Patterns

URL -> MarkdownDocument

Fetch content prioritizing native markdown HTTP headers, appending .md to target paths, and falling back to HTML-to-markdown DOM extraction.

URL -> TargetMarkdown

Traverse upward through a URL's path hierarchy recursively to locate and fetch the nearest context-providing llms.txt file.

Url -> MarkdownText

Fetch a target URL and convert its parsed HTML structure into clean Markdown for LLM consumption.

CatalogRequest -> CatalogResponse

Route and resolve metadata operations across multiple external catalogs (like Hive Metastore and AWS Glue) through a single unified client API.

BlockchainEventStream -> ElasticsearchClient

Stream block and transaction payloads from a node and write them as indexable search documents to an external search cluster.

UnifiedMetadataModel -> EngineSchemaDefinition

Expose dynamically cataloged external datasets to distributed SQL engines via virtual connector plugins.

LogicalQueryPlan -> VectorizedRowBatch

Delegate sub-plan SQL execution from a distributed JVM coordinator to an optimized C++ vectorized execution engine.

AssetURI -> LocalFilePath

Download and cache 3D robot meshes and physical environment descriptions from a remote repository upon first environment instantiation.

Query -> Stream<TimeSeries>

Fan out a metric query to multiple distributed Store API endpoints in parallel and merge the returned partial time-series results.

LakehouseTablePointer -> ColumnarChunk

Query remote cloud storage catalog formats directly by mapping their metadata partitions directly to memory-aligned execution columns.

SpanBatch -> StorageWriteConfirmation

Delegate database storage and retrieval operations to decoupled external driver processes over a gRPC interface.

URL -> LLMOptimizedMarkdown

Fetch a web page and strip raw HTML into structured, LLM-optimized Markdown.

RemoteDatasetURI -> Generator<DataRecord>

Iterate sequentially over a remote multi-file dataset by fetching, buffering, and yielding records from shards on-the-fly without a full disk download.

URL -> MarkdownText

Render a Javascript-heavy target webpage using a headless browser, extract the populated HTML DOM, and parse it into structured Markdown.