Module model 

Core data models shared by agents, tools, and model adapters.

The types in this module form the data contract between Anda runtimes, model providers, agents, tools, and clients. They cover:

• agent and tool inputs/outputs (AgentInput, AgentOutput, ToolInput, ToolOutput);
• chat messages and multimodal content (Message, ContentPart);
• function-call metadata (FunctionDefinition, ToolCall);
• request metadata and usage accounting (RequestMeta, Usage);
• prompt documents and completion requests (Document, Documents, CompletionRequest).

Structs

AgentInput

Request sent to an agent for processing.

AgentOutput

Output produced by an agent execution.

ByteArrayB64

Wrapper around [u8; N] to serialize and deserialize efficiently. If the serialization format is human readable (formats like JSON and YAML), it will be encoded in Base64URL. Otherwise, it will be serialized as a byte array.

ByteBufB64

Wrapper around Vec<u8> to serialize and deserialize efficiently. If the serialization format is human readable (formats like JSON and YAML), it will be encoded in Base64URL. Otherwise, it will be serialized as a byte array.

CompletionRequest

Provider-neutral completion request.

Document

A document with metadata and content.

Documents

Collection of documents that can be injected into a completion prompt.

Function

Represents a function definition with its metadata.

FunctionDefinition

Defines a callable function with its metadata and schema.

Message

Chat message sent to or returned by an LLM provider.

PartialAgentOutput

RequestMeta

Metadata attached to an agent or tool request.

Resource

Represents a resource for AI Agents.

ResourceRef

Borrowed view of a Resource suitable for serialization.

ToolCall

Tool call requested by an LLM or returned by a tool execution pipeline.

ToolInput

Request sent to a tool for processing.

ToolOutput

Output produced by a tool execution.

Usage

Usage statistics for an agent, model, or tool execution.

Xid

Represents a unique identifier with 12 bytes. Based on the xid. See: https://github.com/rs/xid

Enums

ContentPart

A single content item inside a chat message.

ModelEffort

Provider-agnostic reasoning/thinking effort requested for a completion.

PromptCommand

Traits

CompletionFeatures

LLM completion capability exposed by an agent context.

Functions

estimate_tokens

Estimates token count using a small, provider-independent heuristic.

inline_data_from_data_url

Parses a data URL string and extracts the inline data and MIME type, if applicable.

part_to_data_url

Converts a content part with inline data to a data URL string.

platform_text_encoding

Returns the platform-local text encoding used for legacy text files.

prompt_with_resources

select_resources

Removes and returns resources matching any of the supported tags.

text_encoding_for_label

Resolves a text encoding label, accepting both utf8 and standard Encoding Standard labels.

text_encoding_label

Returns the normalized label used by Anda for a decoded text encoding.

text_from

Attempts to decode the given byte vector as text and checks if it looks like text content.

text_from_bytes

Attempts to decode bytes as text and checks if it looks like text content.

text_from_bytes_with_encoding

Attempts to decode bytes as text using UTF-8 first and an explicit fallback encoding second.

text_resource_documents

update_resources

Updates resource metadata before persistence.

utf8_text_from

Attempts to decode the given byte vector as UTF-8 text and checks if it looks like text content.

utf8_text_from_bytes

Attempts to decode the given byte slice as UTF-8 text and checks if it looks like text content.

windows_code_page_encoding

Maps a Windows code page number to an Encoding Standard decoder when supported.