Implementation specification for a future coding agent: vendored ONNX
runtime module, config section, news_engine call site and scoring rules
(gold only, never into USD), archive provenance, tests, ops checklist for
both the bot and the dashboard daemon. Owner decisions recorded: primary
mode with no DeepSeek for headlines, model files outside the repo.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
TF-IDF baseline (news_sentiment.py, models/xauusd_sentiment.joblib) and the
two-head DeBERTa-v3-small transformer (transformer/, news_transformer.py)
with its Google News crawler and DeepSeek labeller (data_collect/), the
golden-set and archive evals (eval/), and the 2026-09-22 and 2026-09-27
results and progress notes.
Transformer weights (model.pt, model.onnx) and the crawl database are not
tracked; rebuild with transformer/train.py and transformer/export_onnx.py.
The DeepSeek label cache in data/ is tracked.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Harwinder Sekhon