list-negative-parser-is-fully-resilient
IN derived (depth 1)
The list-negative LLM response parser handles all degradation levels: regex extraction recovers JSON objects from prose-laden responses, and completely unparseable output returns zero results gracefully rather than raising exceptions.
Summary
The parser that handles LLM classification responses is robust against all forms of messy output. It can pull structured data out of responses cluttered with extra text, and when the output is completely garbled, it simply reports no results instead of crashing — so the system never fails just because the LLM gave a sloppy response.
Justifications
SL — Two independent parser robustness properties combine to cover the full failure spectrum from partial corruption to total garbling
Antecedents (all must be IN):
- list-negative-json-parser-tolerates-prose-preamble — The `list_negative` LLM classification response parser uses `re.finditer` to extract JSON objects from responses that include prose preamble, handling the common LLM pattern of prefacing structured output with natural language rather than requiring clean JSON.
- api-list-negative-graceful-on-malformed-llm — When the LLM returns unparseable output, `list_negative()` returns `count == 0` rather than raising an exception — graceful degradation over failure.
Dependents
These beliefs depend on this one:
- format-resilience-spans-all-external-interfaces — The system tolerates format variation at every external interface — derive output parsers support version fallback, import parsers silently skip unknown fields, storage handles schema evolution — and extends this resilience to LLM response parsing, where the list-negative parser uses regex extraction to recover structured data from prose-laden responses.