Skip to main content

Application Overview

The VectorAmp dashboard is the operator console for governed retrieval. Use it to create datasets, connect sources, monitor ingestion jobs, inspect search behavior, manage API keys, and keep production knowledge workflows healthy.

Dashboard role

The API is for application traffic. The dashboard is for humans operating the retrieval system: creating keys, checking ingestion health, reviewing datasets, and debugging why an answer did or did not have enough evidence.

What you manage in the dashboard

AreaWhat it controls
DatasetsRetrieval boundaries backed by SABLE: vectors, metadata, source content, and search behavior.
SourcesFile uploads and connected content sources that feed ingestion jobs.
Pipelines / jobsParsing, chunking, embedding, indexing, retries, and processing status.
EmbeddingsModel/provider choices and embedding behavior for new content.
API keysServer-side access for apps, ingestion automation, SDKs, CLI, REST, and MCP.
OrganizationsMulti-tenant account boundary for datasets, users, keys, usage, and configuration.

Core workflow

A typical VectorAmp workflow looks like this:

  1. Create a dataset for one retrieval boundary, such as product docs, policy knowledge, customer files, or a tenant-specific corpus.
  2. Connect or upload sources that should become searchable.
  3. Monitor ingestion jobs while VectorAmp parses, chunks, embeds, and indexes content.
  4. Test search and cited answers with representative questions.
  5. Add filters and metadata so application queries are scoped by tenant, role, source, document type, or time window.
  6. Create API keys for the service or workflow that will query the dataset.
  7. Ship with observability so operators can tell whether retrieval is fresh, scoped, and source-grounded.

Dashboard vs API

Use both surfaces together.

SurfaceBest for
DashboardSetup, inspection, demos, troubleshooting, source/job monitoring, key management.
REST APIBackend integration, search, ingestion automation, dataset/job/source operations.
SDKsApp teams that want language-native helpers.
CLIRepeatable local workflows, smoke tests, and scripted ingestion.
MCPAgent tools that need governed access to enterprise knowledge.