Module core 

Core types and traits — available via brainwires::core::* or brainwires::prelude::*.

Modules

content_source

Content source types for tracking where content originates.

embedding

Embedding provider trait for vector operations. Unified embedding abstraction

error

Framework error types and result aliases. Framework error types

graph

Knowledge graph types: entities, edges, and trait interfaces. Graph types and traits for knowledge graph abstraction.

lifecycle

Lifecycle hooks for intercepting framework events. Lifecycle Hooks

message

Message, role, and streaming types for AI conversations.

output_parser

Structured output parsers for LLM responses. Structured output parsing for LLM responses

permission

Permission mode definitions. Permission mode types

plan

Plan metadata, steps, budgets, and serializable plans.

plan_parser

Plan text parser for extracting steps from LLM output. Plan Parser - Extract tasks from plan content

provider

Provider configuration and chat options.

task

Task, priority, and agent response types.

tool

Tool definitions, schemas, contexts, and idempotency.

vector_store

Vector store trait for similarity search. Vector store abstraction

working_set

Working set for file context management with LRU eviction. Working Set for File Context Management

Structs

AgentResponse

Agent response after processing

ChatOptions

Chat completion options

ChatResponse

Response from a chat completion

CommitResult

Result returned by a successful StagingBackend::commit.

GraphEdge

An edge in the relationship graph.

GraphNode

A node in the relationship graph.

IdempotencyRecord

Record of a completed idempotent write operation.

IdempotencyRegistry

Shared registry that deduplicates mutating file-system tool calls within a run.

JsonListParser

Extracts a list of items from a JSON array in the LLM response.

JsonOutputParser

Extracts JSON from LLM responses and deserializes into T.

Message

A message in the conversation

ParsedStep

A parsed step from plan content

PlanBudget

Budget constraints that a serializable plan must satisfy before execution begins. Used in TaskAgentConfig::plan_budget.

PlanMetadata

Metadata for a persisted execution plan

PlanStep

A single step in a serializable pre-execution plan.

RegexOutputParser

Parses LLM output using a regex pattern with named capture groups.

SerializablePlan

A serializable execution plan produced by the agent before any side effects occur. When TaskAgentConfig::plan_budget is set, the agent generates this plan in a separate provider call and validates it against the budget; if the budget is exceeded the run fails immediately with FailureCategory::PlanBudgetExceeded.

StagedWrite

A single write operation that has been staged but not yet committed.

Task

A task being executed by an agent (supports tree structure)

Tool

A tool that can be used by the AI agent

ToolContext

Execution context for a tool.

ToolInputSchema

JSON Schema for tool input

ToolResult

Result of a tool execution

ToolUse

A tool use request from the AI

Usage

Usage statistics for a chat completion

VectorSearchResult

Result from a vector similarity search.

WorkingSet

Manages the set of files currently in the agent’s context

WorkingSetConfig

Working set configuration

WorkingSetEntry

A file entry in the working set

Enums

ContentBlock

Content block for structured messages

ContentSource

Trust level / origin of content injected into an agent’s context.

EdgeType

Types of edges in the relationship graph.

EntityType

Types of entities tracked in the knowledge graph.

FrameworkError

Core framework errors

ImageSource

Image source for image content blocks

MessageContent

Message content can be simple text or complex structured content

PermissionMode

Permission mode for tool execution

PlanStatus

Status of a plan

Role

Role of the message sender

StreamChunk

Streaming chunk from a chat completion

TaskPriority

Task priority levels

TaskStatus

Task status

ToolCaller

Specifies which contexts can invoke a tool. Implements Anthropic’s allowed_callers pattern for programmatic tool calling.

ToolMode

Tool selection mode

Traits

EmbeddingProvider

Trait for text embedding generation.

EntityStoreT

Trait for querying an entity store.

OutputParser

Trait for parsing structured output from LLM text responses.

Provider

Base provider trait for AI providers

RelationshipGraphT

Trait for querying a relationship graph.

StagingBackend

Trait for staging write operations before committing to the filesystem.

VectorStore

Trait for vector database operations.

Functions

estimate_tokens

Estimate tokens for a string (rough: ~4 chars per token)

estimate_tokens_from_size

Estimate tokens for a file by size

parse_plan_steps

Parse plan content into structured steps

serialize_messages_to_stateless_history

Serialize a slice of Messages into the STATELESS protocol format for conversation history.

steps_to_tasks

Convert parsed steps into Task objects

Type Aliases

FrameworkResult

Result type alias using FrameworkError