Streaming Jackson Output Converter
Streaming output converter that extends FilteringJacksonOutputConverter to support JSONL format.
This converter enables streaming LLM responses by:
Converting JSONL (JSON Lines) input to reactive Flux
streams of typed objects Supporting mixed content streams with both objects and thinking content via StreamingEvent
Inheriting JSON schema injection and property filtering capabilities from parent FilteringJacksonOutputConverter
Providing instructions to LLMs for proper JSONL output formatting
Use cases:
Streaming lists of objects from LLM responses in real-time
Processing LLM reasoning (thinking) alongside structured outputs
Real-time agent progress monitoring and incremental results
The converter requests JSONL format from LLMs and parses each line as a separate JSON object, emitting them as reactive stream events as they become available.
Constructors
Functions
Convert streaming JSONL text to a Flux of typed objects. Each line should be a valid JSON object matching the schema. Uses resilient error handling - logs warnings for null conversions but continues processing other lines.
Convert streaming text with thinking blocks into StreamingEvent objects. Supports both object lines and thinking blocks. Uses resilient error handling - logs warnings for individual line failures but continues processing.