kxsystems/kdbx-knowledge
KDB-X workflows and patterns for the next-generation KX time-series database
Use when working with KDB-X modules, AI libraries (vector search, HNSW, IVF, BM25, time-series similarity, anomaly detection), GPU acceleration, Parquet, kURL REST client, object storage, or the module framework. Also use when encountering module-not-found errors, vector dimension mismatches, HNSW index issues, GPU setup issues, or nyi errors from using \l instead of use.
Use this skill whenever working with the KDB-X DB Service - a containerised time-series database distributed via Docker Compose. Trigger for any task involving: setting up or resetting the DB Service, ingesting data (CSV, streaming, REST, or third-party sources like OneTick), writing or optimising queries (Python client, q client, or REST/qSQL), or planning an end-to-end ingest pipeline. Also trigger for schema design questions, troubleshooting ingest jobs, or any reference to dbservice_client, session.import_files, session.import_data, session.query_simple, session.query_sql, session.query_q, session.querySimple, session.querySQL, session.queryQ, or the DB Service REST API on port 8080. Also trigger for: "run the FX recipe", "FX ingest recipe", "run the OneTick recipe", "run the OneTick LSE recipe", "run the OneTick LSE runbook", or any request to set up the full OneTick batch pipeline end-to-end from download through to ingest. Do NOT trigger for: KX Dashboards configuration; generic Docker Compose questions unrelated to the DB Service stack; or plain kdb+/q questions that don't reference the DB Service.
Writing qdoc annotations for aimeta so the compiler publishes tables and functions correctly. Mandatory @kind/@name markers, the chained @col modifier form, q-language traps that silently drop items, and the recompile loop. Use when adding or editing / @kind ... annotation blocks in a q codebase that loads aimeta, or when annotated items aren't surfacing in meta.json. For the wire model & tag vocabulary see ../../reference/agent-guide.md.
Discovering metadata on a running kdb+ process via aimeta — cold-start probing, the /meta HTTP route, kx-meta discover, qIPC fallback, and reading the tier / references[] fields in the response. Use when an agent needs to learn what tables and functions a kdb+ host exposes, when interpreting kx-meta discover output, when handling tier-1 fallback gracefully, or when walking references[] for vocabulary resolution. For the wire model and full meta.json shape see ../../reference/agent-guide.md.