Smart AI Terminal & logs Filter

Filter the noise. Command in plain English.

Smart AI Terminal & logs Filter is a VS Code extension with two core superpowers: it automatically filters terminal output so you only see errors, warnings, and your own logs and it gives you an AI Smart Terminal where you type plain English and get shell commands, confirmed and executed instantly.

Why you need this

The noise problem: Every time you run npm start, python manage.py runserver, or docker compose up, your terminal floods with hundreds of lines you never asked for Next.js banners, webpack progress bars, Spring Boot ASCII art, npm install logs. The one actual error that matters is buried somewhere in the middle.

The command problem: You know what you want to do - "clear npm cache and reinstall dependencies", "kill the process on port 3000", "show me disk usage by folder" - but you have to stop and look up the exact syntax. Every time.

Smart Terminal Filter solves both. Zero configuration required.

Features

AI Smart Terminal

A custom terminal that understands plain English. Open it with Cmd+Shift+Ior by clicking on the icon below on status bar, describe what you want, and the AI translates it into the exact shell command for your platform and shell with a confirmation step before anything runs.

ask ai> kill the process running on port 3000
Thinking...

Suggested command:
  $ lsof -ti:3000 | xargs kill -9
  Finds and kills the process currently listening on port 3000

Run? [y]es / [n]o / [e]dit:

• Destructive command detection: rm, DROP TABLE, truncate, and similar commands get a bold red warning before execution
• Shell passthrough: prefix with ! to bypass AI and run directly (e.g. !./deploy.sh)
• Edit before running : type e at the confirmation prompt to drop the command into the input line for editing
• Context-aware: uses your platform, detected shell, working directory, and last 5 commands as context
• Ctrl+C cancels a running command or dismisses a pending confirmation
• ? or help shows an in-terminal reference
• Falls back to a plain terminal if no API key is configured

Intelligent Noise Filtering

Automatically hides framework boilerplate and build system noise across 8 frameworks and 10+ languages while keeping errors, warnings, and your own console.log output front and center. Up to 95% of irrelevant output is hidden at the default verbosity level.

Verbosity Control

A single slider in the panel gives you 5 levels of detail:

Command Status Banner

A persistent banner at the top of the panel always shows the current state of your terminal:

Stack Trace Grouping

Multi-line stack traces are collapsed into a single expandable block showing the error message, frame count, and source file. Click the block to expand all frames. Click → Go to line to jump directly to the error location in your code.

Error Trend Detection

If the same error appears multiple times across runs it gets a repeat badge - ↻ 2nd time, ↻ 3rd time - so recurring problems are immediately obvious. The first occurrence is retroactively hidden and replaced by the latest one with the badge.

Flash + Toast Notifications

When new errors hit the terminal:

• The status bar flashes red three times
• A toast notification appears: "Smart Terminal: N errors detected in terminal output" with an Open Smart Terminal action button
• Toast notifications are debounced to at most one per 5 seconds

Error Sound Alert

When a terminal command fails with errors, an audio alert plays automatically — so you know something went wrong even when the terminal is off-screen. Disable with smartTerminal.errorSound: false.

Framework Auto-Detection

Detects your project type from package.json, manage.py, pom.xml, go.mod, and other config files, then applies the right noise/signal rules automatically. Supported:

• Next.js / React (CRA + Vite)
• Express / Node.js
• Django / Python
• Spring Boot / Java
• Ruby on Rails
• Docker / Docker Compose
• Go

Custom Patterns

Add your own regex patterns to always hide or always show specific output:

"smartTerminal.customNoisePatterns": ["^\\[HMR\\]", "^webpack compiled"],
"smartTerminal.customSignalPatterns": ["payment", "order.*created"]

AI-Powered Panel Tools (requires API key)

Three AI buttons live inside the filtered log panel:

• Explain Error: plain-English breakdown of what went wrong, the root cause, a suggested fix, and a corrected code snippet. Reads up to 10 lines of source context around the error location automatically.
• Summarize: 2–3 sentence TL;DR of the entire run: outcome, error count, warnings, and key events as bullet points.
• Ask: natural language search through all captured log lines. Type "slow database queries" or "what caused the crash?" and get matching lines plus a conversational answer.

Standalone CLI — smartlog

Filter logs outside VS Code by piping through smartlog:

npm start | smartlog
python app.py | smartlog -v 3
docker logs -f my-container | smartlog -e
go run . 2>&1 | smartlog --summary
cat build.log | smartlog -q "database errors" --json

Status Bar Buttons

After installation, five items appear in the VS Code status bar (bottom-left):

Smart AI Logs Filter Demo

Error Detection

Success State

Smart AI Terminal Demo

Installation

From the VS Code Marketplace (recommended)

1. Open VS Code
2. Press Cmd+Shift+X (Mac) / Ctrl+Shift+X (Windows/Linux) to open Extensions
3. Search Smart Terminal Filter
4. Click Install

The extension activates automatically on startup — no configuration required.

Manual Installation via VSIX

1. Download the .vsix file from the GitHub Releases page
2. Open VS Code
3. Press Cmd+Shift+P → Extensions: Install from VSIX...
4. Select the downloaded .vsix file
5. Reload VS Code when prompted

Install from Source

git clone https://github.com/Apurvashelar/smart-terminal-filter
cd smart-terminal-filter
npm install
npm run build
npm run package        # produces smart-terminal-filter-pro-x.x.x.vsix

Then install the generated .vsix via Extensions: Install from VSIX... as above.

Opening the Filtered Log Panel

The panel docks alongside your terminal. Run any command and filtered output appears immediately.

Opening the AI Smart Terminal

Type natural language or shell commands at the ask ai> prompt. Type ? for a full command reference.

Setting Up AI Features

AI features - Explain Error, Summarize, Ask, and the AI Smart Terminal - require an API key from Claude (Anthropic), OpenAI, or a locally running Ollama instance. Your key is stored in the OS keychain (macOS Keychain / Windows Credential Manager) never in any settings file or on disk.

Step 1: Choose your provider

Open VS Code Settings (Cmd+,) and set:

smartTerminal.ai.provider → claude   (or openai / ollama)

Step 2: Add your API key

Open the Command Palette (Cmd+Shift+P) and run:

Smart Terminal: Set AI API Key

A secure password input box appears. Paste your key and press Enter.

Where to get API keys:

• Claude: console.anthropic.com → API Keys
• OpenAI: platform.openai.com/api-keys
• Ollama: No key needed — set provider to ollama and make sure Ollama is running locally (ollama serve)

Removing your API key

Run Smart Terminal: Clear AI API Key from the Command Palette.

Panel Controls Reference

All Commands

Configuration Reference

CLI Reference — smartlog

Build from source (npm run build), then pipe any command through it:

npm start | smartlog
python app.py 2>&1 | smartlog -v 3
docker logs -f container | smartlog -e
go run . 2>&1 | smartlog --summary
cat build.log | smartlog -q "authentication failures" --json

Output lines are prefixed with level badges: [ERR], [WRN], [USR], [STS], [INF], [DBG].

Languages & Frameworks Supported

Noise filtering for: JavaScript, TypeScript, Python, Ruby, Java, Go, Rust, C#, PHP, Docker

Stack trace detection for: Node.js, Python, Ruby, Java, Go, Rust, C#, PHP

Framework presets for: Next.js, React (CRA/Vite), Express, Django, Spring Boot, Rails, Docker, Go

Privacy

• Your logs never leave your machine unless you explicitly use an AI feature
• AI features send only the relevant error lines, stack trace, and up to 10 lines of source context to the API — not your full log history
• API keys are stored in the OS keychain, never in VS Code settings or any file on disk

License

MIT