@pheme-ai/chat

Angular chat primitives for prototyping AI chat applications with Pheme.

The package contains reusable standalone components, models, services, and stream parsing helpers. It is designed to be used directly in product prototypes or through the demo pages exported from per-page secondary entrypoints.

Exports

Primary entrypoint @pheme-ai/chat:

• provideChat and CHAT_CONFIG for API and route configuration.
• ChatComposer for message input.
• ChatSidenav for chat navigation shells.
• MessageList for rendering user and assistant messages.
• Chat models including Message, UserMessage, AssistantMessage, and ChatSummary.
• ChatClient and ChatStore for API access and UI state orchestration.
• parseJsonLines for NDJSON response streams.

Secondary entrypoint @pheme-ai/chat/pages/home:

• HomePage for starting a new chat.

Secondary entrypoint @pheme-ai/chat/pages/chat:

• ChatPage for displaying and continuing an existing chat.

API Expectations

The default configuration uses /api as the base URL and expects endpoints compatible with the root workspace OpenAPI contract:

• GET /chats
• POST /chats
• GET /chats/{chatId}/messages
• POST /chats/{chatId}/messages

Create and send operations stream newline-delimited JSON events so the UI can show saved user messages, in-progress assistant responses, deltas, completion, and errors.

Configuration

Register the package in the consuming application configuration:

import { provideChat } from '@pheme-ai/chat';

export const appConfig = {
  providers: [provideChat({ apiBaseUrl: '/api' })],
};

Available configuration fields:

• apiBaseUrl - base URL for chat API requests.
• chatPath - router commands used for existing chat links.
• newChatPath - router commands used for starting a new chat.

ChatSidenav also accepts title and brandLogoUrl inputs so a demo shell can carry the consuming app's brand without forking the component.

Build And Test

From the workspace root:

ng build @pheme-ai/chat
ng test @pheme-ai/chat

Status

This package is optimized for fast prototyping rather than a frozen public API. Prefer small, composable imports when building experiments and use the demo app to validate layout, accessibility, and streaming behavior.