@pdpp/read-core

@pdpp/read-core contains pure, adapter-agnostic primitives for shaping PDPP read results into bounded previews, continuation descriptors, handle metadata, and deliberate escalation paths.

It is the shared core consumed by independently-installed read adapters (@pdpp/mcp-server, @pdpp/cli, and future REST/SDK adapters). Given records, record-sets, or field reads, it returns structured, bounded descriptors that any adapter can render — the adapter decides how to present them.

Boundary (what belongs here, and what does not)

In scope — pure read shaping only:

• Bounded previews of records and record-sets.
• Content ladders and field-window evidence (progressive disclosure).
• Continuation descriptors and opaque content handles (encode/decode).
• Binary/blob field metadata.
• Deterministic truncation and stable inline JSON.
• Stable record identity for evidence.

Out of scope — keep these in the adapter or server layers:

• No authorization, grants, or token semantics.
• No HTTP, transport, or networking.
• No filesystem or export I/O.
• No CLI parsing or flag handling.
• No UI rendering beyond minimal stable descriptors.
• No connector-specific or source-specific business semantics.
• No single-client quirks baked in as core semantics (client workarounds belong in the adapter that needs them).

The package keeps a deliberately small, stable public API so adapters can build against it safely.