
 RUN  v4.1.2 /Users/michelkerkmeester/MEGA/Development/Code_Environment/Public/.opencode/skills/system-spec-kit

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[hf-local] Loading onnx-community/embeddinggemma-300m-ONNX (dtype=q8, first load may take 15-30s)...

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[hf-local] Attempting device: mps

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[hf-local] MPS unavailable (Unsupported device: "mps". Should be one of: cpu.), using CPU

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[hf-local] Model loaded in 689ms (device: cpu)

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[node-llama-cpp] ggml_metal_library_init_from_source: error compiling source

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[node-llama-cpp] ggml_metal_device_init: - the tensor API is not supported in this environment - disabling

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[node-llama-cpp] load: control-looking token:    212 '</s>' was not control-type; this is probably a bug in the model. its type will be overridden

stderr | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
[node-llama-cpp] load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect

stdout | mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
{"mean_cosine":0.9677582325103543,"min_cosine":0.9515004519950576,"samples":50}

 ❯ mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts (1 test | 1 failed) 5645ms
     × matches hf-local on the 50-chunk representative panel 5645ms

⎯⎯⎯⎯⎯⎯⎯ Failed Tests 1 ⎯⎯⎯⎯⎯⎯⎯

 FAIL  mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts > llama-cpp vs hf-local embedding parity > matches hf-local on the 50-chunk representative panel
AssertionError: expected 0.9677582325103543 to be greater than or equal to 0.995
 ❯ mcp_server/tests/embeddings-llama-cpp-parity.vitest.ts:112:18
    110|     }));
    111|
    112|     expect(mean).toBeGreaterThanOrEqual(0.995);
       |                  ^
    113|     expect(min).toBeGreaterThanOrEqual(0.99);
    114|   }, 600000);

⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯[1/1]⎯


 Test Files  1 failed (1)
      Tests  1 failed (1)
   Start at  11:54:29
   Duration  5.76s (transform 25ms, setup 9ms, import 21ms, tests 5.65s, environment 0ms)

