Crate atomr_agents_tool 

Tools, tool sets, registry, and the tool-call parser.

Structs

EvidenceTrace

The evidence + rationale behind a tool’s decision.

ExplainedTool

Middleware enforcing an ExplainabilityPolicy on any RichTool. Wraps without touching the inner tool’s constructor (cf. WalledTool).

HandoffTool

KeywordToolStrategy

v1: lexical filter. Returns tools whose name or description contains any of the keywords found in the user’s turn input. (Substring match in v0; promote to TF-IDF later.)

MissingEvidence

Raised when ExplainabilityPolicy::Require is violated.

ParsedToolCall

PermissionSpec

StaticToolStrategy

v0: hand-picked fixed list of tools.

ToolCallParser

Stateful streaming parser. Feed each tool_call_delta value as it arrives; call finish to drain the accumulated calls.

ToolDescriptor

Descriptor advertised to the model. The schema is JSON-Schema shaped; per-provider renderers in this crate adapt it to OpenAI / Anthropic / Gemini call formats.

ToolSchema

JSON-Schema fragment for the tool’s arguments.

ToolSet

Versioned bundle. Lives behind an Arc in the registry.

ToolSetMeta

Free-form metadata attached to a tool set.

ToolSetRegistry

Thread-safe registry of versioned tool sets. Multiple versions of the same id can coexist; consumers pin the version they want.

Enums

EvidenceRef

A reference to a piece of evidence behind a decision.

ExplainabilityPolicy

Per-tool/desk explainability requirement.

Provider

ToolControl

ToolReturn

What a richer tool returns. Agents map this back into the message sequence and (optionally) the workflow state.

Constants

EVIDENCE_KEY

Reserved artifact key under which the evidence trace is stored.

Traits

RichTool

Tool

A tool the agent may call. Tools are also Callable, so they can stand in as workflow steps.

Type Aliases

DynTool

Arc<dyn Tool> — what the registry stores.