Wispr Flow: Dictation That Works EverywhereStop typing. Start speaking. 4x faster.
THURSDAY AI
Your Personal Intelligence
9 followers
Your Personal Intelligence
9 followers
Transform your workspace into a premium, next-generation AI operating layer inspired by Jarvis. Command your entire desktop with natural speech recognition, persistent vector DB memory docks, local sqlite cache databases, hyper-automation macros, and modular glassmorphic HUD telemetry widgets.
Free Options
Launch Team / Built With
How does the local SQLite cache stay in sync across multiple machines if I switch between my desktop and laptop throughout the day?
@dugenciogl23408 The local SQLite database acts as a fast offline cache, not as the source of truth. The canonical data lives in the cloud (for example PostgreSQL with Supabase/Firebase or another backend). Every change is written locally first for instant responsiveness and then synchronized to the server. When another device signs in, it downloads only the changes since its last sync and updates its local SQLite cache. Each record includes metadata such as timestamps or version numbers to detect conflicts, and conflicts are resolved using a defined strategy like last-write-wins or CRDTs depending on the data type. This allows seamless switching between desktop and laptop while still supporting offline usage.
How does the speech recognition hold up in noisy environments, and is the vector DB memory stored locally or synced to the cloud somewhere?
@onurzwag Speech recognition: Thursday AI is designed with a multi-stage speech pipeline rather than relying on raw transcription alone. Incoming audio first passes through voice activity detection (VAD) to ignore silence, followed by noise suppression and echo cancellation to reduce background sounds. The cleaned audio is then sent to the speech recognition engine. In normal office or home environments, recognition remains highly accurate. In noisier places like cafés or traffic, accuracy naturally drops, but using beamforming (when multiple microphones are available), confidence scoring, and optional wake-word verification helps maintain reliable performance.
Vector database memory: The vector database is stored locally by default to prioritize privacy, low latency, and offline functionality. For users who work across multiple devices, Thursday AI can optionally encrypt and synchronize vector embeddings through the cloud. The local vector database acts as a cache, while the cloud maintains the canonical copy. During synchronization, only updated embeddings and metadata are transferred, and conflict resolution is handled using versioning and timestamps. This provides seamless memory continuity across devices while allowing privacy-conscious users to keep all memory completely local if they prefer.
This nails the Jarvis fantasy better than anything I've tried, the voice command latency feels solid and the modular HUD widgets actually look clean instead of cluttered.
@diyard3qn Thank You for your feedback you can join us on instagram "@thursday.ai"