[docs] Add a new quickstart tutorial for auto-incremental column and Agg Merge Engine

[wuchong]

Search before asking

• I searched in the issues and found nothing similar.

Description

A new quickstart tutorial: “Real-Time User Profile” that demonstrates how to combine the auto-increment column and Agg Merge Engine features in a realistic end-to-end scenario.

Original issue:#2598

Willingness to contribute

• I'm willing to submit a PR!

[Prajwal-banakar]

Hi @wuchong Please assign this to me, i'm intrested in this task!

[wuchong]

Thanks, @Prajwal-banakar, for volunteering. However, a use case–focused quickstart requires a thorough understanding of both the feature and the real-world scenario it addresses, so it can effectively attract and guide users to adopt the functionality. Ideally, the use case should reflect a realistic business scenario rather than an abstract or synthetic example—this helps users relate to it more easily and see its practical value.

I’m not certain about your familiarity with this specific feature, but could you briefly outline how you plan to demonstrate it? For instance, which concrete use case would you choose, and why?

[Prajwal-banakar]

Hi @wuchong,

I plan to demonstrate these features through a 'Real-Time User Identity & Activity Tracker' use case. This scenario reflects a common business need: mapping high-cardinality business keys (like emails) to compact system IDs while maintaining real-time metrics.

The plan for the tutorial includes:

Identity Mapping (Auto-Increment): Creating a 'Dictionary Table' where incoming email strings from a raw stream are automatically mapped to a unique BIGINT UID. This highlights the feature’s role as a global ID generator that replaces manual, complex ID management.

Stateless Aggregation (Merge Engine): Creating a 'User Profile' table that uses the aggregation merge engine to track total_spend (sum) and last_active (max) directly in the storage layer. This demonstrates how Fluss allows Flink jobs to remain lightweight and nearly stateless.

End-to-End Pipeline: Showing a single Flink SQL query using the lookup.insert-if-not-exists hint. This is the 'magic' that ties it together, showing how a single join can look up an ID, create it if it's missing, and trigger the downstream aggregation.

Production Reliability: A brief section on Undo Recovery, explaining how Fluss ensures exactly-once accuracy for these aggregates during Flink failovers.

Does this approach align with the project goals for the docs?

[wuchong]

Thanks, @Prajwal-banakar, this sounds great and aligns well with my thinking!

A small suggestion: consider using INT instead of BIGINT for the uid column, as it is more storage and computation efficient and sufficient for most use cases.

Additionally, here are some documentation references for similar use cases:

• https://www.alibabacloud.com/help/en/hologres/use-cases/real-time-pre-aggregated-uv-computing-solution-based-on-hologres-flink?spm=a2c63.p38356.help-menu-113622.d_3_4_2_2.29f12c532DnG8V
• https://www.alibabacloud.com/help/en/hologres/use-cases/user-behavior-analysis-overview/?spm=a2c63.p38356.help-menu-113622.d_3_4_2.44245d55s2qYZL&scm=20140722.H_410616._.OR_help-T_intl~en-V_1
• https://docs.starrocks.io/docs/using_starrocks/query_acceleration_with_auto_increment/
• https://docs.starrocks.io/docs/using_starrocks/distinct_values/Using_bitmap/

[Prajwal-banakar]

Hi @wuchong can we use a diagram!?

[wuchong]

@Prajwal-banakar yes, we can, of course