Enough local data to see real states instead of empty screens.
Governance console
Platform preview
This page explains what the current SkeinRank console is for and how the main workspaces fit together. The preview is organized around one practical loop: govern terminology, bind it to a search context, verify it with evidence, publish a snapshot, and test the runtime query behavior.
The console shows which terminology version is safe for a concrete search surface.
Alias changes are not just stored; they can be inspected before runtime use.
The dashboard is the starting point: it shows whether the seeded profile, binding, job, snapshot, and runtime dependencies are ready.
How to read the preview
Section titled “How to read the preview”Who it is for
Search platform engineers, ML/RAG engineers, support tooling owners, and reviewers who need a visible workflow for domain terminology instead of scattered synonym files.
What it proves
SkeinRank is not only a canonicalization library. The platform preview demonstrates lifecycle: discover, review, bind, enrich, snapshot, and test terminology in a runtime context.
Where to start
Use the dashboard first, then move through Terminology, Bindings, Evidence, Snapshots, and Search Playground as a single operator workflow.
Workflow at a glance
Section titled “Workflow at a glance”Govern
Create canonical terms, attach aliases, assign slots, and keep noisy values out of the profile.
Bind
Connect a profile to an index or alias, choose fields, and make the runtime search context explicit.
Verify
Use evidence, dry-runs, and job state to check whether terminology matches real documents.
Publish
Pin an immutable snapshot so query-time APIs use a stable and auditable version.
Test
Run queries such as k8s pg timeout and inspect canonicalization plus search hits.
Runtime rule:Profile defines terminology. Binding defines where and how that terminology is applied. Snapshot defines the immutable version served to runtime search.
Dashboard
Section titled “Dashboard”The dashboard is the control center for the local platform preview. It makes the current setup state visible before a downstream search UI, backend, or agent depends on the output.
Check API, Elasticsearch, RabbitMQ, Celery, seeded data, and runtime prerequisites.
See profile, aliases, binding health, active snapshot, enrichment jobs, and next actions together.
Jump from a missing setup step to the workspace that fixes it instead of guessing what failed.
Terminology
Section titled “Terminology”The terminology workspace is where domain language becomes governed product data. It keeps canonical values, aliases, slots, and profile-level state visible instead of hiding them in code, CSV files, or Elasticsearch synonym lists.
Create the stable terms that downstream canonicalization and search workflows should use.
Keep variants such as k8s, kube, or pg attached to a clear schema role.
Give reviewers and integration owners a readable place to understand what will enter runtime.
Bindings
Section titled “Bindings”Bindings are the production runtime object. A profile says what terminology means; a binding says where that terminology is applied, which fields are read, where enriched output is written, and which snapshot the runtime should trust.
Attach a profile to an Elasticsearch index or alias and configure the target and source text fields.
Run dry-runs, limit documents, inspect jobs, and avoid writing runtime output before the binding is ready.
Explain how profile, binding, job, and snapshot are connected for a specific search surface.
Evidence review
Section titled “Evidence review”SkeinRank should not automatically turn every proposed synonym into production behavior. Evidence review gives operators a place to inspect where a candidate alias appears, decide whether it is meaningful, and keep noisy terms out of runtime search.
Review suggested aliases before accepting terminology changes into the governed profile.
Use snippets from real indexed documents to validate whether an alias matches the intended domain context.
Close the loop between contributors, moderators, and the runtime dictionary before publishing.
Snapshots
Section titled “Snapshots”Snapshots are the safe handoff from governance to runtime. They make it clear which terminology version is active, whether a binding is stale, and what downstream search behavior should be considered production-ready.
Pin a stable terminology release instead of serving whatever happens to be edited most recently.
See which bindings are active, stale, pending, or waiting for a new runtime snapshot.
Connect enrichment jobs and query-time behavior to a version that can be inspected or rolled back.
Search Playground
Section titled “Search Playground”The Search Playground is the operator-facing proof step. It shows how a query becomes governed runtime context through the selected binding and its pinned snapshot before you wire SkeinRank into an external backend, UI, or agent workflow.
Preview how k8s pg timeout can become a normalized query with matched aliases.
Inspect the selected profile, binding, and snapshot used for query-time behavior.
Validate whether terminology changes improve recall without creating obvious search noise.
Run the seeded preview
Section titled “Run the seeded preview”Local demo
Start the stack and load the platform operations dataset.
The seeded demo gives the console realistic terminology, bindings, evidence, snapshots, and search behavior so the overview above maps to screens you can run locally.
docker compose -f docker-compose.dev.yml up —build -d
make demo-reset