Captured data fields

Reference for every metric a capture carries. Useful when looking at a dashboard row and wondering "what does that column actually mean?", or when integrating with the raw payload via webhook.

The schema is a flat-ish object delivered as JSON. Field names use camelCase on the wire; the dashboard's session-detail page labels them with their human equivalents.

Identity

Field	Type	Description
`id`	uuid	Server-assigned at insert. Stable; safe to deep-link.
`eventName`	string	The KIND of thing happening (e.g. `Order.Complete`, `Window.ChatPresenter`). Set at the call site.
`eventUniqueId`	string \\| null	Per-instance dedup label — the SUBJECT of this specific event (e.g. `"Tutorial Order"`). Null when the call site didn't supply one. See event semantics.
`sessionName`	string	The test-session label that wraps this capture. Auto-tagged `{ProductName}_{Date}` if you didn't set one in Argus Control.
`environment`	string \\| null	Set in Argus Control's Test Session card. One of Development / Staging / Production / Testing / Custom.
`createdAt`	timestamp	ISO-8601 server-side ingest time. Not the in-game capture wall-clock — for that, use `startTime`.

Frame timing

Field	Unit	Description
`frameCount`	frames	How many frames the capture sampled. Bound by `ArgusConfig.maxFrameBuffer` (default 300).
`totalTime`	seconds	Wall-clock duration of the capture window.
`averageFps`	fps	Mean FPS across `frameCount` samples. Computed from `1000 / avgFrameTime`.
`avgFrameTime`	milliseconds	Mean per-frame delta time. The headline regression metric.
`minFrameTime` / `maxFrameTime`	milliseconds	Best / worst single frame in the capture window.

Per-spike attribution (frameSpikes)

Top-10 frame-time outliers, each annotated with code-attribution context. Spike threshold is 2× rolling average.

Field	Type	Description
`frameIndex`	int	Index into the capture's frame array. Useful for cross-referencing with raw `frameTimes`.
`deltaMs`	milliseconds	The spike magnitude (this frame's deltaTime).
`region`	string \\| null	Active `Argus.BeginRegion("…")` at the spike moment. Null when no region wrapped the code path — the dashboard then falls back to `scene` + `dominant` alone.
`scene`	string \\| null	Active scene name (`SceneManager.GetActiveScene().name`) at the spike moment.
`dominant`	string \\| null	Heaviest subsystem this frame: one of `scripts`, `render`, `gc`, or null when nothing clearly dominated.

See code-region attribution in the Unity guide for how to wrap regions in your code.

Memory

Field	Unit	Description
`totalReservedMemory`	MB	Total memory reserved by Unity (used + headroom).
`totalAllocatedMemory`	MB	Memory currently in use by Unity allocations.
`monoHeapSize`	MB	Mono managed heap size. Monitor for slow growth across captures — that's a leak signal.
`monoUsedSize`	MB	Mono managed memory in active use. The headline managed- memory metric.
`gcTime`	milliseconds	Time spent in GC during the capture.
`frameData.gcAllocPerFrameBytes`	bytes[]	Per-frame GC allocations. Big spikes here are the most common source of dropped frames in Unity. The Claude `latest` skill ranks the top 5 GC frames in its report.

GPU / Rendering

Field	Unit	Description
`drawCalls`	count	Last-frame draw calls. Single most actionable GPU metric.
`batches`	count	Last-frame batches. Batches < draw calls means SRP/URP static + dynamic batching is working.
`triangles`	count	Triangles rendered last frame.
`vertices`	count	Vertices rendered last frame.
`renderingTime`	milliseconds	Time spent in the Unity Rendering profiler category. Sums draw calls, GPU sync, post-processing.

Subsystem breakdown

Per-Unity-profiler-category timing for the last frame of the capture. Compare these against avgFrameTime to see which subsystem ate the budget.

Field	Unit	Profiler category
`scriptsTime`	ms	Scripts (every MonoBehaviour Update + game logic)
`animationTime`	ms	Animation (Animator updates, clip evaluation)
`physicsTime`	ms	Physics (rigidbody simulation, collision detection)
`uiTime`	ms	GUI (Canvas rebuild, UI batching)
`audioTime`	ms	Audio (mixer, voice processing)

Audio & physics state

Field	Type	Description
`audioChannelsUsed`	int	From `AudioSettings.GetConfiguration().numRealVoices`.
`audioPlayingVoices`	int	Active `AudioSource`s with `isPlaying`.
`audioLoadPercentage`	0–100	Playing voices / max-real voices, as percent.
`activeRigidbodies`	int	Sum of `Rigidbody` + `Rigidbody2D` instances in the active scene.
`totalColliders`	int	Sum of `Collider` + `Collider2D` instances in the active scene.

Device fingerprint

Field	Source
`deviceModel`	`SystemInfo.deviceModel` — the most reliable field for grouping captures by device.
`operatingSystem`	`SystemInfo.operatingSystem`
`processorType`	`SystemInfo.processorType`
`systemMemorySize`	`SystemInfo.systemMemorySize` in MB
`graphicsDeviceName`	`SystemInfo.graphicsDeviceName`
`graphicsMemorySize`	`SystemInfo.graphicsMemorySize` in MB
`unityVersion`	`Application.unityVersion`
`appVersion`	`Application.version` from PlayerSettings → Other Settings → Version. Lets you correlate regressions to a specific build.
`targetFrameRate`	`Application.targetFrameRate`

Per-frame arrays (frameData)

Held under the frameData jsonb column for cases when you need frame-level detail beyond the summary scalars. The dashboard shows these in a collapsible JSON viewer on the session-detail page.

Field	Type	Description
`frameTimes`	float[]	Per-frame deltaTime in milliseconds.
`frameTimesCPU`	float[]	Per-frame main-thread time in ms.
`frameTimesGPU`	float[]	Per-frame render-thread time in ms.
`topSpikes`	SpikeRecord[]	Hoisted from the raw payload to top-level for easy access. Same shape as the per-spike attribution above.
`raw`	object	The original Unity payload, kept for debugging. JSONB compresses well so the storage cost is small.

Old captures & missing fields

Captures from before any given field landed render as null / — on the dashboard. Specifically:

appVersion, eventUniqueId, the newSpikeRecord shape, regionsObserved — require Unity tool from May 2026 or later. Older captures show legacy frameSpikes (bare float array) in the Top spikes panel.
The dashboard tolerates both shapes — no errors, just less attribution context for old data. Re-export the Unity tool and new captures populate everything.