MonadicSharp.AITyped error handling for LLM pipelines

Exponential backoff, execution tracing, structured output validation, and streaming — all composable with Result.

AiError

Semantic error types for every LLM failure mode — RateLimit, Timeout, TokenExhausted, ContentFiltered. Each carries retry metadata so callers know exactly what to do.

AiError docs

🔄

RetryResult

Exponential backoff with jitter, built on Result. Terminal errors (ContentFiltered, Unauthorized) short-circuit immediately — no wasted retries.

RetryResult docs

✅

ValidatedResult

JSON parsing and domain validation in a single composable step. Parsing errors and validation failures surface as distinct, typed errors.

ValidatedResult docs

🕵️

AgentResult

Multi-step pipeline tracing with per-step timing, token counts, and error context. Full execution history as a value — no external APM required for development.

AgentResult docs

🌊

StreamResult

Handle streaming completions without try/catch. Mid-stream errors are captured as Result — the partial response is preserved for debugging.

StreamResult docs

🔗

Composable

Every type integrates seamlessly with MonadicSharp pipelines via Bind, Map, and Match. Mix AI operations with your domain logic — no impedance mismatch.