Expand description
§synwire-core
Core traits and types — the foundational trait layer for the Synwire AI framework.
This crate provides the foundational trait hierarchy for chat models, embeddings, vector stores, runnables, tools, callbacks, output parsers, retrievers, credentials, and security primitives.
All I/O operations are async-first. Core crate compiles with zero unsafe.
Re-exports§
pub use agents::sampling::NoopSamplingProvider;pub use agents::sampling::SamplingError;pub use agents::sampling::SamplingProvider;pub use agents::sampling::SamplingRequest;pub use agents::sampling::SamplingResponse;pub use state::State;
Modules§
- agents
- Agent runtime traits and types.
- callbacks
- Callback handler traits for observability hooks.
- credentials
- Credential management and secret handling.
- documents
- Document types for retrievable content.
- embeddings
- Embedding traits and types.
- error
- Error types for the Synwire framework.
- language_
models - Language model traits and types.
- loaders
- Document loader traits.
- mcp
- MCP (Model Context Protocol) integration traits and types.
- messages
- Message types and utilities for conversation modelling.
- observability
- Observability stack for the Synwire framework.
- output_
parsers - Output parser traits and implementations.
- prelude
- Convenience re-exports of the most commonly used types and traits.
- prompts
- Prompt template types.
- rerankers
- Reranker traits.
- retrievers
- Retriever traits and types.
- runnables
- Runnable composition types and configuration.
- sandbox
- Sandbox protocols for command execution and approval gates.
- security
- Security primitives including SSRF protection and path traversal guards.
- state
- State trait for agent state management.
- tools
- Tool types, traits, and schemas.
- vectorstores
- Vector store traits and types.
- vfs
- Virtual filesystem abstraction for agent data operations.