feat(llm): add LLM call outcome classifier / state machine#13

Open

GrigoryEvko wants to merge 14 commits into

FusionBrainLab:mainfrom

GrigoryEvko:fix/llm-call-outcome-classifier

Commits on May 15, 2026

feat(llm): add LLM call outcome classifier
GrigoryEvko
committed
feat(llm): audit-driven hardening of the classifier
GrigoryEvko
committed
refactor(llm): drop the arbitrary 24-hour Retry-After cap
GrigoryEvko
committed
feat(bandit): wire classifier into invoke/ainvoke + structured-output
GrigoryEvko
committed
test(bandit): integration-completeness coverage + structured-output process-side hook
GrigoryEvko
committed
fix(llm): log structured-output failure-hook errors at warning level
GrigoryEvko
committed
test(bandit): verify pre-_select failure leaves the ledger clean
GrigoryEvko
committed
style: ruff format after audit commits
GrigoryEvko
committed
fix(core_types): scrub UTF-16 surrogates from StageError text fields
GrigoryEvko
committed
test(bandit): pin no-double-injection + concurrent dispatch invariants
GrigoryEvko
committed
fix(bus): scrub UTF-16 surrogates in MigrantEnvelope.to_stream_fields
GrigoryEvko
committed
test(bandit): agent-layer integration regression locks
GrigoryEvko
committed
fix(token_tracking): defensive coercion in TokenUsage.from_response
GrigoryEvko
committed
fix(llm): raise parsing_error from _process so failure hook fires
GrigoryEvko
committed