Aldehir Rojas
llama.cpp shared-code owner working on chat parsing, tool calls and JSON-schema grammars
- Member of the llama-common code-owner team, llama.cpp project1
Aldehir Rojas is a member of the team that owns llama.cpp's shared common code, according to the project's CODEOWNERS file. Rojas's commits to llama.cpp add model-specific chat parsers and tool-call fixes for model families including Qwen, MiniMax, Gemma and LFM2, along with JSON-schema grammar handling, changes to the reasoning-budget sampler, and llama-server checkpoint and batching changes. Rojas also publishes gpt-oss-adapter, an MIT-licensed proxy that sits between client applications and the LM Studio or llama.cpp servers to manage chain-of-thought reasoning for gpt-oss models.124
Work
llama.cppOpen-source code
CODEOWNERS lists Rojas in the ggml-org/llama-common team, which owns common/. Commits authored from June to October 2026 add specialized chat parsers (Qwen3, MiniMax M3, MiniCPM5), fix tool-call grammars, implement an internal representation for JSON schemas, extend the reasoning-budget sampler, refactor the chat API, and change llama-server checkpoint and batch handling. The repository is MIT-licensed.123
gpt-oss-adapter (external site: github.com)Open-source code
Published under Rojas's GitHub account as an MIT-licensed middleware proxy that manages chain-of-thought reasoning for gpt-oss models, with provider support for LM Studio and llama.cpp.4
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.