In-Car AI That Works Offline | MongoDB Atlas + ObjectBox Demo
Explore the implementation: https://github.com/mongodb-industry-solutions/voice-car-assistant-v2 Subscribe to the MongoDB for Developers YouTube Channel: https://www.youtube.com/@MongoDBDevelopers?sub_confirmation=1 A dashboard warning light can appear where connectivity is limited. This demo shows how an offline-first in-car AI assistant can answer driver questions using vehicle data and the owner’s manual—whether the vehicle is connected or offline. Built with MongoDB and ObjectBox, the architecture keeps essential data and semantic manual search on the vehicle. A local agent routes voice or text requests to tools for manual search, diagnostic-code context, vehicle status, and relevant telemetry. ObjectBox stores data locally and synchronizes it with MongoDB Atlas when connectivity returns. In the cloud, MongoDB Atlas Vector Search supports semantic retrieval, while Atlas Triggers process telemetry snapshots into current vehicle status and long-term history. The demo uses COVESA’s Vehicle Signal Specification (VSS) for hierarchical vehicle signals and Voyage AI’s voyage-4-nano embeddings (1,024 dimensions). ObjectBox HNSW vector search supports local retrieval; MongoDB Atlas Vector Search supports cloud retrieval. Chapters 00:00 The connectivity challenge 00:25 The connected-car scenario 01:02 Meet the offline-first assistant 01:39 Voice, text, and the local agent 02:44 How ObjectBox keeps the assistant available 03:45 On-device semantic search 04:12 Balancing edge and cloud storage 04:43 MongoDB Atlas and vehicle data 06:01 Embeddings and online/offline vector search 07:16 Preserving telemetry history 08:15 Updating shared vehicle knowledge 09:00 Architecture recap 09:59 Explore the implementation #EdgeAI #ConnectedVehicles #MongoDB Visit Mongodb.com → https://mdb.link/MongoDB Read the MongoDB Blog → https://mdb.link/Blog Read the Developer Blog → https://mdb.link/developerblog MongoDB for Developers YouTube Channel → https://www.youtube.com/@MongoDBDevelopers