Services Layer Patterns

This skill documents how to implement services in the Whispering architecture. Services are pure, isolated business logic with no UI dependencies that return Result<T, E> types for error handling.

When to Apply This Skill

Use this pattern when you need to:

• Create a new service with domain-specific error handling

• Add error types with structured context (like HTTP status codes)

• Understand how services are organized and exported

• Implement platform-specific service variants (desktop vs web)

Core Architecture

Services follow a three-layer architecture: Service → Query → UI

┌─────────────┐ ┌─────────────┐ ┌──────────────┐ │ UI │ --> │ RPC/Query │ --> │ Services │ │ Components │ │ Layer │ │ (Pure) │ └─────────────┘ └─────────────┘ └──────────────┘

Services are:

• Pure: Accept explicit parameters, no hidden dependencies

• Isolated: No knowledge of UI state, settings, or reactive stores

• Testable: Easy to unit test with mock parameters

• Consistent: All return Result<T, E> types for uniform error handling

Creating Errors with defineErrors

Every service defines domain-specific errors using defineErrors from wellcrafted. Errors are grouped into a namespace object where each key becomes a variant.

import { defineErrors, type InferError, type InferErrors, extractErrorMessage } from 'wellcrafted/error'; import { Err, Ok, type Result, tryAsync, trySync } from 'wellcrafted/result';

// Namespace-style error definition — name describes the domain const CompletionError = defineErrors({ ConnectionFailed: ({ cause }: { cause: unknown }) => ({ message: Connection failed: ${extractErrorMessage(cause)}, cause, }), EmptyResponse: ({ providerLabel }: { providerLabel: string }) => ({ message: ${providerLabel} API returned an empty response, providerLabel, }), MissingParam: ({ param }: { param: string }) => ({ message: ${param} is required, param, }), });

// Type derivation — shadow the const with a type of the same name type CompletionError = InferErrors<typeof CompletionError>; type ConnectionFailedError = InferError<typeof CompletionError.ConnectionFailed>;

// Call sites — each variant returns Err<...> directly return CompletionError.ConnectionFailed({ cause: error }); return CompletionError.EmptyResponse({ providerLabel: 'OpenAI' }); return CompletionError.MissingParam({ param: 'apiKey' });

How defineErrors Works

defineErrors({ ... }) takes an object of factory functions and returns a namespace object. Each key becomes a variant:

• name is auto-stamped from the key (e.g., key NotFound → error.name === 'NotFound' )

• The factory function IS the message generator — it returns { message, ...fields }

• Each variant returns Err<...> directly — no separate FooErr constructor needed

• Types use InferError / InferErrors — not ReturnType

// No-input variant (static message) const RecorderError = defineErrors({ Busy: () => ({ message: 'A recording is already in progress', }), });

// Usage — no arguments needed return RecorderError.Busy();

// Variant with derived fields — constructor extracts from raw input const HttpError = defineErrors({ Response: ({ response, body }: { response: { status: number }; body: unknown }) => ({ message: HTTP ${response.status}: ${extractErrorMessage(body)}, status: response.status, body, }), });

// Usage — pass raw objects, constructor derives fields return HttpError.Response({ response, body: await response.json() }); // error.message → "HTTP 401: Unauthorized" // error.status → 401 (derived from response, flat on the object) // error.name → "Response"

Error Type Examples from the Codebase

// Static message, no input needed const RecorderError = defineErrors({ Busy: () => ({ message: 'A recording is already in progress', }), }); RecorderError.Busy()

// Multiple related errors in a single namespace const HttpError = defineErrors({ Connection: ({ cause }: { cause: unknown }) => ({ message: Failed to connect to the server: ${extractErrorMessage(cause)}, cause, }), Response: ({ response, body }: { response: { status: number }; body: unknown }) => ({ message: HTTP ${response.status}: ${extractErrorMessage(body)}, status: response.status, body, }), Parse: ({ cause }: { cause: unknown }) => ({ message: Failed to parse response body: ${extractErrorMessage(cause)}, cause, }), });

// Union type for the whole namespace type HttpError = InferErrors<typeof HttpError>;

// Individual variant type type ConnectionError = InferError<typeof HttpError.Connection>;

Anti-Pattern: Discriminated Union Inputs

String literal unions inside error factory inputs are a code smell. When a variant's input contains a field like reason: 'a' | 'b' | 'c' or operation: 'x' | 'y' | 'z' , you're creating a sub-discriminant that duplicates what defineErrors already provides at the top level.

The Problem

// BAD: Sub-discriminant forces double narrowing and dishonest types const ShortcutError = defineErrors({ InvalidAccelerator: (input: { reason: 'invalid_format' | 'no_key_code' | 'multiple_key_codes'; accelerator?: string; // Optional because some reasons don't use it }) => { const messages = { invalid_format: Invalid format: '${input.accelerator}', no_key_code: 'No valid key code found', multiple_key_codes: 'Multiple key codes not allowed', }; return { message: messages[input.reason], ...input }; }, });

Why this is bad:

• Double narrowing: Consumers must narrow on error.name then on error.reason

• Dishonest types: accelerator is optional because some reasons don't need it, but the type doesn't express which ones do

• Obscured intent: The reason field is doing the discriminant's job — that's what variant names are for

The Fix: Split Into Separate Variants

// GOOD: Each variant has exactly the fields it needs const ShortcutError = defineErrors({ InvalidFormat: ({ accelerator }: { accelerator: string }) => ({ message: Invalid accelerator format: '${accelerator}', accelerator, }), NoKeyCode: () => ({ message: 'No valid key code found in pressed keys', }), MultipleKeyCodes: () => ({ message: 'Multiple key codes not allowed in accelerator', }), });

Why this is better:

• Single narrowing: error.name === 'NoKeyCode' — done

• Honest types: InvalidFormat requires accelerator , NoKeyCode takes nothing

• Self-documenting: Variant names describe the error, no lookup table needed

When This Applies

Split whenever you see:

• reason: 'a' | 'b' | 'c' with a message lookup table

• operation: 'x' | 'y' | 'z' with different messages per operation

• errorKind: ... or type: ... acting as a sub-discriminant

• Optional fields that exist because "some variants" don't use them

The whole point of defineErrors is that each variant is a first-class citizen with its own name and shape. Collapsing them behind string unions saves a few lines of definition at the cost of weaker types and double-narrowing at every consumer.

Exception: When It's Genuinely One Error

If the string literal truly is a field and not a sub-discriminant — e.g., the consumer doesn't switch on it — then it's fine:

// OK: 'operation' is metadata for logging, not a sub-discriminant const FsError = defineErrors({ ReadFailed: ({ path, cause }: { path: string; cause: unknown }) => ({ message: Failed to read '${path}': ${extractErrorMessage(cause)}, path, cause, }), WriteFailed: ({ path, cause }: { path: string; cause: unknown }) => ({ message: Failed to write '${path}': ${extractErrorMessage(cause)}, path, cause, }), });

Anti-Pattern: Conditional Logic on Factory Inputs

If a variant constructor uses if/switch on its own input fields to decide the message or behavior, each branch should be its own variant. This is a generalization of the string literal union rule above — any branching inside a constructor means multiple errors are hiding in one variant.

The Problem

// BAD: Constructor branches on inputs — multiple errors hiding in one variant const FormError = defineErrors({ Validation: ({ field, value, receivedType }: { field?: string; // Optional because not every branch uses it value?: string; // Optional because not every branch uses it receivedType?: string; // Optional because not every branch uses it }) => ({ message: (() => { if (field === 'email' && value) return Invalid email address: '${value}'; if (field === 'password' && value) return Password too weak: must be at least 8 characters; if (field === 'confirmPassword') return 'Passwords do not match'; if (receivedType) return Invalid form data: expected string, got ${receivedType}; return 'Form submission failed'; })(), field, value, receivedType, }), });

Symptoms:

• Dishonest optionals: Fields are optional because no single call site uses them all — the type lies about what each error actually carries

• Hidden branching: Consumers must inspect fields beyond name to know the real error kind — name === 'Validation' tells you nothing

• Untypeable messages: The message depends on runtime field combinations, so TypeScript can't narrow to a specific message shape

The Fix: Flatten Each Branch Into Its Own Variant

// GOOD: Each branch becomes its own variant with honest, required fields const FormError = defineErrors({ InvalidEmail: ({ value }: { value: string }) => ({ message: Invalid email address: '${value}', value, }), WeakPassword: () => ({ message: 'Password too weak: must be at least 8 characters', }), PasswordMismatch: () => ({ message: 'Passwords do not match', }), InvalidFormData: ({ receivedType }: { receivedType: string }) => ({ message: Invalid form data: expected string, got ${receivedType}, receivedType, }), SubmissionFailed: () => ({ message: 'Form submission failed', }), });

Why this is better:

• Honest types: InvalidEmail requires value , WeakPassword takes nothing — no dishonest optionals

• Single narrowing: error.name === 'InvalidEmail' tells you everything

• Typeable messages: Each variant has a deterministic message shape

Rule of Thumb

If the constructor branches on its inputs to decide the message, each branch should be its own variant. The branching is the evidence that you have multiple distinct errors collapsed into one.

This applies to:

• If/else chains in message construction (including IIFEs)

• Switch statements on input fields

• Ternary expressions that pick between fundamentally different messages

• Lookup tables keyed on input fields (covered by the string literal union rule above)

See also: docs/core/error-system.mdx § "3b. Avoid Conditional Logic on Factory Inputs" for the canonical reference with full examples.

Service Implementation Pattern

Basic Service Structure

import { defineErrors, type InferErrors, extractErrorMessage } from 'wellcrafted/error'; import { Err, Ok, type Result, tryAsync, trySync } from 'wellcrafted/result';

// 1. Define domain-specific errors — variant names describe failure modes const AutostartError = defineErrors({ CheckFailed: ({ cause }: { cause: unknown }) => ({ message: Failed to check autostart: ${extractErrorMessage(cause)}, cause, }), EnableFailed: ({ cause }: { cause: unknown }) => ({ message: Failed to enable autostart: ${extractErrorMessage(cause)}, cause, }), DisableFailed: ({ cause }: { cause: unknown }) => ({ message: Failed to disable autostart: ${extractErrorMessage(cause)}, cause, }), }); type AutostartError = InferErrors<typeof AutostartError>;

// 2. Create factory function that returns service object export function createAutostartService() { return { async isEnabled(): Promise<Result<boolean, AutostartError>> { return tryAsync({ try: () => isEnabled(), catch: (error) => AutostartError.CheckFailed({ cause: error }), }); }, async enable(): Promise<Result<void, AutostartError>> { return tryAsync({ try: () => enable(), catch: (error) => AutostartError.EnableFailed({ cause: error }), }); }, }; }

// 3. Export the "Live" instance (production singleton) export type AutostartService = ReturnType<typeof createAutostartService>; export const AutostartServiceLive = createAutostartService();

Real-World Example: Recorder Service

// From apps/whispering/src/lib/services/isomorphic/recorder/navigator.ts

const RecorderError = defineErrors({ AlreadyRecording: () => ({ message: 'A recording is already in progress. Please stop the current recording.', }), StreamAcquisition: ({ cause }: { cause: unknown }) => ({ message: Failed to acquire recording stream: ${extractErrorMessage(cause)}, cause, }), InitFailed: ({ cause }: { cause: unknown }) => ({ message: Failed to initialize recorder. ${extractErrorMessage(cause)}, cause, }), }); type RecorderError = InferErrors<typeof RecorderError>;

export function createNavigatorRecorderService(): RecorderService { let activeRecording: ActiveRecording | null = null;

return {
	getRecorderState: async (): Promise<
		Result<WhisperingRecordingState, RecorderError>
	> => {
		return Ok(activeRecording ? 'RECORDING' : 'IDLE');
	},

	startRecording: async (
		params: NavigatorRecordingParams,
		{ sendStatus },
	): Promise<Result<DeviceAcquisitionOutcome, RecorderError>> => {
		// Validate state
		if (activeRecording) {
			return RecorderError.AlreadyRecording();
		}

		// Get stream (calls another service)
		const { data: streamResult, error: acquireStreamError } =
			await getRecordingStream({ selectedDeviceId, sendStatus });

		if (acquireStreamError) {
			return RecorderError.StreamAcquisition({
				cause: acquireStreamError,
			});
		}

		// Initialize MediaRecorder
		const { data: mediaRecorder, error: recorderError } = trySync({
			try: () =>
				new MediaRecorder(stream, {
					bitsPerSecond: Number(bitrateKbps) * 1000,
				}),
			catch: (error) =>
				RecorderError.InitFailed({ cause: error }),
		});

		if (recorderError) {
			cleanupRecordingStream(stream);
			return Err(recorderError);
		}

		// Store state and start
		activeRecording = {
			recordingId,
			stream,
			mediaRecorder,
			recordedChunks: [],
		};
		mediaRecorder.start(TIMESLICE_MS);

		return Ok(deviceOutcome);
	},
};

}

export const NavigatorRecorderServiceLive = createNavigatorRecorderService();

Namespace Exports Pattern

Services are organized hierarchically and re-exported as namespace objects:

Folder Structure

services/ ├── desktop/ # Desktop-only (Tauri) │ ├── index.ts # Re-exports as desktopServices │ ├── command.ts │ └── ffmpeg.ts ├── isomorphic/ # Cross-platform │ ├── index.ts # Re-exports as services │ ├── transcription/ │ │ ├── index.ts # Re-exports as transcriptions namespace │ │ ├── cloud/ │ │ │ ├── openai.ts │ │ │ └── groq.ts │ │ └── local/ │ │ └── whispercpp.ts │ └── completion/ │ ├── index.ts │ └── openai.ts ├── types.ts └── index.ts # Main entry point

Index File Pattern

// services/isomorphic/transcription/index.ts export { OpenaiTranscriptionServiceLive as openai } from './cloud/openai'; export { GroqTranscriptionServiceLive as groq } from './cloud/groq'; export { WhispercppTranscriptionServiceLive as whispercpp } from './local/whispercpp';

// services/isomorphic/index.ts import * as transcriptions from './transcription'; import * as completions from './completion';

export const services = { db: DbServiceLive, sound: PlaySoundServiceLive, transcriptions, // Namespace import completions, // Namespace import } as const;

// services/index.ts (main entry) export { services } from './isomorphic'; export { desktopServices } from './desktop';

Consuming Services

// In query layer or anywhere import { services, desktopServices } from '$lib/services';

// Access via namespace await services.transcriptions.openai.transcribe(blob, options); await services.transcriptions.groq.transcribe(blob, options); await services.db.recordings.getAll(); await desktopServices.ffmpeg.compressAudioBlob(blob, options);

Platform-Specific Services

For services that need different implementations per platform:

Define Shared Interface

// services/isomorphic/text/types.ts export type TextService = { readFromClipboard(): Promise<Result<string | null, TextError>>; copyToClipboard(text: string): Promise<Result<void, TextError>>; writeToCursor(text: string): Promise<Result<void, TextError>>; };

Implement Per Platform

// services/isomorphic/text/desktop.ts const TextError = defineErrors({ ClipboardWriteFailed: ({ cause }: { cause: unknown }) => ({ message: Clipboard write failed: ${extractErrorMessage(cause)}, cause, }), });

export function createTextServiceDesktop(): TextService { return { copyToClipboard: (text) => tryAsync({ try: () => writeText(text), // Tauri API catch: (error) => TextError.ClipboardWriteFailed({ cause: error }), }), }; }

// services/isomorphic/text/web.ts export function createTextServiceWeb(): TextService { return { copyToClipboard: (text) => tryAsync({ try: () => navigator.clipboard.writeText(text), // Browser API catch: (error) => TextError.ClipboardWriteFailed({ cause: error }), }), }; }

Build-Time Platform Detection

// services/isomorphic/text/index.ts export const TextServiceLive = window.TAURI_INTERNALS ? createTextServiceDesktop() : createTextServiceWeb();

Error Message Best Practices

Write error messages that are:

• User-friendly: Explain what happened in plain language

• Actionable: Suggest what the user can do

• Detailed: Include technical details for debugging

Choosing the right approach

• No-input variants for static messages (e.g., Busy: () => ({ message: '...' }) )

• Field-based variants when the message is computed from structured input

• Separate variants when different error conditions need different fields (see Anti-Pattern section above)

const RecorderError = defineErrors({ // Static message — no input needed Busy: () => ({ message: 'A recording is already in progress', }),

// Message computed from fields HttpResponse: ({ status }: { status: number }) => ({ message: HTTP ${status} response, status, }),

// Wrapping an unknown cause with context MicrophoneUnavailable: ({ cause }: { cause: unknown }) => ({ message: Unable to connect to the selected microphone: ${extractErrorMessage(cause)}, cause, }),

// User-actionable message with file context ConfigParseFailed: ({ filename, cause }: { filename: string; cause: unknown }) => ({ message: Failed to parse configuration file. Please check that ${filename} contains valid JSON. ${extractErrorMessage(cause)}, filename, cause, }), });

Key Rules

• Services never import settings - Pass configuration as parameters

• Services never import UI code - No toasts, no notifications, no WhisperingError

• Always return Result types - Never throw errors

• Use trySync/tryAsync - See the error-handling skill for details

• Export factory + Live instance - Factory for testing, Live for production

• Use defineErrors namespaces - Group related errors under a single namespace

• Derive types with InferError/InferErrors - Not ReturnType

• Variant names describe the failure mode - Never use generic names like Service , Error , or Failed . The namespace provides domain context (RecorderError ), so the variant must say what went wrong (AlreadyRecording , InitFailed , StreamAcquisition ). RecorderError.Service is meaningless — RecorderError.AlreadyRecording tells you exactly what happened.

• Split discriminated union inputs - Each variant gets its own name and shape. If the constructor branches on its inputs (if/switch/ternary) to decide the message, each branch should be its own variant

• Transform cause in the constructor, not the call site - Accept cause: unknown and call extractErrorMessage(cause) inside the factory's message template. Call sites pass the raw error: { cause: error } . This centralizes message extraction where the message is composed and keeps call sites clean.

References

• See apps/whispering/src/lib/services/README.md for architecture details

• See the query-layer skill for how services are consumed

• See the error-handling skill for trySync/tryAsync patterns