Crate flashmind_types 

Shared types for the Flashmind AI agent framework.

This crate defines the traits, wire types, and event structures shared by all Flashmind crates. It is the foundation that both library consumers and provider implementations depend on.

Core traits

Trait	Purpose
LlmProvider	Streaming completions from any LLM backend
Tool	Function-calling tools the agent can invoke
MemoryProvider	Store/search/forget long-term memories

Modules

Module	Types
event	AgentEvent, AgentInput, TurnStatus, TurnUsage — telemetry
llm	LlmProvider, CompletionRequest, StreamEvent, FinishReason
memory	MemoryEntry, MemoryMetadata, MemoryProvider
message	Message, Role, ToolCall, ContentPart — LLM wire format
model	Model, Provider, SamplingParams, ReasoningLevel
tool	Tool, ToolRegistry, ToolContext, ToolResult
stream	AgentStream, Outcome — stream-with-final-result pattern
utils	UTF-8-safe string truncation helpers

Example: implementing a provider

use async_trait::async_trait;
use flashmind_types::{
    CompletionRequest, CompletionStream, LlmProvider, Provider, StreamEvent, FinishReason,
};

struct MyProvider;

#[async_trait]
impl LlmProvider for MyProvider {
    fn name(&self) -> &str { "my-provider" }
    fn provider(&self) -> Provider { Provider::OpenAi }

    fn complete(&self, _request: CompletionRequest) -> CompletionStream {
        Box::pin(async_stream::stream! {
            yield Ok(StreamEvent::ContentDelta("Hello!".into()));
            yield Ok(StreamEvent::Finished(FinishReason::Stop));
        })
    }
}

Re-exports

pub use error::LlmError;

pub use error::LlmErrorKind;

pub use error::ParseError;

pub use event::AgentEvent;

pub use event::AgentInput;

pub use event::CompactionReason;

pub use event::Source;

pub use event::TurnResult;

pub use event::TurnStatus;

pub use event::TurnUsage;

pub use interaction_mode::InteractionMode;

pub use llm::AudioFormat;

pub use llm::AudioOutputConfig;

pub use llm::CompletionRequest;

pub use llm::CompletionResponse;

pub use llm::CompletionStream;

pub use llm::FinishReason;

pub use llm::FrameImage;

pub use llm::ImageGenConfig;

pub use llm::ImageGenRequest;

pub use llm::LlmProvider;

pub use llm::Modality;

pub use llm::ModelCapabilities;

pub use llm::ModelCategory;

pub use llm::ModelInfo;

pub use llm::ModelPricing;

pub use llm::ProviderRegistry;

pub use llm::StreamEvent;

pub use llm::SttRequest;

pub use llm::TokenUsage;

pub use llm::ToolDefinition;

pub use llm::TtsRequest;

pub use llm::VideoGenRequest;

pub use llm::Voice;

pub use llm::mime_from_extension;

pub use memory::MemoryEntry;

pub use memory::MemoryMetadata;

pub use memory::MemoryProvider;

pub use message::ContentPart;

pub use message::Message;

pub use message::Role;

pub use message::ToolCall;

pub use message::ToolResult;

pub use model::AgentLlmConfig;

pub use model::AliasedModel;

pub use model::Model;

pub use model::Provider;

pub use model::ProviderPreferences;

pub use model::ProviderSort;

pub use model::Quantization;

pub use model::ReasoningLevel;

pub use model::SamplingParams;

pub use session::SessionId;

pub use session::SessionIdError;

pub use session::SessionUsage;

pub use stream::AgentStream;

pub use stream::Outcome;

pub use tool::ForbiddenCmd;

pub use tool::Tool;

pub use tool::ToolContext;

pub use tool::ToolRegistry;

Modules

error

Error types for the Flashmind framework.

event

Telemetry and control events for the agent turn loop.

humanize

Human-readable formatting for numbers, token counts, and pricing.

interaction_mode

llm

LLM wire types — request / response structures and the LlmProvider trait.

memory

Memory abstraction — MemoryProvider trait and entry types.

message

LLM wire-format types — Role, Message, ToolCall, and ContentPart.

model

Model identifiers, provider variants, and sampling parameters.

session

Session identities and cumulative request usage.

stream

AgentStream — a stream that yields events and terminates with a result.

tool

Tool trait, registry, context, and result types for agent function calling.

utils

Shared utility functions for safe string manipulation.

Constants

MIN_SECRET_REDACT_LEN

Minimum length for a secret value to be redacted from output.