Guides
Layer CLI
The layer CLI ships with the Python SDK. It covers workflows that kubectl
and the SDK don’t: dispatching UDFs against existing namespaces, applying a
spec across a namespace prefix, and producing an agent-readable bundle of the
hev layer and Turbopuffer docs.
For query, scan, and warm-cache work, use the Python SDK. For raw CRDs, use
kubectl. For cluster and pipeline visibility, use the dashboard.
Install
pip install hevlayer
export LAYER_GATEWAY_URL=https://aws-us-east-1.hevlayer.com
export LAYER_GATEWAY_API_KEY=...
See Install for SDK install details and base-URL defaults.
| Variable | Purpose |
|---|---|
LAYER_GATEWAY_URL | Base URL of the gateway. |
LAYER_GATEWAY_API_KEY | API key sent on every request. |
Run a UDF
layer run -f submits a Function spec to the gateway and runs
it against the named index. The gateway resolves defaults (image registry,
compute pool, cache rules) the YAML can omit.
layer run -f tag-products.yaml --index products
The command blocks until the run reaches a terminal state and exits non-zero
on failure. Use --detach to submit and return immediately.
A full build-and-run loop — Depot build, ECR push, Function apply — is
roadmapped as layer push; see the roadmap.
Initialize a namespace
layer init runs a one-shot UDF against an existing namespace to populate
attributes, backfill embeddings, or migrate document shape. It is shorthand
for layer run -f against a built-in initializer Function, parameterized by
flags.
layer init products --embed text --target embedding
The namespace must already exist; layer init does not create indexes.
Apply across a namespace prefix
layer apply -f applies a spec across every namespace matching a
prefix selector, which would otherwise be a shell loop over kubectl apply.
layer apply -f infrarules.yaml --namespaces 'products-*'
Use --dry-run to print the resolved target list without applying.
Agent-friendly docs
layer docs prints a single markdown bundle of the hev layer docs plus the
Turbopuffer reference to stdout. Pipe it into an agent or save it locally:
layer docs > hevlayer.md
layer docs | pbcopy
The bundle follows the llms.txt convention. hev layer publishes its docs at two URLs:
| URL | Contents |
|---|---|
https://hevlayer.com/llms.txt | Navigation index — titles, descriptions, and links per page. |
https://hevlayer.com/llms-full.txt | Every docs page concatenated as one markdown file. |
layer docs fetches llms-full.txt from the gateway-configured docs origin
and from Turbopuffer’s published llms.txt, then concatenates the two. Pass
--source hevlayer or --source turbopuffer to fetch just one half.