Package com.tnsai.rag

Class ContentExtractorRegistry

java.lang.Object

com.tnsai.rag.ContentExtractorRegistry

public final class ContentExtractorRegistry
extends Object

Deterministic registry of installed ContentExtractor providers.

Provider order never decides a collision: providers and duplicate names are sorted by implementation class name. A framework default yields any format a consumer extractor also claims; two non-default claims for one format still fail construction. Missing formats fail when requested.

Since:

0.13.0

Field Summary

Fields

Modifier and Type	Field	Description
static final long	DEFAULT_EXTRACT_TIMEOUT_MILLIS	Default wall-clock bound on one provider call.
static final String	EXTRACT_TIMEOUT_ATTRIBUTE	ExtractionContext.attributes() key for a per-call timeout.

Method Summary

Modifier and Type	Method	Description
static ContentExtractorRegistry	discover()	Discovers providers visible to the core contract's defining classloader.
ExtractedDocument	extract(DocumentInput input, ExtractionContext context)	Resolves and invokes the extractor for one detected document.
String	extractorId(DocumentFormat format)	Returns the registered extractor implementation name for a concrete format.
static ContentExtractorRegistry	of(Collection<? extends ContentExtractor> providers)	Creates an explicit registry, primarily for application registration and deterministic contract tests.
Set<DocumentFormat>	registeredFormats()	Returns the concrete formats registered in this immutable snapshot.
String	version(DocumentFormat format)	Returns the registered extractor version for a concrete format.
String	version(DocumentInput input, ExtractionContext context)	Returns the registered provider version for a concrete input format.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

DEFAULT_EXTRACT_TIMEOUT_MILLIS

public static final long DEFAULT_EXTRACT_TIMEOUT_MILLIS

Default wall-clock bound on one provider call.

See Also:

• Constant Field Values

EXTRACT_TIMEOUT_ATTRIBUTE

public static final String EXTRACT_TIMEOUT_ATTRIBUTE

ExtractionContext.attributes() key for a per-call timeout. Value must be a positive Number of milliseconds.

See Also:

• Constant Field Values

Method Details

discover

public static ContentExtractorRegistry discover()

Discovers providers visible to the core contract's defining classloader. Provider construction and contract-inspection failures use a framework-controlled message while retaining the original provider throwable in the cause chain.

Returns:

immutable deterministic registry

of

public static ContentExtractorRegistry of(Collection<? extends ContentExtractor> providers)

Creates an explicit registry, primarily for application registration and deterministic contract tests.

Parameters:

providers - extractor providers

Returns:

immutable deterministic registry

extract

public ExtractedDocument extract(DocumentInput input,
 ExtractionContext context)

Resolves and invokes the extractor for one detected document.

The provider runs on a worker whose Future is retained. Timeout or caller interrupt cancel(true) that worker. Admission is released exactly once when the worker honours interrupt; a provider that ignores interrupt is quarantined after a bounded grace so ingest is not starved.

Parameters:

input - detected document input

context - source declaration context

Returns:

normalized extracted document

version

public String version(DocumentInput input,
 ExtractionContext context)

Returns the registered provider version for a concrete input format.

extractorId

public String extractorId(DocumentFormat format)

Returns the registered extractor implementation name for a concrete format.

Parameters:

format - detected or declared concrete format

Returns:

implementation class name used in index fingerprints

version

public String version(DocumentFormat format)

Returns the registered extractor version for a concrete format.

Parameters:

format - detected or declared concrete format

Returns:

provider version used in index fingerprints

registeredFormats

public Set<DocumentFormat> registeredFormats()

Returns the concrete formats registered in this immutable snapshot.