Expand description
Error recovery system for AI Agents framework
Structs§
- Backoff
Config - Backoff timing configuration used between retry attempts.
- ByRole
Filter - Classified
Error - Error
Recovery Config - Top-level error recovery configuration for an agent. Controls retry, LLM failure handling, tool failure handling, and parse error handling.
- Keep
Recent Filter - LLMRecovery
Config - LLM-specific recovery settings: what to do on total failure, rate limits, or context overflow.
- Parsing
Recovery Config - Recovery settings for malformed LLM output (invalid JSON, bad tool call format).
- Recovery
Manager - Retry
Config - Retry policy: how many times to retry and with what backoff strategy. Applied to both LLM calls and tool calls unless overridden per-tool.
- Skip
Pattern Filter - Tool
Recovery Config - Tool-level recovery: a default policy plus optional per-tool overrides.
- Tool
Retry Config - Retry and failure policy for a single tool (or the default for all tools).
Enums§
- Backoff
Type - Strategy for computing the wait duration between successive retry attempts.
- Context
Overflow Action - Action to take when the accumulated message history exceeds max_context_tokens.
- Error
Type - Classified error kinds used to decide whether a failure should be retried.
- Filter
Config - Selects which messages are passed to the summarizer during context compression.
- LLMFailure
Action - Action to take when the primary LLM fails after all retries are exhausted.
- Parse
Error Action - Action to take when the LLM response cannot be parsed as expected.
- Rate
Limit Action - Action to take when the LLM returns a rate-limit error.
- Recovery
Error - Tool
Failure Action - Action to take when a tool call fails after all retries are exhausted.