Opportunity: Kibana/Loki+Grafana as a third Phase 3c dashboard option
Projects/Citibank-Compliance-DQ-Audit-Automation's Phase 3c ("SLI/SLA/SLA layer," line 136) frames the dashboard implementation as a binary choice: "Streamlit (fast, less 'real') vs. FastAPI+React (slower, more portfolio-credible)" (line 177), with no third alternative on the table. The project needs a dashboard for SLI/SLO/SLA trends, a queryable result store, and drill-down into failing records over Phase 1/2's accumulated DQ run history (lines 123, 155–170) — a shape that neither option is purpose-built for: Streamlit treats it as a scripting problem, FastAPI+React means hand-building the entire visualization layer.
- Tools/Kibana.md (unresolved) — log/event visualization purpose-built for exactly this kind of drill-down ("why did rule X fail on date Y") over a queryable result store; pairs with Elasticsearch for full-text search across rule descriptions/failure messages.
- Tools/Elasticsearch.md (unresolved) — the queryable store side of the same pair; well suited if the DQ run history needs full-text search, not just label filtering.
- Tools/Loki.md (unresolved) — lighter-weight, label-only log indexing (domain/region/rule/status) — a leaner fit if full-text search isn't needed.
- Tools/Grafana.md (unresolved) — pre-built SLO burn-rate/trend dashboarding, pairs with Loki; also the natural home for the Prometheus+Sloth SLO mechanism already flagged in the 2026-08-22 opportunity pass (Finding 4) — meaning this closes the loop between "how do we compute the SLO" and "how do we display it" with one stack instead of two.
All four Tools notes were created 2026-08-27, after the last Citibank opportunity pass (2026-08-22), which covered the SLI/SLO computation mechanism but explicitly left the dashboard UI choice untouched — this is a genuine gap-fill, not a restatement.
Next step: add a third row to the Phase 3c decision table (line 168) — Kibana+Elasticsearch if full-text search across rule descriptions matters, Loki+Grafana if label-filtered queries (domain/region/rule/status) are enough — alongside a note that Grafana also absorbs the Prometheus+Sloth SLO burn-rate work already flagged as unresolved, making it the option that unifies both open Phase 3c questions (computation + display) into one stack. This is ready now — no missing research, just a table update.