- Add killport.js script for terminating processes on specific ports

- Introduce `supportedModels` in `ClientChatStore` and update model validation logic
- Enhance OpenAI inferencing with local setup adaptations and improved streaming options
- Modify ChatService to handle local and remote model fetching
- Update input menu to dynamically fetch and display supported models
- Add start_inference_server.sh for initiating local inference server
- Upgrade OpenAI SDK to v5.0.1 and adjust dependencies accordingly
This commit is contained in:
geoffsee
2025-05-29 19:28:54 -04:00
committed by Geoff Seemueller
parent c9ee7c7690
commit cc0da17b5f
11 changed files with 204 additions and 23 deletions

View File

@@ -24,6 +24,22 @@
> Note: Subsequent deployments should omit `bun run deploy:secrets`
## Local Inference (Apple Silicon Only)
~~~bash
#####
# install mlx-omni-server (custom homebrew wrapper)
brew tap seemueller-io/tap
brew install seemueller-io/tap/mlx-omni-server
#####
# Run mlx-omni-server
bun run openai:local
####
# Override OPENAI_* variables in .dev.vars
sed -i '' '/^OPENAI_API_KEY=/d' .dev.vars; echo 'OPENAI_API_KEY=not-needed' >> .dev.vars
sed -i '' '/^OPENAI_API_ENDPOINT=/d' .dev.vars; echo 'OPENAI_API_ENDPOINT=http://localhost:10240' >> .dev.vars
### Restart open-gsio server so it uses the new variables
bun run server:dev
~~~
History
---