Skip to content
context-engine
Tools

extract

Read a selected view from registered Context Engine handles.

Use extract after another Context Engine tool returns handles and its API or context is insufficient.

Use when

  • You need implementation, documentation, imports, or a complete selected item.
  • You want resolved inline type annotations for a handle.

Do not use when

  • You need an exact physical line range; use smart_read.
  • You have only a file path or symbol name and no handle.

Request

{"handles":["0004r00o"],"view":"code","output_mode":"type_annotated"}

Parameters

FieldRequiredDefault / bounds
handlesno[]; handles come from Context Engine results.
viewnofull; union: full, code, docs, module-docs, imports.
output_modenotype_annotated; union: type_annotated, plain_text, both.

What it returns

The selected source view for each handle. code reads declarations and implementations; docs reads documentation; full combines the complete item or file; imports reads imports/directives; module-docs reads module-level documentation. Type-annotated output falls back to plain text with a note when LSP inlay hints are unavailable.

Common mistakes

  • Guessing handles instead of copying registered handles from a result.
  • Asking for full when code or docs is sufficient and costly output is unnecessary.
  • Reusing a handle after a rename, move, parent-chain, or signature change; obtain a fresh handle.

grep_definition, outline, jump, line_context