lance-format

Solid

Reference for Lance v9 - the open columnar lakehouse format for multimodal AI - and its Rust crate workspace (`lance`, `lance-table`, `lance-file`, `lance-encoding`, `lance-index`, `lance-io`, `lance-namespace`, and more). Use when building directly on the Lance crates - creating or reading `.lance` datasets, manifests, fragments, deletion files, the 2.x file format and structural encodings, vector / scalar / full-text / FM-Index / geo indexes, MemWAL streaming writes, optimistic-concurrency commits and commit handlers, schema evolution, versioning, time-travel, tags, branches, stable row IDs, namespaces, or object-store config. Triggers on lance crate, .lance file, lance dataset, lance file format, structural encoding, IVF_PQ, IVF_HNSW, IVF_RQ, RaBitQ, FM-Index, lance FTS, zonemap, MemWAL, OCC retry, lance schema evolution, lance namespace, pylance. This is the Lance format and engine (the `lance-format/lance` repo), not LanceDB the database product - but also the right reference for what LanceDB builds on.

Code & Development 31 stars 2 forks Updated 3 days ago MIT

Install

View on GitHub

Quality Score: 85/100

Stars 20%
50
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
80
License 10%
100
Description 5%
100

Skill Content

# Lance v9 reference Lance is an open columnar format for multimodal AI - "a columnar data format that is 100x faster than Parquet for random access." It is not one format but a stack of interoperating specs: a **file format**, a **table format**, **index formats**, **catalog specs**, and a **namespace client spec**. The Rust workspace at `lance-format/lance` implements all of them plus Python (`pylance`) and Java bindings. This skill tracks **`v9.1.0-beta.8`** (the `lance-format/lance` git tag), the current development frontier. Pin against tags, not `main` - Lance ships beta tags every few days and `next`-format encodings can change. v9.0.0 never got a final tag; the line reached **`v9.0.0-rc.2`** (2026-07-21) while `main` advanced to the `9.1.0-beta.*` dev series. **`v8.0.0` final shipped 2026-07-01** - if you need a stable pin rather than the v9 dev betas, track `v8.0.0` (the last final), whose format and API are the frozen predecessor of what this reference describes. Three layers of reference, load what the task needs: - `references/lance-reference.md` - the distilled deep reference. Any concrete schema, parameter, proto, or constraint. - `references/performance.md` - ALL official performance guidance combined, plus field-verified remote-storage practices. Load for any performance, tuning, maintenance-cost, or "why is this slow" question. - `references/docs/` - a **verbatim mirror of the official docs** (`docs/src` at the tracked tag): every guide, quickstar...

Details

Author
tenequm
Repository
tenequm/skills
Created
8 months ago
Last Updated
3 days ago
Language
Python
License
MIT

Similar Skills

Semantically similar based on skill content — not just same category

AI & Automation Listed

lorekeeper-dev

Engineering practices for developing the Lorekeeper codebase. Load this skill when working on Lorekeeper source code, fixing bugs, adding features, writing tests, or reviewing PRs. Covers architecture conventions, SQLite/LanceDB/Chroma quirks, testing patterns, and the verification standard for shipped changes. For backlog/ticket workflow, see backlog-management skill.

3 Updated today
Jessinra
AI & Automation Solid

vllm-deployment

Use this skill when authoring, reviewing, or fixing a vLLM Kubernetes manifest, Docker/Podman pod, or OpenShift ServingRuntime — even when the user does not say "vllm". Triggers on: lab cluster performance practices, cache mount + survival across pod restarts (/root/.cache, VLLM_CACHE_ROOT, TORCHINDUCTOR_CACHE_DIR, TRITON_CACHE_DIR, "do we have caches saved"), HF_TOKEN secret in pod env, liveness + readiness probe tuning (initialDelaySeconds, failureThreshold, "pod takes 12 min to boot"), serve_args review, --enforce-eager rationale, MoE deployment ("ep2 dp2", --enable-expert-parallel, expert-parallel sizing), TP/PP sizing, ConfigMap parser-plugin mount, image tag selection, cold-boot reduction, multi-node LWS + Ray, control planes (llm-d, production-stack, AIBrix, NVIDIA Dynamo, KServe), KEDA autoscaling, GAIE routing, disaggregated prefill/decode (Nixl/Mooncake/LMCache/MORI-IO), RHAIIS on OpenShift (SCC, arbitrary UID, Routes 60s, ModelCar, air-gapped). Lead with operator intent, not vendor names.

3 Updated yesterday
air-gapped
AI & Automation Listed

aidex-conventions

NOT auto-invoked. Shared documentation-canon hub for the aidex-* family — holds the .context/ convention references (references/*.md) that the single-purpose sibling skills delegate into. Routing — plan multi-step work → aidex-plan; record a decision/ADR → aidex-decision; capture a stakeholder/client request → aidex-request; investigate/research how something works → aidex-research; document a settled system reference → aidex-reference; defer/park an idea for later → aidex-backlog; capture/draft a communication received or to send → aidex-comm; check a skill against house conventions → aidex-skill. This skill is the canon home, not an entry point; the siblings are the entry points.

1 Updated today
yacb2