Apaya Coding Assistant

Local-first AI coding assistance for VS Code. Apaya Coding Assistant brings Cursor-style inline suggestions, quick fixes, selected-code edits, and an audit log for generated completions directly into your editor.

Features

Inline Ghost Text - Get AI-powered inline code suggestions while typing and accept them with Tab.
Diagnostic-Aware Fixes - Use the lightbulb action to fix syntax and code issues in place.
Selected-Code Editing - Select code, describe the change, and apply the generated edit from VS Code.
Local Ollama Models - Runs locally by default with Ollama; no cloud API key is required for normal use.
Persistent Inline Log - Review generated, empty, error, and accepted inline suggestion events in a JSONL audit log.
Status Bar Proof - The status bar shows when Apaya is generating, ready, idle, or accepted.
Sidebar Chat - Ask coding questions from the Apaya activity bar when you want a conversational workflow.
Code Generation Commands - Generate common snippets such as React components, TypeScript interfaces, API routes, unit tests, and functions.

Requirements

Install and run Ollama for local AI suggestions.

Recommended local models for a 32 GB RAM machine:

Use	Recommended model
Fast inline autocomplete	`qwen2.5-coder:7b`
Stronger fixes and selected-code edits	`qwen2.5-coder:14b`
Lighter fallback	`qwen2.5-coder:3b`

Pull the recommended models:

ollama pull qwen2.5-coder:7b
ollama pull qwen2.5-coder:14b

Usage

Inline suggestions

Open a code file.
Start typing or place the cursor where you want a suggestion.
Wait for ghost text to appear.
Press Tab to accept the suggestion.

If you want to manually request a suggestion, run:

Apaya Coding Assistant: Trigger Inline Suggestion

Make sure VS Code inline suggestions are enabled:

{
  "editor.inlineSuggest.enabled": true
}

Quick fixes

When VS Code shows a diagnostic, open the lightbulb menu and choose:

Fix with Apaya Coding Assistant

You can also run:

Apaya Coding Assistant: Fix Current Issue

Inline suggestion proof log

Open the persistent audit log with:

Apaya Coding Assistant: Open Inline Suggestion Log

The log records events such as:

request:start
request:suggested
request:empty
request:error
request:accepted

Each suggested or accepted entry includes the file, language, line, column, and preview text generated by Apaya.

Commands

Command	Description
`Apaya Coding Assistant: Trigger Inline Suggestion`	Request ghost-text completion at the cursor
`Apaya Coding Assistant: Open Inline Suggestion Log`	Open the persistent inline suggestion audit log
`Apaya Coding Assistant: Fix Current Issue`	Fix the diagnostic or selected issue in the current editor
`Apaya Coding Assistant: Ask Assistant`	Ask about selected code
`Apaya Coding Assistant: Edit with Assistant`	Edit selected code using an AI instruction
`Apaya Coding Assistant: Generate React Component`	Generate a React component snippet
`Apaya Coding Assistant: Generate TypeScript Interface`	Generate a TypeScript interface snippet
`Apaya Coding Assistant: Generate API Route`	Generate an API route snippet
`Apaya Coding Assistant: Generate Unit Test`	Generate a unit test snippet
`Apaya Coding Assistant: Generate Function`	Generate a function snippet

Settings

Setting	Default	Description
`apayaAssistant.ollama.baseUrl`	`http://localhost:11434`	Ollama server URL
`apayaAssistant.ollama.model`	`qwen2.5-coder:14b`	Model used for stronger fixes, chat, and selected-code edits
`apayaAssistant.inline.enabled`	`true`	Enable ghost-text inline suggestions
`apayaAssistant.inline.model`	`qwen2.5-coder:7b`	Model used for inline autocomplete
`apayaAssistant.inline.debounceMs`	`180`	Delay before requesting an inline suggestion
`apayaAssistant.inline.timeoutMs`	`7000`	Maximum time to wait for an inline suggestion
`apayaAssistant.inline.maxContextLines`	`40`	Surrounding lines sent as local context
`apayaAssistant.inline.fixDiagnostics`	`true`	Suggest same-line repairs when diagnostics are present
`apayaAssistant.context.linesAroundSelection`	`20`	Surrounding lines used for selected-code commands

Privacy

Apaya Coding Assistant is local-first. By default, code context is sent only to your local Ollama server at localhost:11434.

The inline suggestion audit log is stored locally in VS Code global storage. It may contain generated code previews and file paths, so review it before sharing screenshots or logs.

Troubleshooting

If ghost text does not appear, confirm editor.inlineSuggest.enabled is true.
If the status bar shows timeouts, make sure Ollama is running and the configured model is installed.
If another AI extension is also installed, VS Code may show suggestions from either provider. Use Apaya Coding Assistant: Open Inline Suggestion Log to confirm what Apaya generated.
If the activity bar or settings still show an old name after an upgrade, reload VS Code with Developer: Reload Window.

License

See the included LICENSE file.