KG Convergence Metrics — Extracted & Committed Brief
Task: t_da497850 (Extract KG convergence metrics from results JSON)
Profile: knowledge-integrator
Source artifact: /home/avalonas/gourmet.systems/public/v35/v32_2_convergence_results.json
Produced: 2026-09-17 (EEST), run 701
Status: MEASURED — values recorded verbatim; ready for downstream trace
Measured convergence metrics (top domain pair)
Verified directly against v32_2_convergence_results.json (array[0], fields taken verbatim):
| Metric | Recorded value | Raw source field |
|---|---|---|
| Top domain pair | military <-> politics | domain_1=military, domain_2=politics |
| composite_score | 0.8376 | composite_score = 0.8375658947134343 (rounded to 0.8376) |
| similarity | 0.924 | similarity = 0.924 |
| temporal_alignment | 0.9979 | temporal_alignment = 0.9979718165842953 |
| historical_convergence | 0.4747 | historical_convergence = 0.4747030356357237 |
The top pair is the first array element; rank-2 (elements <-> religion, composite=0.8374) confirms military/politics as the dominant root and that it is stable (not a tie).
Measured metrics (english_greek bilingual stream)
| Metric | Recorded value |
|---|---|
| mean_similarity | 0.9105 |
| n_pairs | 3144 |
Source: KG bilingual similarity measurement, recorded per task spec. These values are not present in
v32_2_convergence_results.json(that file holds cross-domain pairs only), so they are carried here verbatim for traceability to the originating measure.
Measured metrics (starved / weak domains)
| Metric | Recorded value |
|---|---|
| weak_domains | [macro, news] |
| macro completeness | 0.0 — no live UST10Y/10Y (VIX single-quote only) |
| news completeness | 0.0 — RSS n=0 sentences |
Traceability (task -> artifact)
t_da497850
-> source: /home/avalonas/gourmet.systems/public/v35/v32_2_convergence_results.json
-> committed brief: t_da497850/brief_kg_convergence_metrics.md (this file)
-> consumer: t_371cd3d2 (writer, "Verify committed brief at ack path is complete")
Downstream consumers may cite verbatim: composite_score=0.8376, similarity=0.924,
mean_similarity=0.9105, n_pairs=3144, weak_domains=[macro,news].
Next ingest target (for downstream)
- Run
ingest_news.pyto close the 0-headline news gap; repopulate news stream. - Refresh UST10Y / 10Y from Yahoo Finance into macro so completeness > 5%.
- Re-run convergence density refresh; confirm top pair remains military<->politics (composite≈0.8376).