Skip to main content

​
Knowledge & research APIs

The Developer Platform is designed to expose Starfire’s retrieval and research systems as programmatic resources rather than limiting them to the web UI.

​
Knowledge resources

Knowledge APIs are intended to let authorized applications create or work with persistent reference collections. Conceptually, a client can work with operations such as:
  • create or list knowledge bases
  • add files or supported source URLs
  • inspect ingestion/indexing state
  • query a knowledge base
  • retrieve relevant source material
The final route names and schemas remain part of the active Alpha contract.

​
Knowledge queries

A knowledge query is different from a normal model prompt. Its primary job is to retrieve material from an authorized source collection. A query can include concepts such as:
  • query text
  • result limit
  • filters
  • project or organization context
  • source constraints
The application can then decide whether to show those results directly or pass them into a model workflow.

​
Embeddings

The v1 architecture includes an embeddings resource so approved developer workloads can use Starfire’s configured embedding providers without binding directly to a single provider implementation. Embedding availability, dimensions, models, pricing, and limits can vary with platform configuration.

​
Research resources

Research APIs represent a higher-level workflow. A research request can create a long-running job that plans queries, gathers sources, evaluates claims, and returns a report or structured result. A research request can include concepts such as:
  • research question
  • depth
  • source limit
  • project context
  • output/report requirements

​
Research status and sources

Because deeper research is asynchronous, the developer model is designed around a research/run identifier. Clients can inspect run state and, where the active contract supports it, retrieve sources or the final report separately.

​
Citation-aware output

Research integrations should preserve enough source metadata for a consuming application to connect claims to their supporting sources.
The API should not be treated as a guarantee that every retrieved source is correct. Applications that make high-stakes decisions should still inspect source quality and the relationship between the source and the claim.

​
Organization ownership

Knowledge bases and research runs can belong to an organization or project. API authorization must respect the same resource boundaries as the web application.

Knowledge & RAG

Learn the retrieval architecture behind the API.

Search & research

See the human-facing research workflow and citation model.