OpenPanel is an open-source web and product analytics platform, an open-source alternative to Mixpanel with optional self-hosting.
Commits on main
ralph(M36-001): Stop computing an answer nobody receives: explicit timeout, abort on client disconnect
1218a59 ralph committed 2 months ago
ralph(M35-001): Anchor the mcp/chart fixtures to a fixed date, removing the midnight flake at source
ca92456 ralph committed 2 months ago
ralph(M34-002): Finish Group C fix 8: the REST retention series and page_titles bounded to range
9c4c087 ralph committed 2 months ago
ralph(M31-004): Group C fix 9: bound or reject very long chart.sankey windows
26a56aa ralph committed 2 months ago
ralph(M31-003): Group C fix 8: give powerUsers, profile.list and retention an explicit window
ac9cf1d ralph committed 2 months ago
ralph(M31-002): Group C fix 7: bound event.pagesTimeseries with LIMIT n BY origin, path
a1e9529 ralph committed 2 months ago
ralph(M31-001): Group C fix 6: LIMIT the chart.getProfiles bucket query
4b7afb8 ralph committed 2 months ago
ralph(M30-003): F2: /live/organization leaks a cross-tenant identifier to a non-member
20228be ralph committed 2 months ago
ralph(M30-001): Stop /metrics costing 70-80 ms of the api's single thread per scrape
80447d7 ralph committed 2 months ago
ralph(M28-002): Give use-property-values.ts a real type - the last place caller/server drift can hide
a0f6c82 ralph committed 2 months ago
ralph(M28-001): M25 Group B: fold the three events scans in page-filtered getMetrics, or prove it is not worth doing
111fd19 ralph committed 2 months ago
ralph(M27-004): Replace three serial await loops with bounded-concurrency Promise.all
0dbb83c ralph committed 2 months ago
ralph(M27-003): Add a bloom_filter skip index on sessions.id so session.byId stops scanning 6.2M rows
2b34180 ralph committed 2 months ago
ralph(M27-002): Re-spell the event-list ORDER BY so it matches the table's own sort key
37e47f1 ralph committed 2 months ago
ralph(M27-001): arrayCompact: turn chart.sankey and overview.userJourney from FAILS into ordinary queries
cd2acc7 ralph committed 2 months ago
ralph(M26-001): Reference-count the shared Redis channel subscription so one socket closing cannot silence every other open socket
f814588 ralph committed 2 months ago
ralph(M21-001): Delete the offset watermark duplicate detector and mark duplicates with an event-keyed Redis key instead - mark only, never drop
cae3829 ralph committed 2 months ago
ralph(M20-001): Dead-letter failed events to a capped Redis list and drop them unconditionally, replacing a Kafka DLQ produce that has never once succeeded
99449f1 ralph committed 2 months ago
ralph(M18-007): Stop the redelivery amplification: drop the re-queue that M18-001 made double-counting, and add backoff so a failed flush cannot spin
22f9256 ralph committed 2 months ago
ralph(M18-003): Redis CACHE client must fail fast: /track cannot block for 42 seconds on a dependency the ingest path already treats as optional
1d213a3 ralph committed 2 months ago
ralph(M18-005): Fix M18-001's false-negative flush gate: resolve on a durability WATERMARK, not on whether this caller happened to start the write
714dae9 ralph committed 2 months ago
ralph(M18-001): Graceful shutdown must flush the event buffer's in-process micro-batch to Redis BEFORE committing Kafka offsets - and must not commit if that flush fails
c9026a6 ralph committed 2 months ago
ralph(M16-005): Turn batching ON by default per ACCEPTED ADR-023: BATCH_SIZE=25, BATCH_LINGER_MS=5, plus the two follow-ups M16-002 could not reach
e004e19 ralph committed 2 months ago
ralph(M16-004): Prove a FRESH install migrates cleanly: 154 Prisma migrations and 27 ClickHouse code-migrations against brand-new databases, and settle the unconfirmed 3-init-ch failure
a5d6f4f ralph committed 2 months ago
ralph(M16-003): docs/SCALING.md: what one box actually sustains, in events/second AND events/month, with the box spec and the peak-vs-average arithmetic spelled out
8774db6 ralph committed 2 months ago
ralph(M16-002): Make the Kafka producer's throughput knobs configurable and add batching behind a default-off flag, so options 2 and 4 can be MEASURED without changing today's behaviour
4b2621c ralph committed 2 months ago
ralph(M16-001): Saturation benchmark harness: a ramp driver, queue/buffer sampling, and a REAL-secret client as the primary configuration
6a60778 ralph committed 2 months ago
ralph(M15-206): CRASH BUG: a NaN duration in http.metrics.ts kills the API process - observeHttpRequest must not hand prom-client a non-finite number
8175c3f ralph committed 2 months ago