Expand description
kernex-core: Foundation types, traits, and error handling for the Kernex runtime.
Re-exports§
pub use config::shellexpand;pub use error::KernexError;pub use guardrails::GuardrailAction;pub use guardrails::GuardrailRunner;pub use guardrails::NoopGuardrailRunner;pub use permissions::PermissionOutcome;pub use permissions::PermissionRules;pub use run::ModelTier;
Modules§
- config
- Configuration types and loading for Kernex.
- context
- Conversation context passed to AI providers.
- error
- Unified error type for all Kernex crates.
- guardrails
- Guardrail trait for intercepting and filtering text at the pipeline layer.
- hooks
- Hook system for intercepting tool execution lifecycle events.
- message
- Request and response types for the Kernex runtime.
- permissions
- Declarative allow/deny permission rules for tool calls.
- pricing
- Per-model token pricing for cost estimation.
- run
- Configuration and outcome types for the agentic runtime loop.
- sanitize
- Input sanitization against prompt injection attacks.
- spawn
- Helpers for spawning subprocesses with skill/MCP-supplied environment maps.
- stream
- Streaming types for SSE-based provider responses.
- traits
- Core traits that define the Kernex runtime contracts.
- utf8
- UTF-8 boundary helpers compatible with the workspace’s MSRV (1.74).