Previously Run/Pause/Stop only drove the pulse animation, so on the Data Flow
tab nothing was generated (the streamer is heartbeat-gated and only the Live
dashboard was sending heartbeats). Now:
- New POST /api/federated/live/heartbeat keep-alive endpoint.
- Run seeds an immediate burst into the sources and keeps the generator alive
(5s heartbeat) while the tab is open; Pause/Stop halts generation.
Data Flow tab:
- Prominent "Generate data" button (500 / 2K / 10K) that inserts a fresh
burst of business rows into all source DBs on demand via a new
POST /api/federated/generate (fresh connections, safe alongside the
background streamer); result toast shows what was inserted, CDC streams it.
- "Scripts" button + a "View generation scripts" action on the Data Generator
node open a modal listing every generator script with full source, served by
GET /api/dataflow/scripts. Sources are the real files: the live streaming
generator (sliced live out of trino_federated.py) and the Airflow per-source
DAGs + Faker scripts (mounted read-only from infra/airflow into the API).
Knowledge Chat:
- New "Vector DB" explorer modal: shows the ChromaDB chunking config
(RecursiveCharacterTextSplitter 800/120, all-MiniLM-L6-v2, 384-dim, HNSW),
collections & documents, and the actual stored chunks with text, metadata and
an embedding preview (bars + values) so you can see exactly how files are
split and written as vectors.
Refactor: generator row-builders shared by the streamer and the on-demand burst.
Live dashboard now feels truly real-time:
- Background generator streams randomly-sized bursts of real rows into
PostgreSQL, MySQL, MongoDB & Cassandra every ~4s (CDC picks them up).
Throughput rises and falls; counters move in lock-step (base snapshot +
generated). Runs only while the Live tab is polling (heartbeat-gated) so
source tables do not grow unbounded; on/off toggle exposed in the UI.
- New /api/federated/live/generator toggle; /live returns per-tick activity
(last burst sizes, orders by region/status, event feed).
- LiveDashboard: live-activity panel, orders-per-tick sparkline, event
stream feed, burst-by-region/status charts, generator status + control.
Data Flow graph now explains how data reaches the assistant:
- Added ChromaDB -> RAG (LangChain) -> vLLM Gateway -> Knowledge Chat lane,
with Trino / OpenMetadata / curated-masked feeding LLM context. Live model
& embed metrics pulled from the RAG /config. New node/edge kinds + legend.
- New "Live" tab: auto-polls /api/federated/live every 2.5s with
animated counters, ingestion throughput sparkline, per-source
write-rate bars, a region scorecard matrix (heat-shaded) and live
business breakdown charts (region/channel/status/customers/
telemetry/supply)
- Backend /api/federated/live: instant source estimates (Postgres),
monotonic max(event_id) for MySQL and Mongo estimated count for
immediate movement, Cassandra from cached matrix; business aggs
cached over the small Hadoop lake tables (short TTL)
- Fix embedded Trino panels being squeezed with internal scrollbars
by making panels/grids shrink-0 so the page scrolls instead
The marquee panel only joined 3 region-keyed sources. Add a
"one SQL across every database" reach matrix that fans a single
Trino query out to PostgreSQL, MySQL, MongoDB, Cassandra and the
Hadoop/Iceberg lake in one UNION ALL (telemetry has no region, so
a per-source summary is used instead of a misleading join).
- New MATRIX_SQL + concurrent execution alongside the region
scorecard so total latency stays ~ the slower query
- Federated tab shows the 5-source matrix (records + headline
metric per engine) above the relabelled 3-source region scorecard
Trino Federation tab (3 sub-views):
- Federated: catalog landscape + a single cross-source SQL that joins
PostgreSQL + MySQL + MongoDB (region scorecard) — the federation proof,
computed in the background and cached (large full scans take ~2 min).
- Hadoop Lake: all federated business data materialized as external Iceberg
tables on HDFS (iceberg.hadoop.*_ext, ~120k rows each) with live, fast
business analytics (revenue by region/channel, top customers, HR by
department, supply by type, telemetry averages). Includes a one-click
"rebuild external tables" job.
- Data Dictionary: every business table + column with masked / visible PII
badges and categories.
Backend trino_federated.py: /catalogs, /marquee(+refresh), /lake,
/materialize(+status), /dictionary. Name-based PII detection flags raw PII
in derived/lake tables as visible vs physically-masked curated layer.
LLM context: platform_context now emits a full BUSINESS DATA CATALOG section
(tables, columns, types, source row counts, federated scorecard) with exact
per-column masked/visible status, so the assistant knows the data in detail
and what is masked vs not.
mo