Dashboard › opencode-lore › Distillation
Distillation
ID: ef0f375a-e525-4b14-b68b-beefc4bad70e
Generation: 0
Tokens: 111
R_compression: 1.054
C_norm: 0.000
Archived: No
Created: 2026-09-17 01:33:16
Source IDs:
["lore_tm_v1_v2OAzCKivdvaOzUa80Lz_08fKJxRnBw6iqv9sZ-evbQ","lore_tm_v1_oMufzmM38jAo7bCyVd4LclkPn2GDsv7AgHutOO0EuEA"]
Observations
Date: Sep 16, 2026
- 🔴 (23:51) User stated embedding execution switched to native ONNX Runtime because it scales with CPU cores.
- 🔴 (23:52) User stated the cached-model purge path is never called for vendored binaries; it applies only when the model resolves to
<cacheDir>/<modelId>/.
- 🔴 (23:52) User requires token truncation to never exceed the configured maximum; texts shorter than the conservative
~1 char/token threshold cannot exceed the limit.