Enum SideEffectClass 

#[non_exhaustive]
pub enum SideEffectClass {
    Pure,
    ReadOnly,
    MutatesLocal,
    MutatesExternal,
    Indeterminate,
}

Side-effect classification — controls whether a tool is safe to run speculatively (i.e. before the LLM asks for it).

Speculative pre-fetch is the killer feature of Paper 3, but it is only safe when re-issuing the call has no observable consequence beyond what the LLM was going to do anyway. Anything that mutates state (local files, remote APIs, user-visible objects) must never be speculated — otherwise we double-execute writes.

The default is the most conservative reading: Indeterminate. New tools are non-speculatable until a provider explicitly opts in.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Pure

Deterministic + idempotent: same input → same output, no state. Safe to speculate freely. Examples: Read of an unchanged file, hash computations, pure functions over args.

ReadOnly

No external mutation, but the result can change between calls (TTL applies). Safe to speculate when freshness_ttl_s has not expired. Examples: get_issues, WebFetch, Glob, Grep, list-style endpoints. The bulk of the planner’s wins.

MutatesLocal

Mutates host-local state (files, in-memory caches). Never speculate — re-running would duplicate the edit. Examples: Edit, Write, MultiEdit, NotebookEdit.

MutatesExternal

Mutates remote state (creates issues, sends messages, runs pipelines, git push). Never speculate — the consequence is visible to other actors. Examples: create_issue, create_merge_request, add_issue_comment, Bash for destructive commands.

Indeterminate

Outcome cannot be classified statically (most prominently Bash — its effect depends on the command string). Default for any tool that has not been annotated. Treated as non-speculatable; the planner only emits a hint to the LLM.

Implementations

impl SideEffectClass

pub fn is_speculatable(&self) -> bool

true iff the planner is allowed to issue this tool ahead of the LLM asking for it. Currently Pure and ReadOnly only; the other variants are bypassed even if enrichment.enabled is on.

Trait Implementations

impl Clone for SideEffectClass

fn clone(&self) -> SideEffectClass

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SideEffectClass

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

Formats the value using the given formatter.

impl Default for SideEffectClass

fn default() -> SideEffectClass

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SideEffectClass

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for SideEffectClass

fn eq(&self, other: &SideEffectClass) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for SideEffectClass

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for SideEffectClass

impl Eq for SideEffectClass

impl StructuralPartialEq for SideEffectClass