Repository Issue Activity (beta)

#4424 design: version PostHog destination event identity for trace-scoped span IDs

#4415 Bare model ids resolve to a neighbouring model's price via findModel's prefix fallback

#4266 [Explore] Cost per agent

#4393 FIX 1 (most effective) — Contrastive cluster naming: name each behaviour against its siblings

#4394 CAUSE 2 — a single build can accept a catch-all child: the dominant-child gate sits at 0.9 and the size floor removes the finer alternatives

#4395 Make Behaviors visualization self-explanatory

#4396 PROBLEM — Behaviour clusters read as near-duplicate groups, and one catch-all group holds most of the traffic

#4397 CAUSE 1 (largest) — clusters are named in isolation, so sibling labels collapse onto project-wide vocabulary

#4398 CAUSE 3 — RULED OUT: the embedding's shared component is not the cause. Do not build a pre-clustering projection.

#4401 STEP 0 — Offline production-vs-treatment experiment on the pilot plus controls, before any fix is written

#4402 FIX 4 (enabler — land before the grouping fixes) — Report largest-leaf share and duplicate-name rate per build

#4297 New models are unpriced until a release refreshes the bundled models.dev data

#4307 Annotation offsets skew on text whose markdown source differs from its rendered value

#4310 Taxonomy: static publish path mutates live cluster rows before commit — fix only if adaptive clustering is discarded

#4325 Project home design

#4371 Stabilize adaptive K-selection flicker on obvious-split projects (udi8q / osp0i) — decide POST-shadow

#4373 latitude-telemetry (Python): exact OpenTelemetry pin silently disables CrewAI tracing

#4174 Facet-on-facet cross-facet analysis

#4218 Improve memory UI

#4249 Remove feature FLAG

#4264 Split cache cost out of cost_input: correct forward, model backward

#4267 [Explore] Cost per behavior / topic — the thing no gateway can do

#4268 [Explore] Wasted spend — errors, retries, abandoned sessions

#4270 [Explore] Cost per completed task — the completion-signal hierarchy

#4271 [Explore] Shape curves — cost vs turn count per agent