Gateway Integration

The l33t gateway is an optional AI processing backend that enhances your dictation before it reaches your editor — formatting, cleanup, and language model enhancement.

How It Works

Mic → OS Dictation → raw text → Gateway → enhanced text → Editor
                                     │
                               Mode-aware routing
                                     │
                         ┌───────────┼───────────┐
                         ▼           ▼           ▼
                    Rules Engine  DeepSeek     OpenAI
                       (free)    (cheapest)  (strongest)

When the gateway is enabled, every utterance goes through a processing pipeline: authentication → rate limiting → credit check → model selection → processing → response → insert.

Prerequisites

• Docker and Docker Compose
• pnpm (corepack enable pnpm)
• Node.js 20+

Starting the Gateway

# From the l33tdev/ monorepo root
export GATEWAY_API_KEY="dev-key-12345"
docker compose up -d --build

This starts four services:

Service	Port	Purpose
Gateway API	3000	Processing endpoint
PostgreSQL	5432	Request/response persistence
Redis	6379	Rate limiting cache
Nginx	80	Reverse proxy

Verify it's running:

curl https://gateway.l33tspeak.dev/health

Enabling the Gateway

1. Open the Command Palette (Cmd+Shift+P)
2. Run "L33t Speak: Set Gateway API Key"
3. Enter your API key (default: dev-key-12345)

Setting the API key automatically enables the gateway. No additional config toggle needed.

You can also control it manually in VS Code settings:

Setting	Default	Description
l33t-speak.enableGateway	false	Toggle gateway processing on/off
l33t-speak.gatewayUrl	https://gateway.l33tspeak.dev	Gateway API base URL
l33t-speak.gatewayApiKey	""	Placeholder — actual key stored in OS keychain

The API key is stored securely in your operating system's keychain via VS Code's SecretStorage — never in plaintext settings.

Viewing Gateway Logs

Every request and response is logged to a dedicated output channel:

1. Open the Output panel (Cmd+Shift+U)
2. Select "L33t Gateway" from the dropdown

You'll see:

[2026-05-29T06:06:13.855Z] POST https://gateway.l33tspeak.dev
  Raw text (104 chars): "Create me a skill that looks up data..."
Mode: markdown (auto-detected)
  Status: 200
  Route: rules-engine
  Model: unknown
  Latency: 0ms
  Cost: $0.000000
  Processed (105 chars): "Create me a skill that looks up data..."

For live container logs:

docker compose logs gateway -f

Processing Indicator

While the gateway processes your dictation, the status bar shows an animated rainbow indicator:

🟥 Enhancing... (red)
🟧 Enhancing... (orange)
🟨 Enhancing... (yellow)
🟩 Enhancing... (green)
🟦 Enhancing... (blue)
🟪 Enhancing... (purple)

The background color cycles through the rainbow while a spinner animates. The indicator is blue/cyan for local processing (rules-engine) and full rainbow for hosted LLM calls (DeepSeek/OpenAI).

One-Command Dev Environment

./scripts/start-dev.sh

This starts Docker, builds the extension, and launches VS Code with the Extension Development Host pre-configured for gateway use.

See Also

• Processing Modes — all 6 modes and how to choose
• Configuration — all gateway settings