Data Dictionary

Telemetry Serving Freshness

Freshness evidence for an Ottto result after it is prepared for display.

When available: Available when Ottto can safely resolve the current data result

experimental

Telemetry Serving Disclosure

Read-path disclosure distinguishing projected, fallback, and legacy responses.

When available: May be unavailable until the response is stamped at the serving boundary

experimental

Session Detail Results Header

SESSION Results facts for the session detail header, with typed field absence reasons.

When available: null while the session detail surface reads the previous pipeline

experimental

Session Results Detail Locator

Opaque locator for canonical Results detail navigation; never a retained session UUID.

When available: Available only when canonical Results detail navigation is admitted for this row

experimental

Session Detail Relationships

Independent bounded SESSION-family evidence with explicit availability and truncation; never family cost or proof that every relation is navigable.

When available: Available only when the independent relationships surface is admitted

experimental

Session Detail Compaction Evidence

Bounded recorded SESSION compaction observations with explicit capture and absence; observed_count is retained evidence, never an exact lifetime compaction total.

When available: Unavailable when the optional section is not served; unavailable capture keeps observed_count and observations unknown, while observed-empty capture proves zero

experimental

Session Row ID

Stable Ottto identifier for opening session details.

When available: Available whenever this metric applies.

stable

Session ID

External telemetry session identifier.

When available: Available after the session is prepared

stable

Session Agent Group ID

Stable identifier for a session in a repeated-job group. This grouping is separate from parent and child session relationships.

When available: Available only when this session is a retained group member root

stable

Session Agent Group Status

Whether the Results Sessions list grouped this row (grouped, not_grouped) or withheld its group fields.

When available: omitted unless the request asked for group fields

stable

Session Group Run Ordinal

This run's position N in "Run N of M": its rank by first activity among its agent group's runs in the rolling 30-day window.

When available: omitted unless the request asked for group fields and the row is grouped

stable

Session Group Member Count

M in "Run N of M": its agent group's runs in the rolling 30-day window.

When available: omitted unless the request asked for group fields and the row is grouped

stable

Session Display Name

Display-safe session title shown in list and detail views.

When available: Available whenever this metric applies.

stable

Session Display Name Source

Source label for the display-safe session title.

When available: Available by default; defaults to a safe fallback label

stable

Session First Seen

Earliest known session timestamp.

When available: May be unavailable until a source supplies timing evidence

stable

Session Last Seen

Latest known session activity timestamp.

When available: May be unavailable until a source supplies timing evidence

stable

Session Tokens Total

Total billable token count excluding reasoning to avoid double counting.

When available: Available whenever this metric applies.

stable

Session Tokens Unattributed Total

Total-only local token evidence that cannot be safely split into input/output/cache dimensions.

When available: 0 when no total-only local evidence exists

stable

Session Tokens Input

Input tokens for the session.

When available: Available whenever this metric applies.

stable

Session Tokens Output

Output tokens for the session.

When available: Available whenever this metric applies.

stable

Session Tokens Cache Read

Cached input tokens read during the session.

When available: Available whenever this metric applies.

stable

Session Tokens Cache Creation

Cached input tokens written during the session.

When available: Available whenever this metric applies.

stable

Session Tokens Cache Creation 5m

Cached input tokens written to the five-minute prompt-cache tier.

When available: Available whenever this metric applies.

stable

Session Tokens Cache Creation 1h

Cached input tokens written to the one-hour prompt-cache tier.

When available: Available whenever this metric applies.

stable

Session Cache Health

Prompt-cache explanation for the session: hit rate, reuse ratio, and the token-weighted hit/miss split, derived from the canonical cache-metrics helper over the session's normalised token totals.

When available: inner ratios are Unavailable when the session has no prompt or cache activity

stable

Session Cache Health Hit Rate

Share of all prompt-side tokens served from cache: cache_read / (input + cache_read + cache_creation).

When available: Unavailable when the session has no prompt tokens

stable

Session Cache Health Reuse Ratio

How well the cache you paid to write is reused: cache_read / (cache_read + cache_creation).

When available: Unavailable when the session has no cache activity at all

stable

Session Cache Health Hit Token Count

Token-weighted cache hits for the session (cache_read tokens).

When available: 0 when no cache reads occurred

stable

Session Cache Health Miss Token Count

Token-weighted cache misses for the session: fresh input plus cache writes (prompt-side tokens not served from cache).

When available: 0 when there is no prompt-side activity

stable

Session Context Fill

Estimated share of the model context window used by the latest prompt-bearing turn.

When available: Unavailable when the session has no prompt-bearing turn or the model's working window cannot be resolved. Requires turn-grain usage_turn_fact rows from OTLP transport; unavailable on local-snapshot-only sessions.

stable

Session Context Fill Reconciliation

Comparison between Ottto's context estimate and a captured Claude Code context measurement.

When available: null for non-Claude sessions or when no Claude Code /context capture exists for the session's machine. The reconstruction (session.context.fill) requires OTLP-collected turn-grain data; both metrics are unavailable on local-snapshot-only sessions.

stable

Session Context Footprint

Session-scoped Claude Code static context footprint from the verified context-footprint data, enriched with scoped usage evidence when attribution facts exist.

When available: Unavailable when the session lacks machine/workspace evidence or no matching context-footprint snapshot exists

stable

Session Context Posture

Content-free local context high-water evidence: first and peak prompt tokens, resolved window and peak percentage, compaction event count, and source activity/collection timestamps. Current prompt tokens stay null until the collector captures a true current watermark.

When available: Unavailable when the session has no keyed local context-posture GOLD evidence

stable

Session Context Curve

Bounded session context curve with prompt-token points, compaction boundaries, and the model-window evidence used for utilization.

When available: Unavailable when the session has no projected context-curve evidence

experimental

Session Long Running

Long-running-session advisory (CC-insights P3 contributing factor): set when the session has WORKED for 8h+ (active_seconds, the decoded Claude Code long-session threshold; wall-clock elapsed span is the fallback for historical sessions with no active signal). A session this long carries its whole history forward, inflating per-turn context and cost. Advisory only — it never changes a token or dollar figure.

When available: Available only when the session crossed the 8h-active threshold

stable

Session Tokens Reasoning Output

Reasoning output tokens, a subset of output tokens.

When available: 0 when the source does not report reasoning tokens

stable

Session Cost Own

Independently admitted own-session COST answer and its availability state.

When available: nullable; only ready own-session answers carry money; never a family total

stable

Session Cost USD

Total session cost in USD.

When available: Available by default; may be 0 when pricing is unresolved

stable

Session Cost USD Availability

Availability of the session's whole cost from admitted writer witnesses: complete, known_subtotal (including zero), unavailable or unclassified. It qualifies cost_usd and never changes the amount.

When available: Unavailable when cost_usd is null (cost_unavailable_reason says why) and on non-Results readers

stable

Session Cost Input USD

Input-token cost for the session.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Cost Output USD

Output-token cost for the session.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Cost Cache Read USD

Cache-read token cost for the session.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Cost Cache Creation USD

Cache-creation token cost for the session.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Cost Cache Creation 5m USD

Five-minute prompt-cache creation cost for the session.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Cost Cache Creation 1h USD

One-hour prompt-cache creation cost for the session.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Cost Kind

Whether the displayed cost is actual or estimated.

When available: Available whenever this metric applies.

stable

Session Cost Source Kind

Internal cost provenance value for the session.

When available: Available whenever this metric applies.

stable

Session Cost Source Label

Human-readable cost source label.

When available: always, using a safe default when unresolved

stable

Session Cost Projection Status

Pricing projection state for the session cost.

When available: Available whenever this metric applies.

stable

Session Cost Unavailable Reason

Exact reason a session's whole cost is unavailable instead of zero.

When available: May be unavailable when the whole session cost is available

stable

Session Cost Materialized At

Timestamp when the selected cost calculation was materialized.

When available: May be unavailable until the pricing calculation completes

stable

Session Cost Is Stale

Whether the cost was projected from a stale pricing catalog snapshot.

When available: Available whenever this metric applies.

stable

Session Cost Pricing Provenance

Sanitized pricing audit snapshot for the selected session cost.

When available: May be unavailable when no pricing projection has been built

stable

Session Request Count

Model turns: model interactions recorded from the agent's local transcripts. Sessions tracked only through live telemetry count telemetry records instead; count_basis names the unit.

When available: Available whenever this metric applies.

stable

Session Count Basis

Which unit the session's activity count uses: local_model_turn (model interactions recorded from local transcripts), telemetry_usage_record (live telemetry usage records), or provider_request (provider-reported requests).

When available: Available whenever this metric applies.

stable

Session Primary Model

Model key with the largest token share in the session.

When available: May be unavailable when no model attribution exists

stable

Session Primary Model Display Name

Display name for the primary model.

When available: May be unavailable when no model attribution exists

stable

Session Primary Reasoning Effort

Observed reasoning-effort tier holding the largest token share in the session. ACTUAL per-turn evidence from verified turn-level usage data; a configured default never stands in for it.

When available: May be unavailable when no turn proved a tier, and withheld when cost_facts_truncated is true, since a partial fact set cannot name a whole-session dominant tier

stable

Session Reasoning Effort Mix

Every observed reasoning-effort tier for the session, dominant first, each with its token count and share. A session is not one tier - Claude Code stamps the applied tier per assistant record and Codex ships it per turn - so the whole mix is carried rather than one lossy scalar. Shares are taken over ALL observed tokens including the effort-unknown remainder, so they sum to <= 1 and the gap is the uncovered portion.

When available: empty when no turn proved a tier, and withheld when cost_facts_truncated is true

stable

Session Source

Canonical AI app or telemetry source for the session.

When available: May be unavailable for legacy rows without source inference

stable

Session Execution Location

Where the observed work executed: locally, in a provider cloud, by remote control, or unknown.

When available: Available whenever this metric applies.

experimental

Session Cloud Entity Kind

Content-free provider cloud entity classification.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Cloud Measurement Basis

Explicit statement that provider cloud work is observed as metadata, not itemized usage.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Cloud Coverage Mask

Compact bit mask recording which supported provider-cloud metadata dimensions were observed.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Cloud Semantic Updated At

Provider-reported time when the cloud entity last changed semantically.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Cloud Lifecycle Status

Normalized lifecycle state for an observed provider-cloud task or session.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Cloud Attempt Count

Observed provider-cloud attempt count when that provider exposes it.

When available: provider-cloud sessions with attempt coverage only

experimental

Session Cloud Environment Kind

Normalized execution-environment class for provider-cloud work.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Cloud Metric Availability

Per-dimension explanation of unavailable or intentionally non-itemized cloud metrics.

When available: provider-cloud sessions only; May be unavailable for local and legacy rows

experimental

Session Machine Applicable

Whether machine attribution is meaningful for this session's execution location.

When available: Available whenever this metric applies.

experimental

Session Originator

Raw session-surface label from the source session_meta (Codex originator), e.g. codex_work_desktop for the unified ChatGPT desktop app's local Work sessions or codex_cli_rs for the CLI. Persisted raw; the source stays codex.

When available: nullable; only present when the collector forwarded a session originator

stable

Session Sub Surface

Derived display sub-surface refining session.source: chatgpt_work for the unified ChatGPT desktop app's local Work sessions (originator codex_work_desktop), else null. Deliberately conservative; the source stays codex.

When available: nullable; only chatgpt_work is derived today

stable

Session Initiator

Session-level human-vs-AI origin classification derived from raw local/live origin signals.

When available: Available by default; defaults to unknown when origin evidence is absent

stable

Session Initiator Label

Display label for the session-level human-vs-AI origin classification.

When available: Available by default; defaults to Unknown when origin evidence is absent

stable

Session Automation Attribution

Positive-only automation origin and neutral workflow facts, each with field-level evidence.

When available: Available by default; empty when no attribution evidence is known

stable

Session Used Workflow Orchestration

Whether the session ran dynamic workflow orchestration (the Workflow tool, e.g. ultracode); detected from the local wf_*.json footprint and/or a tool_name == Workflow turn. Invisible in token usage (ultracode collapses to xhigh effort).

When available: Available by default; defaults to false when no orchestration evidence exists

stable

Session Agent Identity

Where this session sits in a spawned-agent family: its direct parent, its root (the top-level human session), the Claude Code agent type/id, its nesting depth, and the owning Workflow run. Present only for sessions the collector re-keyed off their own subagent transcript.

When available: Available only when the session is itself a spawned subagent

stable

Session Family Status

Whether the spawned-agent family numbers on this response can be trusted: available (exact), partial (the family exceeded the per-request member cap), or unavailable (the family read failed, so the family fields are null rather than zero - 'spawned nothing' and 'could not find out' are different answers).

When available: Available whenever this metric applies.

stable

Session Family Unavailable Reason

Exact reason spawned-agent family totals are withheld from a session row.

When available: May be unavailable when family totals are complete

stable

Session Attribution Status

Whether attribution is known for this Sessions row. unavailable means the page was served without the ATTRIBUTION calculation (the limited SESSION+COST Sessions list, DR-96): attribution chips, agent identity and family are not known yet rather than none, and clients must not link to attribution or family detail from the row.

When available: Available whenever this metric applies.

experimental

Session Detail Unavailable Reason

Whether this Sessions row's id opens the session detail page. Null means id opens GET /sessions/{id}; when set, id is only a row key and clients must not link it, so a list row never leads to a 404 (D-90 section 3). no_detail_row: the detail page is still served by the previous pipeline and has no row for this session yet. detail_not_resolved_in_preview: the dashboard preview does not read detail ids.

When available: May be unavailable when the row's id opens the session detail page

experimental

Session Subagent Count

Number of agents this session started, at every nesting depth. Derived from the transitive parent_session_ref relation, which is the same relation the has_family filter reads, so 0 means nothing names this session as its parent - never that a root_session_ref fact was missing.

When available: Unavailable when family_status is unavailable

stable

Session Family Tokens

Total tokens for this session AND all of its spawned-agent descendants. The session's own total_tokens NEVER includes descendants, so family = own + children, counted exactly once at any depth. Equals the session's own total when it spawned nothing.

When available: Unavailable when family_status is unavailable

stable

Session Family Requests

Model turns for this session and all of its spawned agents, under the same no-double-counting rule as session.family.tokens.

When available: Unavailable when family_status is unavailable

stable

Session Family Cost USD

Selected cost for this session and all of its spawned-agent descendants, under the same no-double-counting rule as session.family.tokens.

When available: Unavailable when family_status is unavailable

stable

Session Family Cost USD Availability

Availability of family_cost_usd (this session plus the sessions below it) on the has_family Results list: complete, known_subtotal (a lower bound, never a complete lower total), unavailable or unclassified. It qualifies family_cost_usd and never changes the amount.

When available: omitted (never null on the wire) on every list except the has_family Results list; there, unavailable when the row's family money was withheld, and omitted on a row whose family money was not read

stable

Session Subagents

The spawned-agent family hanging off this session: descendant-only totals, the family rollup (own + descendants), and the child list with enough per agent to render a row and deep-link to it.

When available: Unavailable when the session spawned no agents

stable

Session Parent

Backlink from a subagent session to the session that spawned it, plus the root of the whole family.

When available: Available only when the session is itself a spawned subagent

stable

Session Machine ID

Best-known computer identifier for the session.

When available: May be unavailable when no machine evidence exists

stable

Session Machine Name

Display-safe computer name for the session.

When available: May be unavailable when no display-safe machine name exists

stable

Session Machine Icon ID

Selected icon identifier for the session computer.

When available: May be unavailable when no icon is configured

stable

Session Machine Icon Url

Renderable icon URL for the session computer.

When available: May be unavailable when no icon is configured

stable

Session Machine Icon Version

Version counter for the session computer icon.

When available: Available whenever this metric applies.

stable

Session Project Label

Display-safe workspace/project label.

When available: May be unavailable when workspace evidence is absent or unsafe

stable

Session Project Label Source

Source marker for the display-safe project label.

When available: Available whenever this metric applies.

stable

Session Repository Hash

Organization-scoped opaque repository identity for grouping sessions.

When available: May be unavailable when no repository can be resolved locally

stable

Session Repository Label

Display-safe repository name derived locally without a path or remote URL.

When available: May be unavailable when no safe repository label can be resolved

stable

Session Repository Label Source

Local derivation source for the display-safe repository label.

When available: May be unavailable when repository_label is absent

stable

Session Repository Identity Source

Local derivation source for the opaque repository identity.

When available: May be unavailable when repository_hash is absent

stable

Session Workspace Kind

Local workspace relationship to the repository, such as root or worktree.

When available: May be unavailable when no repository workspace can be resolved

stable

Session Branch Label

Display-safe branch label for the session workspace.

When available: May be unavailable when branch evidence is absent or unsafe

stable

Session Worktree Label

Display-safe worktree label for the session workspace.

When available: May be unavailable when worktree evidence is absent or unsafe

stable

Session Lifecycle Status

Lifecycle state inferred for the session.

When available: Available whenever this metric applies.

stable

Session Source Plan Profile ID

Durable source-plan attribution profile id.

When available: May be unavailable when attribution is unknown

stable

Session Source Plan

Display-safe source-plan summary for subscription/billing attribution.

When available: May be unavailable when no profile is attributed

stable

Session Subscription Parts

Subscriptions charged for the session, most money first (at most four), each with its share of the row's cost, its source-plan label, when the plan was first observed, and whether usage preceded that observation.

When available: empty when no subscription was charged or when the row's COST dimensions are unavailable; shares are May be unavailable when the row's cost is unavailable

experimental

Session Other Subscription Count

Subscriptions charged for the session beyond the four named parts.

When available: Available whenever this metric applies.

experimental

Session Unidentified Usage

Reserved; always false. Unknowns are never shown externally, so usage charged to no identified subscription is not reported.

When available: always false externally; kept for schema compatibility

experimental

Session Unidentified Cost Share

Reserved; always null. Unknowns are never shown externally, so no unidentified cost share is reported.

When available: always null externally; kept for schema compatibility

experimental

Session Pricing Status

Fine-grained pricing coverage status for the row.

When available: Available whenever this metric applies.

stable

Session Usage Mode

Usage mode selected from source-plan and telemetry attribution.

When available: Available whenever this metric applies.

stable

Session Runtime Defaults

Display-safe runtime defaults captured for the session when local status evidence exists.

When available: May be unavailable when no runtime-default evidence was captured for the session

stable

Session Attribution

Display-safe billing/model attribution snapshot.

When available: May be unavailable when no attribution evidence exists

stable

Session Usage Origin

Source path that currently owns the session usage totals.

When available: Available whenever this metric applies.

stable

Session Timeline Origin

Source path that currently owns session timing.

When available: Available whenever this metric applies.

stable

Session Reconciled At

Timestamp of the latest local/live reconciliation pass.

When available: May be unavailable until reconciliation runs

stable

Session Reconciliation Status

Status comparing Local enriched and Live telemetry evidence.

When available: Available whenever this metric applies.

stable

Session Cost Facts Truncated

Whether this session's turn-level usage records were truncated at the per-session fact cap (25,000 OTLP turns). True means the FACT-BACKED numbers (cost breakdown, cost rollups, fact-derived model rollups) cover only the session's first N turns; the session's own totals come from the uncapped prepared session data and stay complete, so the two disagree by design.

When available: Available by default; defaults to false and is true only past the fact cap

stable

Session Tool Usage Truncated

Whether the Local enriched transcript aggregate omitted additional distinct tool names after reaching its per-session bound. Counts for the retained names remain cumulative session-level transcript counts, not turn-grain Live telemetry facts.

When available: Available by default; defaults to false and is true only when the local transcript aggregate exceeds its distinct-tool bound

stable

Session Error Count

Count of errors associated with the session detail.

When available: Available whenever this metric applies.

stable

Session Elapsed Seconds

Elapsed wall-clock seconds for a session: last_seen minus first_seen. This is the headline Duration stat on session detail and spans the whole observed run, including idle gaps between bursts of telemetry.

When available: May be unavailable until both session bounds are known

stable

Session Active Seconds

Active wall-clock seconds for a session: total time spent on requests with each idle gap capped at 5 minutes. The session header's Duration stat shows the full wall-clock span (last_seen minus first_seen) on top and pairs it with this active-only figure beneath, so a long-running session with idle stretches reads as both a long wall time and a shorter active time.

When available: 0 when no activity records exist

stable

Session Model Breakdown[*]

Per-model usage and cost rows for a session.

When available: Available whenever this metric applies.

stable

Session Tool Summaries[*]

Redaction-safe per-tool session summaries with per-element source tags.

When available: empty when neither path has tool evidence

stable

Session Mcp Server Summaries[*]

Redaction-safe per-MCP-server call rollup (tool calls grouped by server).

When available: empty when the session invoked no MCP tools

stable

Session Capability Summaries[*]

Privacy-safe positive Chrome, browser, and Computer Use calls classified locally from the session record.

When available: empty until a compatible collector positively observes a supported capability

stable

Session Activity Summary

Content-free local activity evidence for tool, shell, patch, file, line-change, web, MCP, and subagent-spawn counts, with bounded tool, MCP-tool, and consent-gated skill labels.

When available: null until a compatible collector uploads a keyed activity summary

stable

Session File Summaries[*]

Redaction-safe file/workspace evidence summaries with per-element source tags.

When available: empty when neither path has file evidence

stable

Session Conversation Turns[*]

Redaction-safe conversation turn summaries with per-element source tags.

When available: empty when turn evidence is unavailable

stable

Session Conversation Turn Count

Total number of token-bearing telemetry records for the session. Lets the UI render an honest 'N of M turns' header when conversation_turns is capped and evenly sampled.

When available: Available whenever this metric applies.

stable

Session Workflow Markers[*]

Display-safe workflow behavior markers derived from telemetry metadata.

When available: empty when no known workflow marker is detected

stable

Session Tool Summaries[*] Source Path

Source path attached to each redaction-safe tool summary element.

When available: Available whenever this metric applies.

stable

Session File Summaries[*] Source Path

Source path attached to each redaction-safe file summary element.

When available: Available whenever this metric applies.

stable

Session Conversation Turns[*] Source Path

Source path attached to each redaction-safe turn summary element.

When available: Available whenever this metric applies.

stable

Session Artifacts

VCS artifacts (PR / issue / commit) a session produced; opt-in, local-derived.

When available: Available whenever this metric applies.

stable

Session Model Breakdown[*] Input Cost USD

Per-model input-token cost.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Model Breakdown[*] Output Cost USD

Per-model output-token cost.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Model Breakdown[*] Cache Read Cost USD

Per-model cache-read token cost.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Model Breakdown[*] Cache Creation Cost USD

Per-model cache-creation token cost.

When available: May be unavailable when dimension pricing is unresolved

stable

Session Model Breakdown[*] Cache Efficiency Ratio

Per-model cache reuse ratio: cache_read / (cache_read + cache_creation) over this model's normalised counts, from the canonical cache-metrics helper. 1.0 for read-only (cache_creation == 0) rows.

When available: Unavailable when the model row has no cache activity at all

stable

Session Records[*]

Live telemetry record rows for a session.

When available: empty when Live telemetry is disabled or has no records

stable

Session Records[*] ID

Database id for a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Received At

Server receive timestamp for a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Telemetry Type

OpenTelemetry signal type for a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Raw Record Material Exposed

Invariant proving session record rows omit raw resource and payload material.

When available: Available whenever this metric applies.

stable

Session Records[*] Model

Model identifier attached to a live telemetry record.

When available: May be unavailable when the record does not carry model attribution

stable

Session Records[*] Tokens Input

Input tokens attached to a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Tokens Output

Output tokens attached to a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Tokens Cache Read

Cache-read tokens attached to a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Tokens Cache Creation

Cache-creation tokens attached to a live telemetry record.

When available: Available whenever this metric applies.

stable

Session Records[*] Reasoning Effort

Actual reasoning effort the Codex turn ran at, distinct from the configured default.

When available: nullable; Codex turns that emit a per-turn reasoning signal

stable

Session Records[*] Duration Ms

Wall-clock duration of one model request (a single turn) for agents that report per-turn latency over OTLP — Claude Code and Pi. Session detail averages this across the loaded turns as the 'Response time' stat and surfaces the slowest turn (max) alongside it; time to first token is the responsiveness portion.

When available: nullable; agents that emit per-turn latency over OTLP (Claude Code, Pi)

stable

Session Records[*] Time To First Token Ms

Latency to the first output token for the Codex turn attached to a live telemetry record.

When available: nullable; Codex turns that emit per-turn latency

stable

Session Session Avg Duration Ms

Average wall-clock duration of a Codex agentic turn across the session — a rollout `task_complete` task, i.e. model time plus tool execution and idle, so long-running turns can span minutes. The local collector aggregates it per session because Codex never emits per-turn latency over OTLP. Shown in session detail as the 'Turn duration' stat; the average smooths over the longest turn, and time to first token is the model's responsiveness.

When available: nullable; Codex sessions with completed turns

stable

Session Session Avg Time To First Token Ms

Average latency to the first output token across the session's turns, aggregated by the local collector from the Codex rollout (never emitted over OTLP).

When available: nullable; Codex sessions with completed turns

stable

Session Session Max Duration Ms

Duration of the slowest single Codex agentic turn in the session (the max the average smooths over), aggregated by the local collector from the rollout. Surfaced next to the 'Turn duration' average as the tail signal.

When available: nullable; Codex sessions with completed turns

stable

Session Session Max Time To First Token Ms

Slowest time to the first output token across the session's Codex turns, aggregated by the local collector from the rollout (never emitted over OTLP).

When available: nullable; Codex sessions with completed turns

stable

Cost Period Start

Inclusive start date for a cost breakdown window.

When available: Available whenever this metric applies.

stable

Cost Period End

Inclusive end date for a cost breakdown window.

When available: Available whenever this metric applies.

stable

Cost Range Preset

Calendar preset used for a cost breakdown window.

When available: May be unavailable for custom or rolling day windows

stable

Cost Bucket Granularity

Bucket granularity used for cost trend arrays.

When available: Available whenever this metric applies.

stable

Cost Data Basis

Data projection basis used for the cost breakdown.

When available: Available whenever this metric applies.

stable

Cost Date Basis

Session date field used to assign records to the requested cost window.

When available: Available whenever this metric applies.

stable

Cost Cost Basis

Cost field selected for cost totals and grouped breakdowns.

When available: Available whenever this metric applies.

stable

Cost Source Filter Applied

Whether the cost breakdown was filtered to one or more AI app sources.

When available: Available whenever this metric applies.

stable

Cost Machine Filter Applied

Whether the cost breakdown was filtered to one or more machines.

When available: Available whenever this metric applies.

stable

Cost Source Plan Filter Applied

Whether the cost breakdown was filtered to one or more source-plan profiles.

When available: Available whenever this metric applies.

stable

Cost Selector Filter Applied

Whether the cost breakdown was filtered by selector dimension values.

When available: Available whenever this metric applies.

stable

Cost Model Filter Applied

Whether the cost breakdown was filtered to one or more models.

When available: Available whenever this metric applies.

stable

Cost Cloud

Organization-scope cloud provider billing composed into this window. Present only with include_cloud=true and the integrations.org_scope.read permission; absent from the payload otherwise, so an un-opted-in response is unchanged. Its own gross/credits/net, per-day and per-model rows, tokens, precedence overlap and freshness live inside the block; combined local+cloud figures appear ONLY under cloud.composed and never mutate a session-grain total on this response.

When available: Available whenever this metric applies.

stable

Cost Diagnostics

Display-safe diagnostics for the requested cost breakdown.

When available: Available whenever this metric applies.

stable

Cost Total USD

Total cost over the requested window.

When available: Available whenever this metric applies.

stable

Cost Money Availability

Availability of selected money from admitted writer witnesses: complete, known_subtotal (including zero), unavailable or unclassified. Unavailable and unclassified money is null; this is not a pending-work signal.

When available: Available whenever this metric applies.

stable

Cost Total Session Count

Exact distinct session count for the complete response window; clients must not sum by_source[*].session_count.

When available: Available whenever this metric applies.

stable

Cost Kind

Whether the cost breakdown is actual or estimated.

When available: Available whenever this metric applies.

stable

Cost Source Kind

Internal cost provenance value for the breakdown.

When available: Available whenever this metric applies.

stable

Cost Source Label

Human-readable cost source label for the breakdown.

When available: Available whenever this metric applies.

stable

Cost Projection Status

Pricing projection state for the cost breakdown.

When available: Available whenever this metric applies.

stable

Cost Is Stale

Whether the cost breakdown was projected from stale pricing data.

When available: Available whenever this metric applies.

stable

Cost Stale Reason

Why the cost totals carry a freshness qualifier: an in-flight local-data reconciliation, ordinary gold-rollup lag, or fresh.

When available: Available whenever this metric applies.

stable

Cost Reconciliation Active

Whether a totals-pausing local-data reconciliation (account/machine backfill) is in flight for the cost breakdown.

When available: Available whenever this metric applies.

stable

Cost Pricing Provenance

Sanitized pricing audit snapshot for the breakdown.

When available: May be unavailable when no pricing projection has been built

stable

Cost By Model[*]

Cost and usage totals grouped by model.

When available: Available whenever this metric applies.

stable

Cost By Model Daily[*]

Cost and usage buckets grouped by model and day or hour.

When available: Available whenever this metric applies.

stable

Cost By Day[*]

Daily cost and usage buckets for the requested window.

When available: Available whenever this metric applies.

stable

Cost By Source[*]

Cost and usage totals grouped by AI app/source.

When available: Available whenever this metric applies.

stable

Cost By Selector[*]

Current, previous, and delta cost totals grouped by selector dimension.

When available: Available whenever this metric applies.

stable

Cost By Reasoning Effort[*]

Cost and usage totals grouped by per-turn reasoning-effort tier, per source.

When available: Available whenever this metric applies.

stable

Cost By Source Daily[*]

Daily cost buckets grouped by AI app/source.

When available: Available whenever this metric applies.

stable

Cost By Source Hourly[*]

Hourly cost buckets grouped by AI app/source.

When available: Available whenever this metric applies.

stable

Cost By Machine[*]

Cost and usage totals grouped by computer.

When available: Available whenever this metric applies.

stable

Cost By Machine Daily[*]

Daily cost buckets grouped by computer.

When available: Available whenever this metric applies.

stable

Cost By Machine Hourly[*]

Hourly cost buckets grouped by computer.

When available: Available whenever this metric applies.

stable

Cost By Subscription[*]

Period cost and usage grouped by backend-authored subscription membership; request and session counts are distinct within each row's own scope.

When available: omitted unless complete authored subscription grouping is published

experimental

Cost By Subscription Daily[*]

Daily cost and usage grouped by backend-authored subscription membership; request and session counts are distinct within each row's own scope.

When available: omitted unless complete authored subscription grouping is published

experimental

Cost By Subscription Hourly[*]

Hourly cost and usage grouped by backend-authored subscription membership; request and session counts are distinct within each row's own scope.

When available: omitted unless complete authored subscription grouping is published at hour granularity

experimental

Cost By Source Plan[*]

Cost and usage totals grouped by source-plan profile.

When available: Available whenever this metric applies.

stable

Cost By Source Plan Daily[*]

Daily cost buckets grouped by source-plan profile.

When available: Available whenever this metric applies.

stable

Cost By Source Plan Hourly[*]

Hourly cost buckets grouped by source-plan profile.

When available: Available whenever this metric applies.

stable

Cost Machine Grouping State

Disclosure for the by-machine groupings: unavailable when the served answer cannot reproduce the recorded machine for every recorded profile (DR-122); totals and other groupings stay valid.

When available: omitted unless the machine grouping is withheld

experimental

Cost Billing State

Disclosure for the billing-identity groupings: bound, recovered_not_recorded, or unavailable when exact billing identity is not supported for the served answer (DR-115).

When available: omitted unless the response states its billing grouping

experimental

Cost By Billing Identity[*]

Cost and usage totals grouped by billing identity.

When available: Available whenever this metric applies.

stable

Cost By Billing Identity Daily[*]

Daily cost buckets grouped by billing identity.

When available: Available whenever this metric applies.

stable

Cost By Billing Identity Hourly[*]

Hourly cost buckets grouped by billing identity.

When available: Available whenever this metric applies.

stable

Agent Context Cost Total USD

Total cost in the bounded agent context pack for the requested scope.

When available: Available whenever this metric applies.

stable

Agent Context Cost Total Tokens

Total tokens in the bounded agent context pack for the requested scope.

When available: Available whenever this metric applies.

stable

Agent Context Cost Total Requests

Model turns in the bounded agent context pack for the requested scope.

When available: Available whenever this metric applies.

stable

Agent Context Cost Total Sessions

Session count in the bounded agent context pack for the requested scope.

When available: Available whenever this metric applies.

stable

Agent Context Cost Provenance

Cost, freshness, and customer-facing provenance labels for the agent context cost summary.

When available: Available whenever this metric applies.

stable

Agent Context Cost Top Models[*]

Top model cost rows included in the bounded agent context pack.

When available: Available whenever this metric applies.

stable

Agent Context Cost By Source[*]

Source cost rows included in the bounded agent context pack.

When available: Available whenever this metric applies.

stable

Agent Context Sources[*]

Source health rows included in the bounded agent context pack.

When available: Available whenever this metric applies.

stable

Agent Context Subscriptions[*]

Subscription attribution rows included in the bounded agent context pack.

When available: empty when subscription data is unavailable or cannot be safely scoped

stable

Agent Context Subscription Totals

Subscription monthly-vs-PAYG rollup included in the bounded agent context pack.

When available: May be unavailable when subscription totals cannot be safely scoped

stable

Dashboard Summary Period Start

Inclusive start date for a dashboard summary window.

When available: Available whenever this metric applies.

stable

Dashboard Summary Period End

Inclusive end date for a dashboard summary window.

When available: Available whenever this metric applies.

stable

Dashboard Summary Days

Rolling day count used for the dashboard summary window.

When available: Available whenever this metric applies.

stable

Dashboard Summary Account First Telemetry At

All-time earliest telemetry received_at for the user. Authoritative anchor for the new-account ramp UI (Day-N pill, pre-tracking spend clip, hour preview, activity backfill). Null when the user has never emitted telemetry.

When available: Available whenever this metric applies.

stable

Dashboard Summary Account First Telemetry Status

unavailable when a Results-served summary has no Results owner for the all-source first-telemetry anchor (D-57 §2); the anchor is then withheld rather than read from previous stores.

When available: omitted except on a Results-served summary

experimental

Dashboard Summary Cloud Work Status

unavailable when a Results-served summary has no Results owner for the cloud-collector metadata (D-57 §2); cloud work is then withheld rather than read from previous stores.

When available: omitted except on a Results-served summary

experimental

Dashboard Summary Unavailable Fields

Requested summary fields a Results-served body cannot answer from its grouped cost read, each with its reason: by_device needs each session's last activity; by_source_hourly is not read on the day-bucket summary. Those lists stay empty rather than zero-filled.

When available: omitted except on a Results-served summary that requested breakdowns

experimental

Dashboard Summary Cloud Work

Opt-in provider-cloud work counts, lifecycle mix, coverage, and collector freshness for a supported window.

When available: nullable unless explicitly requested; unavailable when rollout policy, consent, or collector health does not authorize a current view

experimental

Dashboard Summary Total Usage

Total token usage over the dashboard summary window.

When available: Available whenever this metric applies.

stable

Dashboard Summary Total Cost USD

Total cost over the dashboard summary window.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Kind

Whether the dashboard summary cost is actual or estimated.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Source Kind

Internal cost provenance value for the dashboard summary.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Source Label

Human-readable cost source label for the dashboard summary.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Projection Status

Pricing projection state for the dashboard summary.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Is Stale

Whether the dashboard summary cost was projected from stale pricing data.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Stale Reason

Why the dashboard summary totals carry a freshness qualifier: an in-flight local-data reconciliation, ordinary gold-rollup lag, or fresh.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Reconciliation Active

Whether a totals-pausing local-data reconciliation (account/machine backfill) is in flight for the dashboard summary totals.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Pricing Provenance

Sanitized pricing audit snapshot for the dashboard summary.

When available: May be unavailable when no pricing projection has been built

stable

Dashboard Summary Total Requests

Model turns over the dashboard summary window. Sessions tracked only through live telemetry count telemetry records instead.

When available: Available whenever this metric applies.

stable

Dashboard Summary Total Sessions

Total session count over the dashboard summary window.

When available: Available whenever this metric applies.

stable

Dashboard Summary By Model[*]

Dashboard summary usage and cost totals grouped by model.

When available: Available whenever this metric applies.

stable

Dashboard Summary Daily[*]

Dashboard summary daily usage and cost buckets.

When available: Available whenever this metric applies.

stable

Dashboard Summary By Source[*]

Dashboard summary usage and cost totals grouped by source.

When available: Available whenever this metric applies.

stable

Dashboard Summary By Source Daily[*]

Dashboard summary daily cost buckets grouped by source.

When available: Available whenever this metric applies.

stable

Dashboard Summary By Source Hourly[*]

Dashboard summary hourly cost buckets grouped by source.

When available: Available whenever this metric applies.

stable

Dashboard Summary Cost Breakdown

Cost split for the dashboard summary.

When available: Available whenever this metric applies.

stable

Dashboard Summary By Device[*]

Dashboard summary local device contribution rows.

When available: Available whenever this metric applies.

stable

Dashboard Summary Session Breakdown[*]

Dashboard summary session counts grouped by AI app and computer.

When available: Available whenever this metric applies.

stable

Dashboard Summary Session Model Breakdown[*]

Dashboard summary distinct-session counts per model; multi-model sessions count once per model they touched.

When available: Available whenever this metric applies.

stable

Dashboard Summary Previous Period

Previous equivalent dashboard summary window.

When available: nullable unless compare_previous is requested

stable

Activity Hourly Period Start

Inclusive local calendar start for the hourly activity query.

When available: Available whenever this metric applies.

stable

Activity Hourly Period End

Inclusive local calendar end for the hourly activity query.

When available: Available whenever this metric applies.

stable

Activity Hourly Complete Before

Exclusive UTC boundary before which hourly buckets are complete enough for recurring-period denominators.

When available: Available whenever this metric applies.

stable

Activity Hourly Days

Rolling day count for the hourly activity query.

When available: Available whenever this metric applies.

stable

Activity Hourly Timezone

IANA timezone used for local day, hour, and weekday fields.

When available: Available whenever this metric applies.

stable

Activity Hourly Dimension

Grouping dimension requested for the hourly activity query; total deduplicates sessions across sources within each hour.

When available: Available whenever this metric applies.

stable

Activity Hourly Origin

Activity source path used for this response.

When available: Available whenever this metric applies.

stable

Activity Hourly Coverage Status

Whether local activity buckets fully cover the requested period.

When available: complete only when every eligible latest local snapshot has matching bucket totals; partial, unsupported, or empty responses must not be rendered as heatmaps

stable

Activity Hourly Unavailable Reason

Why the Results hourly activity body withheld an answer.

When available: nullable; present for a typed unavailable Results body

stable

Activity Hourly Eligible Session Count

Latest local snapshots considered for hourly activity coverage.

When available: Available whenever this metric applies.

stable

Activity Hourly Covered Session Count

Eligible latest local snapshots with complete activity bucket coverage.

When available: Available whenever this metric applies.

stable

Activity Hourly Missing Session Count

Eligible latest local snapshots missing complete activity bucket coverage.

When available: Available whenever this metric applies.

stable

Activity Hourly Last Bucket At

Newest covered UTC activity bucket in the response window.

When available: May be unavailable when no complete bucket rows exist