solidity, Bash, PowerShell, SQL, and more.

Long output and continuation

• Local models can use a larger output budget when appropriate.
• If a response is cut by a token/context limit, OrvixCode can request continuation from the missing point.
• Repeated overlap is cleaned when continuation chunks are merged.
• Content from an unfinished fenced code block can still be preserved as code.

Large-project planning

1. The request is analyzed and a project structure is planned.
2. A manifest/file plan is created.
3. Files are generated while dependencies and contracts are tracked.
4. If imports reveal a missing local module, the plan can be expanded.
5. Generation continues when output limits are reached.
6. Basic consistency checks run at the end.

Use the GitHub import flow to bring a repository into a local OrvixCode workspace. After the clone/import completes, the project behaves like a normal local folder and can participate in AI context.

OrvixCode can keep backup information around file edits to reduce the risk of data loss. Depending on the edit workflow, earlier content can be retained so changes can be reviewed or reverted.

The VS Code integration also provides an undo flow for supported recent AI changes.

Open tabs and saved chats are separate concepts. Closing a tab does not have to delete its conversation history.

• Reopen saved chats from the navigation area.
• Delete chats from the relevant management controls.
• Keep multiple chat tabs open at once.
• Synced VS Code sessions can appear under their own titles.

The settings area groups the main OrvixCode preferences:

• General: default model, creativity, writing animation, sounds, navigation, and language.
• Customize: theme, font family, and text appearance.
• AI Providers: provider connections, model refresh, and Combo configuration.
• Local Models: Ollama service, compatibility, downloads, and logs.
• Storage & Backup: data location and automatic file backup behavior.
• Chat Management: review and remove stored conversations.
• Setup / VS Code: desktop and extension integration options.
• Troubleshooter: local system and connection checks.
• Keyboard Shortcuts: Rebind actions such as send, new chat/project, save, undo/redo, run, terminal, stop AI response, Combo, and main navigation.

Open Settings → Keyboard Shortcuts to rebind major application actions. Selecting an action opens a key-capture dialog for the new combination.

Examples of configurable actions

• Send message / new line
• Focus the message box
• New chat and new project
• Attach file and save
• Ctrl+Z undo / Ctrl+Y redo
• Run code and open terminal
• Stop AI response
• Toggle Combo Mode
• Docs, feedback, and main navigation
• Minimize, maximize, close, or send the app to background/tray

Open Settings → Customize to change OrvixCode's visual appearance.

OrvixCode aims to keep AI-driven file operations visible and reviewable. The edited file and change state can be shown in the chat or edit report.

When automatic apply is disabled, a prepared edit can wait for user approval before being written.

External AI providers require the credentials issued by the relevant service.

The VS Code extension is designed to use the provider configuration exposed through OrvixCode's connection layer rather than asking you to copy secrets into every editor workflow.

Multi-file tasks can have a wider change scope than a single-file edit. Depending on your version and settings, OrvixCode can request permission for one file or for a broader proposed change.

AI is not responding

• Confirm that an active model is selected and available.
• For cloud models, verify provider credentials and internet access.
• For local models, confirm that the Ollama service is running.
• In VS Code, use the connection test and review any error details.

File changes are not applying

• Check the edit package state.
• If Auto Apply is enabled, the change may already be applied.
• Verify write permissions and VS Code Workspace Trust.
• Review the diff and retry the edit if needed.

This error can appear when the active file, project context, and chat history together exceed the model request limit.

OrvixCode can reduce the context size and retry with a smaller context profile after an oversized request.

Extension is missing from search

Refresh the Extensions view, verify internet access, and search for OrvixCode again. Newly published updates can take time to propagate.

Workspace is not trusted

File-editing features may be restricted until the workspace is trusted in VS Code.

Panel opens but AI does not respond

Run the connection test, then check cloud-provider access or the local OrvixCode/Ollama status.

Simple chat includes irrelevant project context

Update the extension and reload VS Code. Current builds are designed to isolate simple greetings from project context.

Context-menu command does not start

Confirm that code is selected, retry the command, and use a new OrvixCode chat if the issue persists.

Ollama is not running

Open Settings → Local Models, check the runtime state, and use the service start control. Review logs if startup continues to fail.

“failed to allocate buffer / kv cache”

The selected model or context length may be too large for available RAM/VRAM. OrvixCode can retry with smaller context levels.

8192 → 6144 → 4096 → 3072 → 2048

CUDA / GPU error

A safe CPU fallback can keep the model usable. You can retry GPU mode later from the local model controls.

Local model is too slow

Try a smaller model, close memory-heavy applications, and confirm that compatible GPU acceleration is active.

OrvixCode Learn complète la documentation de référence avec des leçons pratiques, progressives et directement applicables.

La page officielle affiche l’état du site, de la documentation, des téléchargements et des services IA. Les annonces publiées sont immédiatement disponibles sur la page publique et les flux lisibles par machine.

Les liens de téléchargement utilisent un réglage de version centralisé. Lorsqu’un administrateur publie une nouvelle version stable, le numéro et la destination sont mis à jour sur toute la plateforme.

https://orvixcode.orvixgames.com/download.php?lang=fr-fr

Télécharger la dernière version →