Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsClickHouse Analytics Architect
Design and tune ClickHouse analytical systems around real query, ingestion, reliability, and cost constraints.
$9.99
ClickHouse Analytics Architect
Example session with this skill installed
We run ClickHouse [version and Cloud/self-managed deployment]. Business questions: [questions]. Tables/DDL and sample rows: [redacted schema]. Ingestion: [events/sec, batch size, freshness], data volume/growth, retention, query examples with EXPLAIN, concurrency, latency SLOs, and cost constraints: [details]. Diagnose the main bottleneck, recommend a table/model/ORDER BY/partition strategy and only justified indexes or pre-aggregation, identify risks and unknowns, give validation queries and rollout/rollback steps, and state which conclusions need measurement. Use the documented version; do not invent metrics or performance gains.
- Read your context and instructions
- Compiled the clickhouse analytics architect
Executive finding
No schema, query, metrics, or version details were supplied, so this example cannot identify a real bottleneck or promise a speedup. It provides a version-aware investigation plan and a safe design review template.
1. Establish the workload
Collect the exact ClickHouse version and deployment model; business questions; representative SQL; expected results; data freshness; event and row volumes; ingest frequency and batch size; retention; growth; concurrency; latency targets; and cost limits. Redact credentials, private identifiers, and sensitive row values. Preserve column types and query structure when redacting.
2. Evidence to inspect
| Area | Evidence | What it can reveal |
|---|---|---|
| Query | SQL, EXPLAIN output, rows/bytes read, duration, memory | Scan volume, pruning, joins, aggregation, and plan behavior |
| Table | DDL, engine, ORDER BY, partition key, codecs, granularity | Physical layout and lifecycle tradeoffs |
| Ingestion | Batch sizes, inserts, parts, merges, deduplication | Small-insert pressure, merge health, write amplification |
| Operations | Version, replicas, shards, disk, memory, errors, SLOs | Capacity, reliability, and deployment constraints |
Treat each observation as scoped to its timestamp, environment, and representative workload. Confirm version-specific semantics in the official documentation for the target release.
3. Architecture review
Design the table from dominant filters, grouping, ordering, cardinality, locality, and retention needs. Choose a MergeTree-family engine based on correctness and data lifecycle. Use ORDER BY to support common access patterns; do not treat partitioning as a substitute for sorting keys. Keep partition cardinality operationally manageable. Select codecs and data types deliberately.
Consider skipping indexes, projections, materialized views, dictionaries, joins, or pre-aggregation only when measured query shapes justify their storage, write, freshness, and maintenance costs. Plan materialized-view backfills and source semantics explicitly. Batch ingestion to avoid excessive parts; investigate merge pressure before increasing insert frequency. Treat mutations, deletes, TTLs, and schema changes as version- and workload-dependent operations.
4. Candidate plan format
| Decision | Recommendation | Evidence needed | Cost or risk | Validation |
|---|---|---|---|---|
| Table engine and model | Choose for correctness and lifecycle | DDL, update/delete patterns | Storage and operational complexity | Test representative inserts and reads |
| ORDER BY | Align with dominant filters/grouping | Query distribution and cardinality | Poor choice can increase scans | Compare rows/bytes read across workload |
| Partitions | Support retention and lifecycle | Retention and partition counts | Too many partitions add overhead | Inspect partition and part counts |
| Pre-aggregation | Add only for recurring expensive patterns | Query frequency and freshness needs | More storage and write complexity | Compare results and cost under concurrency |
5. Safe validation and rollout
Create a reproducible baseline using representative data and queries. Compare correctness first, then bytes and rows read, latency distributions, memory, ingestion behavior, merge health, and resource cost under realistic concurrency. Include warm/cold conditions where relevant and record the test environment. Change one major factor at a time. Stage the change, set success and rollback thresholds, monitor after release, and retain the previous schema/query path until verified.
Next inputs
Provide the exact version, deployment model, table DDL, representative query and EXPLAIN output, ingestion pattern, retention, observed metrics, concurrency, and SLO. With those facts, this skill can rank bottlenecks and produce a concrete design and validation plan. No benchmark or production outcome is claimed here.
Connects securely to your tools. The creator never sees your data.
About this skill
ClickHouse Analytics Architect
Design, diagnose, optimize, migrate, and document ClickHouse analytical systems using the actual workload, schema, deployment, version, and service goals. Translate business questions into query patterns, data models, MergeTree engine choices, sorting keys, partitioning, codecs, indexes, projections, materialized views, ingestion, retention, replication, sharding, observability, backup, and recovery plans.
Start with freshness, retention, ingest rate, volume, concurrency, SLOs, and cost constraints. Optimize measured rows and bytes processed, ingestion health, merges, memory, tail latency, and operational risk. Explain tradeoffs and validate recommendations against representative queries, query logs, explain plans, system metrics, and version-specific official documentation.
Do not invent workload facts, metrics, benchmark gains, provider capabilities, or production guarantees. Treat recommendations as hypotheses until measured; include rollback and safe rollout steps. ClickHouse features and behavior vary by version and deployment, so verify the target version before implementation. Do not make production changes automatically.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scanned
Verified clean 9 days ago
- Passed all security checks, Safe to install