ThoughtEcho

What ThoughtEcho is

ThoughtEcho, also called Xinji in Chinese, is an open source inspiration and quote notebook built with Flutter. The README describes it as an elegant, local first, AI powered cross platform app for capturing fleeting thoughts, organizing reading excerpts, and supporting creative work. The local first design means the user's data is stored on their own device using SQLite and MMKV rather than being sent to a mandatory cloud service. Sensitive notes can be hidden and protected with biometric authentication such as Fingerprint or Face ID, and the README stresses that there is no tracking and no forced cloud lock in. The app targets Windows, Android, and iOS, with Web explicitly listed as unsupported. It is distributed through the Microsoft Store for Windows and through GitHub Releases as an APK for Android. The README states full support for English, Simplified Chinese, Japanese, and Korean, with German, Spanish, and French listed as in progress. The overall positioning is a personal tool for people who collect quotes and ideas, rather than a collaborative or team note product. The privacy section is one of the first things the README highlights, which reflects the project's emphasis on user owned data and on device storage as the default location for notes and attachments.

Features and AI architecture

The feature list covers rich text notes, AI assistance, periodic insights, and synchronization. Notes use the Quill rich text engine and can carry multimedia attachments including images, audio, and video, with both plain and rich storage. The AI side is a multi provider architecture built on the OpenAI compatible protocol, and the README names presets for Ollama, OpenAI, DeepSeek, Gemini, Claude, OpenRouter, and SiliconFlow. A Thoughter agent is described as having cross session long term memory that learns the user's writing personality, and the app can produce weekly and monthly reflection insights, annual review reports, and AI generated share cards. Context capture is automatic: the app records geocoding location, weather, and time of day for an entry. Synchronization uses LocalSend for LAN direct transfer with encrypted TLS, plus optional WebDAV cloud backup and restore. The tech stack table lists Provider and GetIt for state management, sqflite for the local database, flutter_quill for rich text, MMKV and flutter_secure_storage for encrypted API keys, and LocalSend plus WebDAV for sync. The README also shows a roadmap with completed items such as the rich text editor, the multi provider AI architecture, and biometric protected hidden notes, alongside longer term ideas like on device offline model inference and OCR. Nothing in the README claims a specific accuracy or quality level for the AI output.

Building and contributing

The README gives a quick start that assumes Flutter 3.29 or newer and Dart 3.5 or newer, verified with flutter doctor. A user clones the repository, runs flutter pub get, and then flutter run to launch the app. AI services are optional and configured in Settings under AI Settings, where the user picks a provider preset and pastes an API credential. The contribution section welcomes issue reports, localization help for languages still in progress, and code via fork and pull request, with a note that code should pass analysis and tests. The project ships English and Simplified Chinese READMEs plus a user manual and web guide, and it credits open source projects and services it builds on, including Flutter, LocalSend, Sentry, the Hitokoto and ZenQuotes quote providers, Open-Meteo for weather, and OpenStreetMap Nominatim for geocoding. With 44 stars in the metadata, ThoughtEcho is an early stage but actively documented project. The README presents a long list of completed roadmap items, which suggests a working application rather than a prototype, while the long term roadmap items remain clearly marked as ideas. The MIT license lets users modify and redistribute the app. The combination of a local first stance, multiple AI provider presets, and a published Microsoft Store build makes the project a concrete, installable notebook rather than a concept description.

Editorial conclusion

The application is licensed under the MIT license and its most recent commit was on 2026-08-26.