Discover and explore top open-source AI tools and projects—updated daily.
Speech and vision personal assistant
Top 10.4% on SourcePulse
Lucida is a speech and vision-based intelligent personal assistant designed to integrate various backend services like Automatic Speech Recognition (ASR), Image Matching (IMM), Question Answering (QA), and more. It allows users to interact with these services through a unified interface, enabling flexible customization and extension with new functionalities. The project is targeted at developers and researchers looking to build or enhance multimodal AI assistants.
How It Works
Lucida employs a modular architecture where backend services communicate via Thrift RPC. A central "command center" (CMD) orchestrates requests, determining the necessary services based on user input and routing queries through a defined service graph. This graph, specified in configuration files, dictates the data flow between services, allowing for complex interactions and dependencies. Services can be written in various languages (C++, Java, Python) and integrated by implementing the Thrift interface and configuring the command center.
Quick Start & Requirements
make local
to install dependencies and compile services. make start_all
starts all services, accessible at http://localhost:3000/
.tools/deploy/
for Docker and Kubernetes deployment instructions.LD_LIBRARY_PATH
. Specific service dependencies might vary.Highlighted Details
Maintenance & Community
The project is released under a BSD license, with submodules potentially having their own licensing. Contributions are welcomed, with details provided in the CONTRIBUTING
section.
Licensing & Compatibility
BSD license. Submodules may have different licenses.
Limitations & Caveats
The REST API for the command center is in active development and subject to change. Adding or removing services requires careful modification of Makefiles and configurations to maintain dependency integrity. The project's complexity in managing dependencies and service graphs might pose a challenge for new contributors.
2 years ago
Inactive