Struct RecallArgs 

pub struct RecallArgs {

    pub context: String,
    pub namespace: Option<String>,
    pub limit: usize,
    pub tags: Option<String>,
    pub since: Option<String>,
    pub until: Option<String>,
    pub tier: Option<String>,
    pub as_agent: Option<String>,
    pub budget_tokens: Option<usize>,
    pub context_tokens: Option<Vec<String>>,
    pub session_default: bool,
    pub include_archived: bool,
    pub has_citations: bool,
    pub source_uri_prefix: Option<String>,
    pub kind: Option<String>,
    pub confidence_tier: Option<String>,
    pub verbose_provenance: bool,
    pub format: String,
    pub session_id: Option<String>,
}

Clap-derived arg shape for the recall subcommand. Definition moved from main.rs verbatim in W5b — fields and attrs unchanged.

Fields

context: String

namespace: Option<String>

limit: usize

tags: Option<String>

since: Option<String>

until: Option<String>

tier: Option<String>

Feature tier for recall: keyword, semantic, smart, autonomous

as_agent: Option<String>

Task 1.5: querying agent’s namespace position. Enables scope-based visibility filtering (private/team/unit/org/collective).

budget_tokens: Option<usize>

Task 1.11: context-budget-aware recall. Return the top-ranked memories whose cumulative estimated tokens fit within N. Omit for unlimited (limit-based only).

context_tokens: Option<Vec<String>>

v0.6.0.0 contextual recall. Comma-separated list of recent conversation tokens used to bias the query embedding at 70/30 (primary/context). Shifts the recall towards memories that match both the explicit query and the conversation’s nearby topics.

session_default: bool

v0.7.0 (issue #518) — when set, splice defaults from [agents.defaults.recall_scope] in config.toml for any filter field not explicitly passed on the command line. Resolution: explicit args > recall_scope defaults > compiled defaults. Default false preserves v0.6.x recall semantics.

include_archived: bool

v0.7.0 WT-1-E — when set, recall returns archived sources (those replaced by their atoms after WT-1-B atomisation) alongside the atoms. Default false surfaces atoms only, which is the canonical post-atomisation recall unit.

has_citations: bool

v0.7.0 Form 4 (issue #757) — restrict results to memories whose citations array is non-empty. Composes with the other filters; default false (no provenance filter).

source_uri_prefix: Option<String>

v0.7.0 Form 4 (issue #757) — restrict results to memories whose source_uri starts with this prefix. Matches the substring exactly (no glob/regex). Typical use: --source-uri-prefix doc: to surface every atom or memory pointing at a substrate doc; --source-uri-prefix uri:https:// to surface every memory citing an HTTP source.

kind: Option<String>

v0.7.x Form 6 (issue #759) — Batman-taxonomy memory-kind filter. Comma-separated. Examples: –kind concept –kind concept,entity,claim –kinds concept,entity,claim (plural alias for MCP parity) Recognised values: observation, reflection, persona, concept, entity, claim, relation, event, conversation, decision. OR-of-kinds within the flag; AND with the other filters. Pass ‘all’ or omit for no filter.

Cluster E audit API-3 (issue #767): the MCP tool param is kinds (plural), so the CLI accepts both spellings via an alias for cross-interface ergonomics.

confidence_tier: Option<String>

v0.7.0 #1098 — restrict to memories whose confidence tier matches one of {high, medium, low}. Wired through to crate::models::RecallRequest::confidence_tier via RecallRequest::from_cli_args; the MCP / HTTP surfaces have accepted this filter since RC, the CLI surface closes the three-surface parity gap.

verbose_provenance: bool

v0.7.0 #1098 — when set, emit per-row provenance decoration (Gap-7 #890): citations, source_uri, source_span, confidence_source, confidence_signals. The flag flows through the DTO so MCP / HTTP / CLI agree on the verbose envelope shape; the JSON renderer downstream owns the actual expansion (today’s CLI emits the full Memory row already, so the flag is preserved for cross-surface parity).

format: String

v0.7.0 #1098 — response format selector: human (default pretty text), json (the same envelope --json produces), or toon (TOON compact format, ~79% smaller than JSON; see crate::toon). The MCP / HTTP surfaces accept the same vocabulary via RecallRequest::format. Default human preserves v0.6.x CLI semantics.

session_id: Option<String>

v0.7.0 #1257 — session-id parity flag (DTO C2 #967, +0.05 rerank boost under #518). Pre-#1257 this was hard-coded to None in RecallRequest::from_cli_args, so a CLI caller could not reach the in-session ring boost even though MCP ({"session_id": "…"} param) and HTTP (?session_id=… or JSON body) callers could. Optional; omit to preserve v0.6.x recall semantics.

Trait Implementations

impl Args for RecallArgs

fn group_id() -> Option<Id>

Report the ArgGroup::id for this set of arguments

fn augment_args<'b>(__clap_app: Command) -> Command

Append to Command so it can instantiate Self via FromArgMatches::from_arg_matches_mut

fn augment_args_for_update<'b>(__clap_app: Command) -> Command

Append to Command so it can instantiate self via FromArgMatches::update_from_arg_matches_mut

impl FromArgMatches for RecallArgs

fn from_arg_matches(__clap_arg_matches: &ArgMatches) -> Result<Self, Error>

Instantiate Self from ArgMatches, parsing the arguments as needed.

fn from_arg_matches_mut( __clap_arg_matches: &mut ArgMatches, ) -> Result<Self, Error>

Instantiate Self from ArgMatches, parsing the arguments as needed.

fn update_from_arg_matches( &mut self, __clap_arg_matches: &ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.

fn update_from_arg_matches_mut( &mut self, __clap_arg_matches: &mut ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.