Enum InferenceEvent 

pub enum InferenceEvent {
    Token(String),
    MutedToken(String),
    Thought(String),
    VoiceStatus(String),
    ToolCallStart {
        id: String,
        name: String,
        args: String,
    },
    ToolCallResult {
        id: String,
        name: String,
        output: String,
        is_error: bool,
    },
    ApprovalRequired {
        id: String,
        name: String,
        display: String,
        diff: Option<String>,
        mutation_label: Option<String>,
        responder: Sender<bool>,
    },
    Done,
    Error(String),
    ProviderStatus {
        state: ProviderRuntimeState,
        summary: String,
    },
    OperatorCheckpoint {
        state: OperatorCheckpointState,
        summary: String,
    },
    RecoveryRecipe {
        summary: String,
    },
    McpStatus {
        state: McpRuntimeState,
        summary: String,
    },
    CompactionPressure {
        estimated_tokens: usize,
        threshold_tokens: usize,
        percent: u8,
    },
    PromptPressure {
        estimated_input_tokens: usize,
        reserved_output_tokens: usize,
        estimated_total_tokens: usize,
        context_length: usize,
        percent: u8,
    },
    TaskProgress {
        id: String,
        label: String,
        progress: u8,
    },
    UsageUpdate(TokenUsage),
    RuntimeProfile {
        model_id: String,
        context_length: usize,
    },
    VeinStatus {
        file_count: usize,
        embedded_count: usize,
        docs_only: bool,
    },
    VeinContext {
        paths: Vec<String>,
    },
    SoulReroll {
        species: String,
        rarity: String,
        shiny: bool,
        personality: String,
    },
    CopyDiveInCommand(String),
    EmbedProfile {
        model_id: Option<String>,
    },
    ShellLine(String),
}

Variants

Token(String)

A text token to append to the current assistant message.

MutedToken(String)

A text token to be displayed on screen but NOT spoken (e.g. startup greeting).

Thought(String)

Internal model reasoning (shown in side panel, not dialogue).

VoiceStatus(String)

Critical diagnostic feedback from the voice synthesis engine.

ToolCallStart

A tool call is starting – show a status line in the TUI.

Fields

id: String

name: String

args: String

ToolCallResult

A tool call completed – show result in the TUI.

Fields

id: String

name: String

output: String

is_error: bool

ApprovalRequired

A risky tool requires explicit user approval. The TUI must send true (approved) or false (rejected) via responder. When diff is Some, the modal renders a coloured before/after diff preview.

Fields

id: String

name: String

display: String

diff: Option<String>

Pre-formatted diff: lines starting with “- “ are removals, “+ “ are additions, “—” is a file header. None means a plain high-risk approval (no diff).

mutation_label: Option<String>

Intent label for mutation protocol (Cyan box trigger).

responder: Sender<bool>

Done

The current agent turn is complete.

Error(String)

An error occurred during inference.

ProviderStatus

Compact provider/runtime state for the operator surface.

Fields

state: ProviderRuntimeState

summary: String

OperatorCheckpoint

Typed operator checkpoint/blocker state for SPECULAR and recovery UIs.

Fields

state: OperatorCheckpointState

summary: String

RecoveryRecipe

Typed recovery recipe summary for operator/debug surfaces.

Fields

summary: String

McpStatus

Compact MCP/runtime server health for the operator surface.

Fields

state: McpRuntimeState

summary: String

CompactionPressure

Current compaction pressure against the adaptive threshold.

Fields

estimated_tokens: usize

threshold_tokens: usize

percent: u8

PromptPressure

Current total prompt-budget pressure against the live context window.

Fields

estimated_input_tokens: usize

reserved_output_tokens: usize

estimated_total_tokens: usize

context_length: usize

percent: u8

TaskProgress

A generic task progress update (e.g. for single-agent tool execution).

Fields

id: String

label: String

progress: u8

UsageUpdate(TokenUsage)

Real-time token usage update from the API.

RuntimeProfile

The current runtime profile detected from LM Studio.

Fields

model_id: String

context_length: usize

VeinStatus

Vein index status after each incremental re-index.

Fields

file_count: usize

embedded_count: usize

docs_only: bool

VeinContext

File paths the Vein surfaced as relevant to the current turn. Used to populate ACTIVE CONTEXT with retrieval results.

Fields

paths: Vec<String>

SoulReroll

A new companion was hatched mid-session via /reroll.

Fields

species: String

rarity: String

shiny: bool

personality: String

CopyDiveInCommand(String)

A “Dive-In” command (cd

&& hematite) to be copied to the clipboard.

EmbedProfile

Embed model loaded/unloaded mid-session.

Fields

model_id: Option<String>

ShellLine(String)

A single line of live shell output, streamed while the command runs. Displayed in the SPECULAR panel so the operator sees progress without waiting for the full command to finish.

Trait Implementations

impl Debug for InferenceEvent

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.