Struct SessionRecord 

pub struct SessionRecord {

    pub id: String,
    pub provider: Provider,
    pub provider_session_id: String,
    pub title: Option<String>,
    pub summary: Option<String>,
    pub cwd: Option<String>,
    pub repo_root: Option<String>,
    pub created_at: Option<DateTime<Utc>>,
    pub updated_at: Option<DateTime<Utc>>,
    pub last_message_at: Option<DateTime<Utc>>,
    pub preview_text: String,
    pub source_path: String,
    pub message_count: Option<i64>,
    pub parse_version: String,
    pub raw_metadata_json: Option<String>,
    pub parse_warning: Option<String>,
    pub discovery_source: String,
    pub parent_session_id: Option<String>,
    pub agent_label: Option<String>,
}

Fields

id: String

provider: Provider

provider_session_id: String

title: Option<String>

summary: Option<String>

cwd: Option<String>

repo_root: Option<String>

created_at: Option<DateTime<Utc>>

updated_at: Option<DateTime<Utc>>

last_message_at: Option<DateTime<Utc>>

preview_text: String

source_path: String

message_count: Option<i64>

parse_version: String

raw_metadata_json: Option<String>

parse_warning: Option<String>

discovery_source: String

parent_session_id: Option<String>

The SessionRecord::id of the session that spawned this one, when this session is a subagent run.

None for an ordinary top-level session. Every provider marks subagent runs differently (claude by a subagents directory under its parent, codex in a thread_spawn payload, cursor and pi by directory layout), so detection stays provider-specific while this field is the one shape they all produce. Typed rather than buried in raw_metadata_json so “every subagent of this session” is a query, which is what made codex’s richer spawn data unusable.

Holds the whole id, provider prefix included, so it reads as the same value the parent row carries in id and needs no rule about stripping a prefix before comparing.

PATTERN: this names a row that need not exist. Providers keep a spawned run’s transcript after the parent conversation is rotated away — in one project directory, 549 runs referenced 69 distinct parents and not one of those parent transcripts was still on disk. Recovering exactly that otherwise-unreachable work is a reason this is indexed, so do NOT add references sessions(id) to this column: the constraint would reject every such run at insert, and a cascade would delete the runs when a parent is reindexed away.

agent_label: Option<String>

Human-meaningful name for the spawned agent when the provider records one: claude’s agentType (Explore, general-purpose), codex’s agent_nickname, or the agent’s own directory or file name where that is all a provider records. Display and grouping only; the link is parent_session_id.

Trait Implementations

impl Clone for SessionRecord

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SessionRecord

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for SessionRecord

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Row for SessionRecord

fn headers() -> &'static [&'static str]

Column headers, in order.

fn cells(&self) -> Vec<String>

Cell values for this row, in the same order as Row::headers.

impl Serialize for SessionRecord

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

where __S: Serializer,

Serialize this value into the given Serde serializer.