rendezqueue / rendezllama / 21254213748

86%

trunk: 87%

push

github

Refactor assistant_cli to use rendezllama classes and reduce verbosity

- Rewrote `eg/assistant_cli/main.cc` to use `GlobalScope`, `Vocabulary`, `Inference`, `ChatTrajectory`, and `ChatDisplay`.
- Suppressed `llama.cpp` logging using a no-op callback.
- Fixed an assertion failure in `ChatTrajectory::maybe_rollforget_within_limit` when no message prefixes are found.
- Updated `eg/assistant_cli/test_assistant_cli.py` to use greedy sampling expectation.
- Configured `assistant_cli` to use deterministic sampling by default.
- Prevented echoing of user input in the CLI output.

57 of 61 new or added lines in 2 files covered. (93.44%)

2124 of 2394 relevant lines covered (88.72%)

524.96 hits per line