Amr Taha (dc9f9738) at 20 Mar 09:23

... and 1 more commit

Amr Taha (c96c4304) at 20 Mar 09:23

What does this MR do and why?

• Adds Duo Code Suggestions aggregation engine to the GraphQL API.
• Renames the AiUsage GraphQL type to Analytics for a more generic naming. Updates all references across group/project types, resolvers, spec files, documentation, and the GraphQL introspection result. No functional changes — purely a rename/refactor

References

Screenshots or screen recordings

How to set up and validate locally

1. Setup ClickHouse in GDK
2. Run db seeds with FILTER=ai_usage_stats bundle exec rake db:seed_fu
3. Execute a query similar to the one shown in the screenshot above.
4. Validate output.

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

Thanks Adam! Approving with two main question

What does this MR do and why?

This MR adds the Siphon table configurations for the merge_requests clickhouse table.

You can test the replication with GDK with these instructions: gitlab-development-kit!5842

Related issue: #594122

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

Question: Don't we need to have the same setup and the other yml files for replication_targets?

Question: Is awardable_id enough here or do we need to include awardable_type? I think we need the type as well for correct referencing no?

Same question for label_links

@ahegyi Would you mind re-approving? Your approval got lost amid the latest commit pushes

Amr Taha (c14be789) at 20 Mar 08:14

... and 1 more commit

Amr Taha (f097f945) at 20 Mar 08:14

What does this MR do and why?

The branch rules link to a literal ^All branches$ instead of just showing all branches.

This MR fixes that by removing the query literal.

References

Screenshots or screen recordings

How to set up and validate locally

The video attached above shows the outcome and how to test.

1. Go to any project
2. Go to settings > merge requests
3. Click on branch rules
4. Click on all branch rules
5. Observe the link

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

Pipeline failuers are false-postiives. Starting a new pipeline and setting auto-merge

@ahegyi I can review this in the next hour or two. Is it WIP or will you assign another reviewer?

Thanks @pshutsin 💯

Three questions from my side with nothing blocker

What does this MR do and why?

• Added a new Aggregation Engine (AE) for AI usage events
• Extended the AE with filter formatting support — allows converting string values to integers and GIDs to regular IDs.

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

ClickHouse review

SELECT
    ch_aggregation_inner_query.aeq_user_id                                    AS aeq_user_id,
    toStartOfInterval(ch_aggregation_inner_query.aeq_timestamp, INTERVAL 1 month) AS aeq_timestamp,
    COUNT(*)                                                                   AS aeq_total_count

FROM (
    SELECT
        ai_usage_events.user_id        AS aeq_user_id,
        timestamp                      AS aeq_timestamp,
        ai_usage_events.user_id,
        ai_usage_events.namespace_path,
        ai_usage_events.event,
        ai_usage_events.timestamp

    FROM ai_usage_events

    WHERE
        CASE
            WHEN event IN (1, 2, 3, 4, 5)            THEN 'code_suggestions'
            WHEN event IN (6)                         THEN 'chat'
            WHEN event IN (7)                         THEN 'troubleshoot_job'
            WHEN event IN (8, 9, 19, 20, 21, 22)     THEN 'agent_platform'
            WHEN event IN (10, 11, 12, 13, 14, 15, 16, 17, 18) THEN 'code_review'
            WHEN event IN (23, 24)                    THEN 'mcp'
            ELSE NULL
        END IN ('troubleshoot_job')

        AND ai_usage_events.timestamp >= 1758284081.859363
        AND startsWith(namespace_path, '9970/')

    GROUP BY ALL
) ch_aggregation_inner_query

GROUP BY ALL
ORDER BY
    aeq_timestamp    ASC,
    aeq_total_count  DESC

SELECT
    ch_aggregation_inner_query.aeq_user_id                                        AS aeq_user_id,
    toStartOfInterval(ch_aggregation_inner_query.aeq_timestamp, INTERVAL 1 month) AS aeq_timestamp,
    COUNT(DISTINCT ch_aggregation_inner_query.aeq_features_count)                 AS aeq_features_count

FROM (
    SELECT
        ai_usage_events.user_id AS aeq_user_id,
        timestamp               AS aeq_timestamp,
        CASE
            WHEN event IN (1, 2, 3, 4, 5)                      THEN 'code_suggestions'
            WHEN event IN (6)                                   THEN 'chat'
            WHEN event IN (7)                                   THEN 'troubleshoot_job'
            WHEN event IN (8, 9, 19, 20, 21, 22)               THEN 'agent_platform'
            WHEN event IN (10, 11, 12, 13, 14, 15, 16, 17, 18) THEN 'code_review'
            WHEN event IN (23, 24)                              THEN 'mcp'
            ELSE NULL
        END                     AS aeq_features_count,
        ai_usage_events.user_id,
        ai_usage_events.namespace_path,
        ai_usage_events.event,
        ai_usage_events.timestamp

    FROM ai_usage_events

    WHERE ai_usage_events.timestamp >= 1758284354.372006
    AND startsWith(namespace_path, '9970/')

    GROUP BY ALL
) ch_aggregation_inner_query

GROUP BY ALL
ORDER BY
    aeq_timestamp     ASC,
    aeq_features_count DESC

Related to #589610

Suggestion (Non Blocker): Similar to the pervious suggestion on table-based tests pattern

I have no strong opinion about it and I think it's non-linear convention as some specs are unique and more readable with being table-based. So I'll leave it to your judgement

Question: I see we are moving from ai_usage to analytics from the previous MR.

Should we still need to mount AiUsageEvents AE in ee/app/graphql/types/analytics/ai_usage/ai_usage_type.rb or it can be done in a separate MR once the other one is merged?

Suggestion (Non Blocker): Do we need to define extras here or until we have a use case for it?

What does this MR do and why?

• Adds Duo Code Suggestions aggregation engine to the GraphQL API.
• Renames the AiUsage GraphQL type to Analytics for a more generic naming. Updates all references across group/project types, resolvers, spec files, documentation, and the GraphQL introspection result. No functional changes — purely a rename/refactor

References

Screenshots or screen recordings

How to set up and validate locally

1. Setup ClickHouse in GDK
2. Run db seeds with FILTER=ai_usage_stats bundle exec rake db:seed_fu
3. Execute a query similar to the one shown in the screenshot above.
4. Validate output.

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

Nicely done on ignoring these fields, approving 👍