lib_llama_cpp_server 0.7.5
lib_llama_cpp_server: ^0.7.5 copied to clipboard
Local OpenAI-compatible HTTP server for lib_llama_cpp.
0.7.5 #
- Add asynchronous native ownership on a worker isolate while preserving the synchronous constructor. Close awaits native terminate/join/free.
- Preserve streaming payload/reasoning/tools/finish/usage, reject DONE-only streams and propagate socket disconnects to native cancellation.
- Expose long-context configuration and actual capability reports; return HTTP 400 for invalid parameters/overflow and preserve structured native error codes.
- Fix delayed HTTP body reads and align default Linux/Windows library names.
- Accept native Windows drive paths in the CLI instead of treating drive letters as URI schemes.
Unreleased #
- Documented use through the app-facing
lib_llama_cppfacade.
0.7.0 #
- Align package metadata with the workspace minor release.
0.1.0 #
- Initial local OpenAI-compatible HTTP server package.
- Added persistent
llcs_enginewrapper, CLI entrypoint, HTTP routes, and a small Dart client.