Exe-finops adds a reporting layer on ClickHouse that removes ETL, Databricks, and Snowflake

The core change is the introduction of exe-finops, a tool that reads every billing webhook event written into ClickHouse and layers a lightweight reporting interface on top. Instead of building a full ETL pipeline that moves data into Databricks or Snowflake, the system keeps raw events in ClickHouse and generates reports from them on demand. This shift replaces a multi-tool stack with a single ClickHouse-centric workflow.

Adopting the approach requires configuring webhook firehose endpoints for each billing provider and ensuring those events are persisted in ClickHouse. Developers also need to maintain the exe-finops code that pulls events, optionally enriches them via provider APIs, and writes the final report data into a SQLite database. The cost is limited to the engineering effort for setup and the modest storage needs of ClickHouse and SQLite, with no additional SaaS licensing for ETL or warehousing services.

The method stops working when required events are not captured by the firehose or when a provider’s API does not expose needed details, forcing the system to fall back to manual investigation. SQLite, while convenient for small-scale reporting, may become a bottleneck for very large query volumes or complex joins. Consequently, the solution is best suited for teams that can guarantee comprehensive webhook coverage and moderate reporting loads.

In practice the new stack enables rapid answers to support questions, such as tracing shared payment methods across accounts, by letting an LLM-driven agent generate SQL against ClickHouse. This reduces time spent navigating provider dashboards and improves the detail and speed of financial dispute resolution. The example shows how combining a columnar event store with prompt-based agents can replace heavyweight data pipelines for specific operational reporting needs.