User guide

Spec Search

Search for a field, type, enum value or default across every CRD of a release at once.

Spec Search (/spec-search) answers questions like "which CRDs have a vlanId field?" or "where is the enum value lacp used?". It searches inside the schemas of every CRD in a release.

#Using it

1. Pick a release and, optionally, a version scope:
   • Latest per CRD (default): the newest version of each CRD;
   • a specific version name;
   • all versions (version=* in the URL).
2. Type a query: a field path (spec.members), part of a name (vlan), a type (integer), an enum value or a default.
3. Tick search descriptions to also match field descriptions (slower, more results).

Results appear while the search runs, grouped by CRD and version, with the matching fields listed. Click a result to open the CRD window with those fields highlighted.

Tip You can leave out spec. or status.: a query for members also matches spec.members.

#Limits

• At most 100 grouped results are shown. Make the query more specific if you hit the limit.
• State CRDs are skipped, because their fields mirror the config CRDs.
• Typing is debounced by 250 ms, so a search starts when you pause.

#How it works

All of the work happens in your browser:

flowchart TB
  Q["Your query"] --> WQ["Work queue: one item per CRD and version"]
  WQ --> ORD["Already-parsed schemas first"]
  ORD --> F["Fetch and parse YAML, 16 at a time"]
  F --> IDX["Build a flat search index per schema"]
  IDX --> PRE{"Index might match?"}
  PRE -->|"no"| SKIP["Skip"]
  PRE -->|"yes"| PRUNE["Prune the schema to matching branches"]
  PRUNE --> GRP["Group by CRD and version, max 100"]
  GRP --> UI["Show results progressively"]

• Warm-up. As soon as you pick a release, schemas start downloading in the background (10 at a time), so the first search is faster.
• Index. Each schema is turned into one string of tokens: field paths, types, enum values, defaults, titles and required/optional markers (plus descriptions if enabled). It matches both spaced and compact forms of your query.
• Pruning. For a match, only the branches of the schema that contain it are kept, which is what you see in the result.
• Responsiveness. The search yields to the browser every few batches so the page stays usable.