Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Overview skill for OceanBase SeekDB. Routes to specialized sub-skills covering the full SeekDB lifecycle — install/deploy on a target machine, build from source, look up documentation, use the seekdb-cli for SQL/schema/vector ops, import CSV/Excel data, and query/export results. Use as a starting point when the user mentions seekdb, pyseekdb, "install seekdb", "build seekdb", "seekdb docs", "seekdb-cli", "import to seekdb", "query seekdb", or wants a lightweight standalone OceanBase-compatible d
.claude/skills/oceanbase-seekdb/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 1% | 0% |
SeekDB is a lightweight OceanBase-compatible database that can run as a standalone server (Homebrew / Docker / yum / apt / Windows MSI) or as an embedded Python module (pyseekdb). It also supports vector search, hybrid search, and AI functions.
This is the entry point. Pick the sub-skill that matches the user's task.
| Sub-skill | Use When | |-----------|----------| | install | Install / deploy SeekDB on a target machine — Homebrew, Docker, yum, apt, Windows MSI, or pip embedded. | | build | Build SeekDB binaries and packages from source — macOS, Linux, Android (cross), Windows, Python wheel. | | docs | Look up SeekDB documentation (SQL syntax, vector/hybrid search, integrations, SDK APIs, deployment guides) via catalog-based search. | | cli | Use seekdb-cli to interact with a SeekDB/OceanBase instance from the shell — SQL, schema exploration, table profiling, vector collections, AI models. | | importing | Import CSV / Excel files into SeekDB, with optional column vectorization for semantic search. | | querying | Query a SeekDB collection — scalar (metadata) search, hybrid (fulltext + semantic) search, export to CSV / Excel. |
| User says... | Go to | |--------------|-------| | "Install seekdb on my Mac / Linux / Windows" | install | | "I want to use seekdb from Python" | install — pip embedded mode | | "Build seekdb from source" / "make a release rpm/deb/tgz/apk/installer" | build | | "How does seekdb's hybrid search / vector index / SQL function X work?" | docs | | "Connect to seekdb from shell" / "run SQL via CLI" / "list tables" / "profile this table" | cli | | "Import this CSV / Excel into seekdb" / "vectorize a column" | importing | | "Search this collection" / "filter by metadata" / "export search results to Excel" | querying | | "Deploy seekdb primary-standby with obd, do switchover/failover" | OBD-managed SeekDB lifecycle and HA are not included in this tested skill set. Do not substitute the standalone installation or data-plane workflows; explain the limitation and stop before executing OBD lifecycle or HA commands. |
The sub-skills naturally split into two directions — pick by what the user wants to do:
install, build) — get SeekDB running on a machine, or produce a build artifact.docs, cli, importing, querying) — interact with a running SeekDB: look up documentation, run SQL, ingest data, search.| Port | Purpose | |------|---------| | 2881 | MySQL protocol | | 2886 | HTTP / obshell |
docs/): <https://github.com/oceanbase/seekdb-ecology-plugins>| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | pass→pass | 11,672 | 7,862 | -33% | 1 | 1 | 0% | 846 | 1,246 | +47% | 0 | 0 | — |
case-13 | pass→pass | 14,476 | 8,479 | -41% | 1 | 1 | 0% | 1,783 | 1,252 | -30% | 0 | 0 | — |
case-14 | fail→pass | 15,417 | 8,226 | -47% | 1 | 1 | 0% | 1,336 | 1,311 | -2% | 0 | 0 | — |
case-01 | pass→pass | 32,699 | 22,467 | -31% | 1 | 1 | 0% | 3,329 | 3,531 | +6% | 0 | 0 | — |
case-02 | fail→pass | 26,778 | 9,348 | -65% | 1 | 1 | 0% | 3,790 | 1,401 | -63% | 0 | 0 | — |
case-03 | pass→pass | 19,480 | 14,404 | -26% | 1 | 1 | 0% | 2,487 | 3,247 | +31% | 0 | 0 | — |
case-04 | fail→fail | 18,101 | 39,682 | +119% | 1 | 1 | 0% | 1,931 | 1,446 | -25% | 0 | 0 | — |
case-16 | pass→pass | 15,211 | 8,515 | -44% | 1 | 1 | 0% | 1,433 | 1,312 | -8% | 0 | 0 | — |
case-05 | fail→pass | 20,063 | 3,953 | -80% | 1 | 1 | 0% | 2,034 | 1,368 | -33% | 0 | 0 | — |
case-06 | pass→pass | 20,087 | 7,858 | -61% | 1 | 1 | 0% | 670 | 1,130 | +69% | 0 | 0 | — |
case-07 | fail→fail | 11,011 | 6,681 | -39% | 1 | 1 | 0% | 1,783 | 1,799 | +1% | 0 | 0 | — |
case-08 | fail→pass | 17,113 | 8,394 | -51% | 1 | 1 | 0% | 1,887 | 1,356 | -28% | 0 | 0 | — |
case-15 | pass→pass | 12,467 | 3,939 | -68% | 1 | 1 | 0% | 1,078 | 1,319 | +22% | 0 | 0 | — |
case-09 | pass→pass | 5,795 | 4,736 | -18% | 1 | 1 | 0% | 770 | 1,489 | +93% | 0 | 0 | — |
case-10 | pass→pass | 17,586 | 9,084 | -48% | 1 | 1 | 0% | 1,923 | 1,292 | -33% | 0 | 0 | — |
case-11 | pass→pass | 25,199 | 3,690 | -85% | 1 | 1 | 0% | 3,130 | 1,369 | -56% | 0 | 0 | — |
case-12 | pass→pass | 13,705 | 6,830 | -50% | 1 | 1 | 0% | 1,117 | 1,532 | +37% | 0 | 0 | — |
case-18 | fail→pass | 9,992 | 4,166 | -58% | 1 | 1 | 0% | 1,351 | 1,361 | +1% | 0 | 0 | — |
case-19 | pass→pass | 9,750 | 8,068 | -17% | 1 | 1 | 0% | 1,377 | 1,271 | -8% | 0 | 0 | — |
case-20 | fail→pass | 26,066 | 11,901 | -54% | 1 | 1 | 0% | 3,237 | 1,790 | -45% | 0 | 0 | — |
case-21 | pass→pass | 20,020 | 7,608 | -62% | 1 | 1 | 0% | 2,215 | 1,811 | -18% | 0 | 0 | — |
case-22 | fail→fail | 18,768 | 16,064 | -14% | 1 | 1 | 0% | 1,993 | 2,717 | +36% | 0 | 0 | — |
case-23 | fail→pass | 14,935 | 10,648 | -29% | 1 | 1 | 0% | 1,258 | 1,365 | +9% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +30 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 9/7/2026 | +48% |
| gemini-3.6-flash | verified | 8/6/2026 | +82% |
Other measured skills in the registry, with their headline benchmark lift.