KNOWLEDGE

Knowledge infrastructure

Connect PDFs, URLs, databases, APIs, GitHub, and Notion. Ingest, index, and retrieve them with a record of what each answer used.

Knowledge shouldn't just exist. It should be understood.

Who this is for

Teams whose answers have to come from a known corpus: policies, tickets, code, product docs, contracts. A vector dump without version or conflict handling will look fine until two documents disagree or a source goes stale.

How the pipeline works

You connect a source. Obliq ingests and normalizes it, indexes it with authority and version, and serves retrieval only to agents that are allowed to see that collection. When an agent answers, the trace shows the sources that were assembled into context.

PipelineSOURCE → RETRIEVE

  1. 01SOURCE

    Register the system of record: a PDF set, URL, database, API, GitHub repo, or Notion workspace.

  2. 02INGEST

    Pull content on a schedule or on change so the index is not a one-time dump.

  3. 03NORMALIZE

    Turn heterogeneous documents into a consistent shape agents can retrieve without custom parsers per source.

  4. 04INDEX

    Store chunks with authority and version so a later answer can point at what was current.

  5. 05RETRIEVE

    Return scoped context to an agent and record which sources were used on that execution.

Health signals

  • Stale sources

    Flag material that has not been refreshed inside the window you set, before an agent treats it as current.

  • Conflicting sources

    Surface two authorities that disagree so you can decide which one an agent is allowed to prefer.

  • Version and authority

    Keep who owns a source and which revision was retrieved, instead of a single undifferentiated blob.

  • Missing information

    See gaps in coverage when a question has no supporting source, rather than letting the model fill them in silently.

  • Confidence

    Attach a retrieval signal to the context an agent actually received on a run.

  • Coverage

    Know which collections an agent can see, and which collections were never connected.

Questions

What is AI knowledge infrastructure?

Knowledge infrastructure is the pipeline that connects source systems, keeps them current, and makes retrieval traceable. In Obliq that pipeline is source, ingest, normalize, index, and retrieve, with health signals for staleness, conflict, and coverage.

Which sources can Obliq connect?

PDFs, URLs, databases, APIs, GitHub, and Notion. The point of the pipeline is that agents retrieve from those sources under scope, and the execution records which source was used.

How does Obliq handle conflicting documents?

Conflicting sources are a health signal, not a silent merge. You can see disagreement and authority so an agent is not left to invent a winner.