Duo workflows

This data source provides aggregated metrics about GitLab Duo Agent Platform flows (also called Duo workflows) run across your project or group, including how many credits they used.

Allowed modes

• analytics

Allowed scopes

To aggregate GitLab Duo workflows across multiple groups or projects, use a list with the in operator. For more information, see multiple groups and projects.

Query filters

Use these fields in the query parameter to filter your results.

Active days

Description: Filter flows by how many distinct days their user created flows in the selected period. activeDays = 1 matches users who were active on exactly one day.

Allowed value types: Number

Notes:

• The count covers the whole selected period, per user.
• You can combine two range operators for a window, for example activeDays >= 2 and activeDays <= 5, but not = with another comparison on the same field.

Created

Description: Filter flows by date they were created. Use multiple created conditions to define a specified time period.

Allowed value types:

• AbsoluteDate (in the format YYYY-MM-DD)
• RelativeDate (in the format <sign><digit><unit>, where sign is +, -, or omitted, digit is an integer, and unit is one of d (days), w (weeks), m (months), or y (years))

Notes:

• For the = operator, GLQL considers the time range from 00:00 to 23:59 in the user’s time zone.

Drop reason

Description: Filter failed flows by the reason recorded when they were dropped.

Allowed value types:

• Enum, one of job_failed, no_status_reported, no_activity, workload_not_created, api_reported, preconditions_not_met, or start_failed
• List (use in or not in operator for multiple values)

Notes:

• no_activity is the stuck-session cleanup, no_status_reported means the CI workload ended without a terminal status, job_failed is a CI job that did not succeed, workload_not_created, preconditions_not_met, and start_failed mean the flow never started, and api_reported is the agent reporting an error.
• Flows that did not fail, or that were dropped before reasons were recorded, have no reason. No operator matches them, so != and not in exclude them as well.

Flow type

Description: Filter flows by their type. For example, software_development or code_review/v1.

Allowed value types:

• String
• List (use in or not in operator for multiple values)

Flow types used

Description: Filter flows by how many distinct flow types their user ran in the selected period. flowTypesUsed >= 2 matches users who used two or more flow types.

Allowed value types: Number

Notes:

• The count covers the whole selected period, per user.
• The same combination rule applies: use two range operators for a window, but not = with another comparison on the same field.

Status

Description: Filter flows by their lifecycle status.

Allowed value types:

• Enum, one of created, running, paused, finished, failed, stopped, input_required, plan_approval_required, or tool_call_approval_required
• List (use in or not in operator for multiple values)

Notes:

• The three *_required statuses are flows waiting on the user, and are distinct from paused.
• Statuses are matched case-insensitively. Other spellings, like completed or aborted, are not accepted.

User

Description: Filter by the user who ran the flow.

Allowed value types:

• Number (user ID)
• List (use in or not in operator for multiple user IDs)

Dimensions

Metrics

Notes:

• The returningUsersCount, joinedUsersCount, churnedUsersCount, and previousPeriodUsersCount metrics compare each created bucket with the preceding one, so they are only valid when the created dimension is also selected. A created filter alone is not enough.
• The mergedMrCount, closedMrCount, and openMrCount metrics count merge requests a flow created that were later merged, closed without merging, or are still open. They can lag behind the current state of those merge requests, so a merge request merged recently can still be counted as open.

Sort fields

Sort by any field included in your selected dimensions or metrics. For more information, see analytics mode sorting.

Examples

• Flows by flow type for the last 30 days:

  yaml

  ```glql
  title: "Duo workflows by flow type (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  dimensions: flowType as "Flow"
  metrics: totalCount as "Flows", usersCount as "Users", creditsUsedSum as "Credits"
  sort: totalCount desc
  ```

• Monthly credit usage trend:

  yaml

  ```glql
  title: "Monthly Duo workflow credit usage trend"
  display: columnChart
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -90d
  dimensions: created(monthly) as "Month"
  metrics: creditsUsedSum as "Credits used"
  sort: created asc
  ```

• Monthly credit usage excluding chat flows:

  yaml

  ```glql
  title: "Monthly Duo workflow credits, excluding chat"
  display: columnChart
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -90d and flowType not in ("chat", "agentic_chat/v1")
  dimensions: created(monthly) as "Month", flowType as "Flow"
  metrics: creditsUsedSum as "Credits used"
  sort: created asc
  ```

• Credits by model for the last 30 days:

  yaml

  ```glql
  title: "Duo workflow credits by model (last 30 days)"
  display: barChart
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  dimensions: model as "Model"
  metrics: creditsUsedSum as "Credits used"
  sort: creditsUsedSum desc
  ```

• Top users by credits for the last 30 days:

  yaml

  ```glql
  title: "Top users by Duo workflow credits (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  dimensions: user as "User"
  metrics: totalCount as "Flows", creditsUsedSum as "Credits", creditsUsedMean as "Avg credits per flow"
  sort: creditsUsedSum desc
  limit: 10
  ```

• Flows by status for the last 30 days:

  yaml

  ```glql
  title: "Duo workflows by status (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  dimensions: status as "Status"
  metrics: totalCount as "Flows", createdMrCountSum as "MRs created", mergedMrCountSum as "MRs merged", openMrCountSum as "MRs open"
  sort: status asc
  ```

• Weekly user retention:

  yaml

  ```glql
  title: "Weekly Duo workflow user retention"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -90d
  dimensions: created(weekly) as "Week"
  metrics: usersCount as "Users", returningUsersCount as "Returning", joinedUsersCount as "Joined", churnedUsersCount as "Churned"
  sort: created desc
  ```

• Users by activity tier and flow type for the last 30 days:

  yaml

  ```glql
  title: "Duo workflow users by activity tier (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  dimensions: userTier(thresholds=[4, 25, 100]) as "Tier", flowType as "Flow"
  metrics: totalCount as "Flows", usersCount as "Users"
  sort: userTier asc
  ```

• Flow counts by status side by side, for the last 30 days:

  yaml

  ```glql
  title: "Duo workflows by outcome (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  metrics: totalCount as "All", totalCount(status="finished") as "Completed", totalCount(status=["paused", "input_required"]) as "Waiting", totalCount(status="failed") as "Failed"
  ```

• Group comparison for the last 30 days:

  yaml

  ```glql
  title: "Duo workflow usage by subgroup (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  dimensions: group(depth=2) as "Group"
  metrics: usersCount as "Users", creditsUsedSum as "Credits"
  sort: usersCount desc
  ```

• Overall usage without grouping:

  yaml

  ```glql
  title: "Duo workflow usage overview (last 30 days)"
  display: table
  mode: analytics
  query: type = DuoWorkflow and group = "gitlab-org" and created > -30d
  metrics: totalCount as "Flows", usersCount as "Users", projectsCount as "Projects", creditsUsedQuantile(0.5) as "Median credits"
  ```