Crate serdes_ai_core 

serdes-ai-core

Core types, messages, and error handling for the serdes-ai framework.

This crate provides the foundational types used throughout the serdes-ai ecosystem:

• Messages: Request/response message types for LLM interactions
• Errors: Comprehensive error types with context
• Usage: Token usage tracking and limits
• Settings: Model configuration options
• Identifiers: Type-safe IDs for conversations, messages, runs

Feature Flags

• tracing-integration: Enable tracing instrumentation
• otel: Enable OpenTelemetry integration
• full: Enable all optional features

Example

use serdes_ai_core::{
    messages::{ModelRequest, ModelResponse, UserContent},
    usage::{RequestUsage, RunUsage, UsageLimits},
    settings::ModelSettings,
    identifier::{generate_run_id, now_utc},
};

// Build a request
let mut request = ModelRequest::new();
request.add_system_prompt("You are a helpful assistant.");
request.add_user_prompt("Hello!");

// Configure settings
let settings = ModelSettings::new()
    .max_tokens(1000)
    .temperature(0.7);

// Track usage
let mut usage = RunUsage::new();
usage.add_request(RequestUsage::with_tokens(100, 50));

// Check limits
let limits = UsageLimits::new().max_total_tokens(10000);
limits.check(&usage).expect("Within limits");

Re-exports

pub use errors::Result;

pub use errors::SerdesAiError;

pub use format::format_as_xml;

pub use format::format_as_xml_with_options;

pub use format::XmlFormatError;

pub use format::XmlFormatOptions;

pub use identifier::now_utc;

pub use identifier::ConversationId;

pub use identifier::RunId;

pub use identifier::ToolCallId;

pub use messages::BinaryContent;

pub use messages::BuiltinToolCallPart;

pub use messages::BuiltinToolReturnContent;

pub use messages::BuiltinToolReturnPart;

pub use messages::CodeExecutionResult;

pub use messages::FilePart;

pub use messages::FileSearchResult;

pub use messages::FileSearchResults;

pub use messages::FinishReason;

pub use messages::ModelRequest;

pub use messages::ModelRequestPart;

pub use messages::ModelResponse;

pub use messages::ModelResponsePart;

pub use messages::ModelResponsePartDelta;

pub use messages::ModelResponseStreamEvent;

pub use messages::PartDeltaEvent;

pub use messages::PartEndEvent;

pub use messages::PartStartEvent;

pub use messages::SystemPromptPart;

pub use messages::TextPart;

pub use messages::ThinkingPart;

pub use messages::ToolCallPart;

pub use messages::ToolReturnPart;

pub use messages::UserContent;

pub use messages::UserContentPart;

pub use messages::UserPromptPart;

pub use messages::WebSearchResult;

pub use messages::WebSearchResults;

pub use settings::ModelSettings;

pub use usage::RequestUsage;

pub use usage::RunUsage;

pub use usage::UsageLimits;

Modules

errors

Error types for serdes-ai.

format

Format data as XML for LLM prompts.

identifier

ID generation utilities.

messages

Message types for model interactions.

prelude

Prelude module for common imports.

settings

Model settings and configuration.

usage

Token usage tracking for model requests.