> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.vapi.ai/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.vapi.ai/_mcp/server. # Get traffic split metrics GET https://api.vapi.ai/traffic-allocations/metrics Per-assistant-version call metrics over a time window: call counts, configured vs actual traffic share, duration, cost per minute, latency, ended reasons, a bucketed timeseries, and optional structured output aggregations. Calls with no assistant version are reported as an 'unversioned' row. Traffic splitting is in beta, rolling out to select organizations; requests from organizations without access receive a 403. Reference: https://docs.vapi.ai/api-reference/traffic-allocations/traffic-allocation-controller-metrics-get ## Authentication - `Authorization` header (bearer token, required) — Authenticate server-side requests with a private Vapi API key. Create or copy a key from the [Vapi Dashboard](https://dashboard.vapi.ai) and send it in the `Authorization` header as `Bearer `. Keep private API keys out of client-side code and public repositories. ## Request ### Query parameters - `assistantId` (string, required) — The assistant whose per-version traffic metrics to return. - `from` (datetime, required) — Start of the metrics window (inclusive), ISO 8601 datetime. - `to` (datetime, required) — End of the metrics window (exclusive), ISO 8601 datetime. Must be later than from, and the window may hold at most 2000 buckets of the requested width. - `bucket` (enum, optional) — Bucket size for the timeseries. Defaults to 'day'. - Allowed values: `hour`, `day` - `structuredOutputIds` (list of any, optional) — Structured output ids to aggregate per version, as an array or a comma-separated string. At most 5. ## Response ### 200 - `window` (TrafficTabMetricsWindowDTO, required) - `totalCalls` (double, required) — All calls for the assistant in the window, every version. - `versions` (list of TrafficTabVersionMetricsDTO, required) — One row per version with calls in the window, plus the unversioned row when it has any. Ordered by calls descending, then by version label (numeric-aware), the unversioned row last among equals. - `currentAllocation` (TrafficAllocation, optional, nullable) — The allocation currently governing the assistant. Null when traffic splitting was never configured. ## Errors ### 400 Bad Request Error assistantId is missing or not a UUID, from/to are missing, malformed, or inverted, the window spans more than 2000 buckets, or structuredOutputIds has more than 5 entries or a non-UUID entry. - `any` ### 403 Forbidden Error Traffic splitting is not enabled for your organization. The feature is in beta, rolling out to select organizations. - `any` ## Types ### TrafficTabMetricsWindowDTO - `from` (string, required) — Start of the window (inclusive), ISO 8601. - `to` (string, required) — End of the window (exclusive), ISO 8601. - `bucket` (enum, required) - Allowed values: `hour`, `day` ### TrafficTabVersionMetricsDTO - `calls` (double, required) - `actualPercentage` (double, required) — calls / totalCalls * 100 over the window, rounded to 3 decimals. - `durationSeconds` (TrafficTabDurationSecondsDTO, required) - `costPerMinute` (TrafficTabCostPerMinuteDTO, required) - `latencyMs` (TrafficTabLatencyMsDTO, required) - `endedReasons` (list of TrafficTabEndedReasonCountDTO, required) — Full per-reason counts; the UI unions and top-Ns. - `timeseries` (list of TrafficTabTimeseriesPointDTO, required) - `assistantVersionId` (string, optional, nullable) - `assistantVersionLabel` (string, optional, nullable) — The published version label, such as "v7". Null when no version row resolves (unversioned calls, or a since-purged version). - `unversioned` (boolean, optional) — True for the row aggregating calls with no assistant version (stored as SQL NULL). - `configuredPercentage` (double, optional, nullable) — The version's share in the currently-governing allocation. Null when the version is not in the current split — never rendered as 0%. - `structuredOutputs` (list of TrafficTabStructuredOutputDTO, optional) — Present only when structuredOutputIds was given. ### TrafficAllocation - `id` (string, required) — Unique identifier. The most recently created allocation for an assistant is the one in effect. - `orgId` (string, required) - `allocationIntent` (enum, required) — 'explicit' splits calls across this allocation's targets. 'follow-latest' sends every call to the newest published version. - Allowed values: `follow-latest`, `explicit` - `createdAt` (datetime, required) — When this allocation was created, which is also when it took effect. - `actorType` (enum, required) — Who created this allocation. 'system' means Vapi created it automatically, for example when a publish advances a follow-latest allocation. - Allowed values: `user`, `api-key`, `system` - `targets` (list of TrafficAllocationTarget, required) — The versions this allocation splits calls across, in position order. Empty for follow-latest allocations. - `assistantId` (string, optional) — The assistant this allocation splits calls for. - `actorId` (string, optional) — The user id or API key id that created this allocation. Absent for system rows. - `actorEmail` (string, optional) — Email of the user who created this allocation, as of that time. - `description` (string, optional) — The note given when this allocation was created, if any. ### TrafficTabDurationSecondsDTO - `mean` (double, optional, nullable) - `stddev` (double, optional, nullable) - `p50` (double, optional, nullable) - `p95` (double, optional, nullable) ### TrafficTabCostPerMinuteDTO - `breakdown` (TrafficTabCostBreakdownPerMinuteDTO, required) - `mean` (double, optional, nullable) ### TrafficTabLatencyMsDTO - `model` (TrafficTabLatencyStatsDTO, required) - `voice` (TrafficTabLatencyStatsDTO, required) - `transcriber` (TrafficTabLatencyStatsDTO, required) - `turn` (TrafficTabLatencyStatsDTO, required) ### TrafficTabEndedReasonCountDTO - `reason` (string, required) - `count` (double, required) ### TrafficTabTimeseriesPointDTO - `bucketStart` (string, required) — Start of the bucket, ISO 8601 (UTC). - `calls` (double, required) - `durationSecondsMean` (double, optional, nullable) - `costPerMinuteMean` (double, optional, nullable) - `turnLatencyMsMean` (double, optional, nullable) ### TrafficTabStructuredOutputDTO - `structuredOutputId` (string, required) - `type` (enum, required) - Allowed values: `number`, `string`, `boolean`, `unknown` - `present` (double, required) — Calls in the window where this output produced a result. - `numeric` (TrafficTabStructuredOutputNumericDTO, optional) - `valueCounts` (list of TrafficTabStructuredOutputValueCountDTO, optional) — Top 10 values by count, for string/boolean outputs. ### TrafficAllocationTarget - `assistantVersion` (string, required) — The assistant version this target sends calls to, such as "v7". - `position` (double, required) — The target's place in the split, starting at 0. Set from the order of the targets array. - `percentage` (double, required) — Share of calls sent to this version, from 0 to 100 with up to three decimal places. Targets add up to exactly 100. A 0% target keeps the version in the split without sending it calls. ### TrafficTabCostBreakdownPerMinuteDTO - `llm` (double, optional, nullable) - `stt` (double, optional, nullable) - `tts` (double, optional, nullable) - `vapi` (double, optional, nullable) - `transport` (double, optional, nullable) ### TrafficTabLatencyStatsDTO - `mean` (double, optional, nullable) - `p50` (double, optional, nullable) - `p95` (double, optional, nullable) ### TrafficTabStructuredOutputNumericDTO - `mean` (double, optional, nullable) - `stddev` (double, optional, nullable) - `p50` (double, optional, nullable) - `p95` (double, optional, nullable) - `sum` (double, optional, nullable) ### TrafficTabStructuredOutputValueCountDTO - `value` (string, required) - `count` (double, required) ## Examples **Response** ```json { "window": { "from": "string", "to": "string", "bucket": "hour" }, "totalCalls": 1.1, "versions": [ { "calls": 1.1, "actualPercentage": 1.1, "durationSeconds": { "mean": 1.1, "stddev": 1.1, "p50": 1.1, "p95": 1.1 }, "costPerMinute": { "breakdown": { "llm": 1.1, "stt": 1.1, "tts": 1.1, "vapi": 1.1, "transport": 1.1 }, "mean": 1.1 }, "latencyMs": { "model": { "mean": 1.1, "p50": 1.1, "p95": 1.1 }, "voice": { "mean": 1.1, "p50": 1.1, "p95": 1.1 }, "transcriber": { "mean": 1.1, "p50": 1.1, "p95": 1.1 }, "turn": { "mean": 1.1, "p50": 1.1, "p95": 1.1 } }, "endedReasons": [ { "reason": "string", "count": 1.1 } ], "timeseries": [ { "bucketStart": "string", "calls": 1.1, "durationSecondsMean": 1.1, "costPerMinuteMean": 1.1, "turnLatencyMsMean": 1.1 } ], "assistantVersionId": "string", "assistantVersionLabel": "string", "unversioned": true, "configuredPercentage": 1.1, "structuredOutputs": [ { "structuredOutputId": "string", "type": "number", "present": 1.1, "numeric": { "mean": 1.1, "stddev": 1.1, "p50": 1.1, "p95": 1.1, "sum": 1.1 }, "valueCounts": [ { "value": "string", "count": 1.1 } ] } ] } ], "currentAllocation": { "id": "string", "orgId": "string", "allocationIntent": "follow-latest", "createdAt": "2024-01-15T09:30:00Z", "actorType": "user", "targets": [ { "assistantVersion": "string", "position": 1.1, "percentage": 1.1 } ], "assistantId": "string", "actorId": "string", "actorEmail": "string", "description": "string" } } ```