Dashboard › opencode-lore › Distillation
593bece0-a25f-49c3-b583-46e49ba01736["lore_tm_v1_qT8zYeACY8JS4MNVVyqRSwGJcYTHNCdlcGnqzA_2o0g","lore_tm_v1_81hAN93SOoC4ILfNMk6kD76zQXOjvMLoxxaAsXBvJso"]
Date: Sep 16, 2026
@loreai/gateway@0.40.0: dist/index.cjs 16.7 MB in 2932 ms; dist/index.bun.js 16.2 MB in 1819 ms; dist/embedding-worker.cjs 565.1 KB in 181 ms; dist/embedding-worker.js 564.9 KB in 162 ms; dist/vector-worker.cjs 11.1 KB in 10 ms; dist/vector-worker.js 10.3 KB in 12 ms; source map dist/index.cjs.map was 43.7 MB.4357cfcf-e64a-4b10-89ad-7002b69bbcfb; sourcemap upload was skipped because SENTRY_AUTH_TOKEN was absent.packages/core/test/semantic-lint.test.ts, the local embedding provider failed to initialize after 3 attempts because ONNX model loading ended in Protobuf parsing failed; the affected model was nomic-ai/nomic-embed-text-v1.5/onnx/model_quantized.onnx, and the error originated through packages/core/src/embedding/local.ts:1008 and packages/core/src/embedding/local.ts:735.packages/core/test/semantic-lint.test.ts had 84 tests with 19 failures. Reported failures included: parseDiffResult > bounds hunks from diffs larger than Node's default buffer before judging; checkInvariants (funnel, stubbed LLM) > flags a hunk that the judge says violates a cosine-near invariant, where prompt was expected once but called 0 times at line 1027; repairs invalid prose once, then records an unresolved candidate, where prompt was expected twice but called 0 times at line 1077; and passes a self-contained minimum judge budget that survives an empty models.dev cache, which expected an array length of 1 but received 0.packages/gateway/test/recall-codex-stream.test.ts failed to deliver the expected continuation text "Here is the answer."; instead, the client received response.created followed by response.failed with type: "server_error", code: "server_error", and message "Gateway request failed". The assertion failed at packages/gateway/test/recall-codex-stream.test.ts:293.