Skip to content
ChangelogRSS

What shipped.

New diagnoses, performance, fixes, and security updates. Subscribe via RSS or follow us on Twitter.
Looking for a public roadmap? denpex.com/roadmap.

July 2026

Shipped2026-07-25

Predictive node health scoring is GA

Per-node 0-100 pre-crash score on the Fleet page and GET /api/fleet/node-health: live ECC/thermal trajectories, Xid/NVLink/PCIe counters and incident recurrence, each deduction named with its evidence. Trend arrows (improving/stable/deteriorating), a recommended action (watch / cordon / drain+RMA), and a failure ETA that only appears when a real monotonic trend supports one, no invented countdowns. Nodes with no telemetry read UNKNOWN, never healthy.

Shipped2026-07-25

Checkpoint rollback planning + runbooks

POST /api/remediation/rollback-plan builds a verify → cordon (when hardware-implicated) → resume-from-verified-checkpoint plan from a stored diagnosis, with the launcher-specific command detected from the logs. The plan never deletes checkpoints and data-corruption classes stay human-gated. Remote execution remains limited to controlled design-partner pilots until the standard on-prem receiver passes production execute, readback, verification, and rollback qualification.

Shipped2026-07-25

NCCL Topology Linter, free at /preflight

The deep pre-flight: GPU→NIC affinity from your real nvidia-smi topo -m matrix, PCIe ACS detection (the silent P2P/GDR killer), NCCL_IB_HCA and NCCL_SOCKET_IFNAME validated against your actual HCAs and interfaces, typo detection across 60+ NCCL variables, and memlock ceilings. Runs 100% in your browser, nothing uploaded. Also server-side for CI at POST /api/preflight/topology (Scale+). Corrected 2026-08-12: this entry originally claimed the lint also shipped in the agent. It did not, until the Python port landed on 2026-08-12.

Shipped2026-07-25

Cross-Run Environment Diff, free at /diff

Paste artifacts from your last good run and the failing one; driver, CUDA, NCCL, PyTorch/JAX, tracked packages, and NCCL_* env are extracted, diffed, and ranked by correlation with the observed failure class, with the weighting shown so the ranking is auditable. Team+ gets the same engine at POST /api/diagnose/env-diff.

Shipped2026-07-25

Cost optimization advice in every diagnosis

Diagnoses now carry a costAdvisor block: right-sizing backed by measured memory headroom (e.g. your job fits L40S at 60% lower $/GPU-hour), spot↔on-demand reliability trades when a job is crash-looping on spot, and utilization advice, every figure with its basis stated, and a 'measure this first' answer when the evidence isn't there.

Shipped2026-07-25

GPU Reliability Index, public dataset

The reliability data procurement never gets, at /gpu-reliability: cited field studies (Meta's 419-interruption Llama-3 run on 16,384 H100s; the NCSA Delta A100/H100 resilience study; ORNL's Summit memory-corruption work) alongside the documented failure surface per GPU family, interconnect and provider stack computed from the 559-class public encyclopedia. Machine-readable at GET /api/reliability-index. Deliberately NOT included: a Denpex-fleet MTBF table, that lights up when 90 days of real cross-tenant history exists, not before.

Shipped2026-07-25

On-call shift handoff reports

GET /api/reports/handoff (and /dashboard/handoff): what failed, what was diagnosed, what auto-remediation did, what the incoming shift inherits, each open incident with its NEXT fix given how far escalation has gone, plus a watchlist (stale running jobs, remediations that need a human). Copy-as-Markdown pastes clean into Slack or a ticket.

Shipped2026-07-25

Multi-cloud single pane on the Fleet page

Every monitored host across AWS, GCP, Azure, Lambda, CoreWeave, RunPod, OCI and on-prem grouped in one view via GET /api/fleet/multicloud, with per-provider GPU totals, degraded counts, and Denpex-identified open issues. Attribution order: the agent's explicit --provider tag, then hostname conventions; unknown stays its own bucket instead of being folded into on-prem.

Shipped2026-07-25

CI/CD gates: GitHub Action + GitLab CI template

Gate merges on preflight, the deep topology lint (preflight-deep), version known-bads, or checkpoint validation. Docs at /docs/ci, GitLab template at /ci/denpex-preflight.gitlab-ci.yml. No API key; everything runs on your runner and nothing leaves it. Corrected 2026-08-12: the GitHub example referenced a published Action repository that was never created, and preflight-deep was not implemented in the agent. Both are fixed; the example now calls the single-file agent directly, and a parity test keeps the agent lint identical to the web engine.

Fixed2026-07-25

Encyclopedia hardening: the 404 wall can't come back

Entry pages now render on demand from static asset copies when the prerender cache misses (dynamicParams=true + a dual-source loader), so the failure mode that once 404'd every /failures/<slug> page now degrades to an edge render instead. Also fixed: every entry title was rendering '… Denpex | Denpex' in search results (double brand), meta descriptions are now answer-first and length-capped, and the live smoke probes cover 11 entry pages across categories including the one dotted slug.

Shipped2026-07-01

16,384-GPU fleets: mass-crash coalescing

Denpex now supports fleets up to 16,384 GPUs end-to-end. The new mass-crash coalescing engine handles the worst day at that scale: when one fabric event (a bad IB switch, a PDU trip) crashes ~2,000 nodes at once, all their crash reports are grouped by cluster and failure signature, ONE diagnosis runs, and every affected job links to it. One alert with the fix instead of thousands of pages, zero dropped diagnoses, and the whole event bills as a single diagnosis. Crash ingest also moved to a dedicated budget that counts distinct root causes, never raw crash reports. Fleet heartbeats already shard per-tenant with headroom at 2,000 nodes, and cascade analysis is tested to 65,536 ranks. Data Center plan copy updated to match.

June 2026

Security2026-06-25

Diagnosis history auto-purged at 90 days

The daily retention cron now deletes diagnoses, follow-up chat, user feedback, and async-diagnosis rows older than 90 days; login_attempts and stripe_events older than 30 days. Combined with the existing crash_logs (30d), alert_log (90d), rate_limit_hits (2d), and consent_receipts (3y) purges, this makes §5 of the privacy policy 100% code-true. Previously, the policy said 90 days for signatures but only crash_logs and alert_log were actually auto-deleted. Diagnosis history was kept forever. No customer action required.

Security2026-06-25

Privacy policy: retention values now match the code

Rewrote §4 and §5 to reflect the actual retention cron (raw crash logs: 30 days, diagnosis signatures: 90 days, alert log: 90 days, rate-limit hits: 2 days, consent receipts: 3 years, audit log: 2 years, billing: 7 years). Account deletion is immediate hard-delete, not 30-day grace. Tightened §2 to say pattern-matching improvements use aggregate signatures, never raw logs. No code changes. These are policy corrections to match the implementation that already runs in production.

Shipped2026-06-20

Diagnosis engine: ranked matches, evidence trail, and an accuracy benchmark

The encyclopedia matcher moved from naive token-overlap to IDF-weighted retrieval with error-signature anchoring (Xid codes, CUDA error numbers, exception classes). It returns ranked candidates with the evidence that fired the match and a multi-step fix. The public benchmark now counts unique GitHub incidents, discloses mined-label noise, and is guarded against duplicate IDs and result drift. Added dedicated CUDA device-side-assert and straggler / slow-rank classes.

Shipped2026-06-15

Trust Center launched

SOC 2 status, sub-processor list, data flow, shared-responsibility matrix, and a one-click request form for DPA, BAA, SOC 2, pen-test, CAIQ, and MSA documents. /security/trust-center.

Shipped2026-06-15

Stripe billing fulfillment is live

Paid checkouts now upgrade the user's plan in the database via webhook. Idempotent on event id. /api/billing/webhook.

Shipped2026-06-15

Cookie consent banner (GDPR)

Accept / reject / customize; analytics and marketing toggles; persisted to localStorage and broadcast as a `denpex:consent` event.

Security2026-06-15

HSTS, CSP, X-Frame-Options, Referrer-Policy

Site-wide security headers via Cloudflare _headers. CSP starts in report-friendly mode and tightens over time.

Shipped2026-06-14

Architecture brief request flow

A 6-page PDF + a follow-up from sales, with an NDA option. /architecture-brief.

Fixed2026-06-12

Dashboard: brief 500 spike during Stripe webhook burst

Resolved in 14 minutes. Idempotency tightened on the subscription.updated path.

Shipped2026-06-10

GitHub OAuth login

Continue with GitHub on the login page. Most ML platform teams run on GitHub Enterprise, so this was overdue.

Shipped2026-06-09

Failure Encyclopedia: first entries live

The first failure-class entries have root cause, symptoms, fix, and prevention. New entries ship weekly (312 classes as of this entry; the encyclopedia has since grown past 360).

Shipped2026-06-05

Annual pricing toggle

Save 2 months on annual plans. Available on /pricing. Team, Scale, and Data Center.

Security2026-06-01

Sub-processor list published

Live list on the Trust Center. 30 days' notice of changes via email and the changelog RSS.

May 2026

Fixed2026-05-28

Console: diagnosis latency above SLO for ~22 minutes

Resolved in 22 minutes. Upstream rate limit on the GitHub tier triggered premature escalation; tier escalation is now rate-limit-aware.

Shipped2026-05-20

Customer responsibility matrix

Plain-English breakdown of who does what, called out from the Trust Center.

Shipped2026-05-15

Data flow diagram

End-to-end: agent → local masking → TLS 1.3 → engine → artifact store → alert fan-out. On the Trust Center.

Fixed2026-05-09

Auth: OAuth callback returned 502 for Google sign-ins

Resolved in 9 minutes. Cloudflare worker pool warm-up. No customer action required.

Shipped2026-05-01

In-VPC agent (Scale and Data Center)

Logs never leave your cluster. Configurable PII/PHI masking on the agent. Available to all Scale and Data Center customers.