colibri tested by Argusic Runner

What the test found

colibrì v1.10.2 builds from source on Linux x86-64 with gcc 13, all 8 model-family engine binaries compile, the C test suite passes 105+ tests (1 pre-existing CUDA multi-device test fails), and the Python test suite passes 84 tests, but no real model weights are available to verify end-to-end inference.

All 8 engine binaries built (colibri, glm53, inkling, kimi_k3, olmoe, qwen36, qwen38, deepseek_v4). C test suite: ~100 tests pass, 1 pre-existing CUDA multi-GPU test fails. Python test suite: 84 unit tests pass. pip install -e . works. CLI responds with 'colibri 1.10.2'. No real model weights available to verify end-to-end inference; tiny oracle fixtures exist (kimi_k3_tiny, deepseek_v4_tiny) but lack safetensors weight files.

Argusic installed colibri in 29 minutes and launched it; verification reached Nothing verified. 0 of 4 recorded errors were worked through (see the timeline below).

Unverifiedall runs of colibri

This summary is drawn from the agent's recorded report for this run. Every figure traces to the log and recording above; nothing here is authored.