Guides ClickHouse OLAP insert, query, and schema patterns from Rust and Node.js.
npx skills add https://github.com/avbel/ai-skills --skill clickhouse-avbel
Please help me install this Agent Skill. Skill: clickhouse Source: https://github.com/avbel/ai-skills/tree/main/skills/clickhouse Command: npx skills add https://github.com/avbel/ai-skills --skill clickhouse-avbel
clickhouse crate (typed Row structs, buffered inserts, RowBinary) and the official @clickhouse/client for Node.js, including the browser/edge build. - Insert and query patterns: Large batch inserts, async inserts, ReplacingMergeTree deduplication, materialized-view rollups, TTL retention, and ORDER BY/skip-index design. - Antipattern catalog: A table of common mistakes (per-row inserts, heavy mutations, FINAL everywhere, over-partitioning) with the correct alternative for each. - Use Case: When building an event-ingestion pipeline that writes millions of rows per day, use this Skill to generate batched insert code with async_insert settings and a ReplacingMergeTree table instead of row-by-row writes. ## Quick Start Ask the agent to write a Rust or Node.js function that batch-inserts events into a ClickHouse table following ClickHouse best practices.High-intent search queries and answers about installing and using this skill.
Use @clickhouse/client with client.insert({ table, values, format: 'JSONEachRow' }) passing a large array of objects per call. Aim for tens of thousands of rows per insert, or enable async_insert=1 so the server batches many small producers for you.
Use the clickhouse crate, the recommended async HTTP client with typed Row derives and buffered insert APIs. The older clickhouse-rs TCP crate still exists, but the HTTP crate is preferred for new code and ClickHouse Cloud compatibility.
The error comes from many small frequent inserts creating thousands of tiny parts that background merges cannot keep up with. Fix it by batching inserts into large blocks or enabling async_insert so the server merges small inserts server-side.
ClickHouse is append-mostly and not suited for OLTP-style mutations. Use lightweight DELETE sparingly, model updates with ReplacingMergeTree or CollapsingMergeTree, and avoid ALTER TABLE UPDATE mutations which rewrite entire parts.
Use a ReplacingMergeTree table with a version column and apply FINAL or GROUP BY with argMax(ver) at query time. Deduplication happens eventually during background merges, so never assume duplicates are gone immediately after insert.
MaelwalserOptimize ClickHouse schemas and queries for analytical workloads.
archibateProvide ClickHouse schema and query optimization patterns for analytical workloads.
nksethWrite, debug, and refactor JavaScript code with ES2023+ and async/await.
beratcelik1Optimize PostgreSQL SQL queries and schema design with index and configuration guidance.