BetterChatGPT

Manages conversation state using Zustand store with automatic localStorage persistence, enabling real-time UI updates without server round-trips. Implements unidirectional data flow pattern with minimal boilerplate, storing ChatInterface objects (conversations with messages, metadata, and configuration) directly in browser storage. Supports state migrations for schema evolution and atomic updates across chat, folder, and configuration slices.

Abstracts OpenAI and Azure OpenAI API calls through a service layer that handles streaming responses, token counting, and cost calculation in real-time. Implements fetch-based streaming with incremental message updates, supporting custom proxy endpoints for regional bypass. Automatically calculates token usage per message using model-specific pricing tiers and updates conversation cost metadata without blocking the UI.

Integrates with ShareGPT API to publish conversations publicly and generate shareable links, enabling discovery and reuse of high-quality conversation examples. Implements one-click sharing that uploads conversation JSON to ShareGPT and returns a public URL. Supports importing shared conversations from ShareGPT links back into the application.

Maintains a library of pre-written prompt templates organized by category (e.g., writing, coding, analysis), stored in application state or JSON files. Enables quick insertion of templates into the system prompt or message input with variable substitution. Supports user-created custom prompts saved to the library for reuse across conversations.

Packages the web application as native desktop applications using Electron or similar framework, enabling installation and usage without a web browser. Maintains feature parity with web version while providing native OS integration (system tray, keyboard shortcuts, file associations). Supports auto-updates and offline usage with cached assets.

Integrates with Google Drive API to automatically backup conversations and sync state across devices. Implements OAuth authentication for secure credential handling and periodic sync of chat data to Google Drive. Supports selective sync (backup only, sync only, or bidirectional) and conflict resolution for conversations modified on multiple devices.

Organizes conversations into a tree-structured folder hierarchy stored in Zustand state, with color-coded visual differentiation and search/filter capabilities. Folders are FolderInterface objects with metadata (name, color, nested folder IDs) that enable drag-and-drop reorganization and bulk operations. Supports auto-generation of chat titles and filtering by folder, with UI components (Navigation and Chat Organization) rendering the folder tree and managing folder CRUD operations.

Calculates token usage per message using tiktoken-js library with model-specific encoding, then applies OpenAI's published pricing tiers to compute real-time conversation costs. Integrates with the streaming API layer to update token counts and costs incrementally as responses arrive, storing cumulative usage in message metadata. Supports multiple model pricing (gpt-4, gpt-3.5-turbo, etc.) with separate input/output token rates.

Exports conversations to JSON (full metadata), Markdown (readable text), or PNG image formats, and imports previously exported JSON chats back into the application. Implements format-specific serialization logic that preserves message history, metadata, and configuration in export, while import validates JSON schema and merges conversations into existing chat list. Supports batch export and Google Drive integration for cloud backup.

Allows per-conversation customization of system prompts and model parameters (temperature, max_tokens, top_p, frequency_penalty, presence_penalty) through ConfigInterface objects stored in Zustand state. Each chat maintains its own configuration, enabling A/B testing of different prompts and parameters. Configuration UI (Configuration Interfaces component) provides sliders and text inputs for parameter tuning, with sensible defaults for each model.

Enables in-place editing of message content, reordering messages within a conversation, and reassigning message roles (user/assistant/system) through UI controls. Implements optimistic updates in Zustand store, immediately reflecting changes in the UI while persisting to localStorage. Supports inserting new messages at arbitrary positions and deleting messages, with automatic conversation regeneration from edited point.

Implements i18n using a localization structure (Localization Structure component) that maps UI strings to language-specific translations stored in JSON files. Supports dynamic locale switching at runtime without page reload, with automatic persistence of language preference to localStorage. Covers API model names, pricing tiers, and UI labels across supported languages (English, Chinese, etc.).

Automatically generates descriptive chat titles by sending the first few messages to the API and extracting a summary, then allows users to manually override with custom titles. Stores titles in ChatInterface metadata and updates them in the folder navigation UI. Regeneration is optional and can be triggered manually for existing conversations.

Allows configuration of custom proxy endpoints to route API requests through alternative servers, enabling access to OpenAI API from regions with restrictions. Proxy URL is stored in configuration and used for all API calls instead of direct api.openai.com endpoint. Supports both OpenAI and Azure proxy endpoints with transparent request/response forwarding.