GitLab

Enhance GPU detection logic to support WSL2 environments

This commit updates the GPU detection logic in the setup script to accommodate environments where NVIDIA GPUs might not be directly listed by `lspci`, such as in Windows Subsystem for Linux 2 (WSL2). The revised logic now attempts to identify NVIDIA GPUs using both `lspci` and `nvidia-smi`, ensuring broader compatibility and recognizing GPUs that are accessible in WSL2. This change allows the script to correctly detect NVIDIA GPUs in a wider range of setups, improving user experience and script robustness in diverse environments.

fix: rag message body format

refac: model select option wording

fix: use exec the uvicorn process to handle Docker's SIGTERM correctly

Unhardcode some stuff, flexibilize dotenv requirement

Add default voice setting and apply to chat

feat: Add excel parser for RAG

Signed-off-by: Xiaodong Ye <yeahdongcn@gmail.com>

fix: search

Replace the shell with the uvicorn process, so it becomes PID 1 and receives the signals directly

chore: optimize dockerfile order for caching

add excel file ext

Signed-off-by: lucasew <lucas59356@gmail.com>

Bugfix: Fix toast error popup when front end can't figure out file type.

feat: Add epub support

feat: Add reStructuredText specific parser for RAG

feat: custom interface support

fix: openai

refac: response message