Struct Session 

pub struct Session {

    pub session_id: String,
    pub machine_id: String,
    pub framework: Option<String>,
    pub first_started_at: DateTime<Utc>,
    pub last_started_at: DateTime<Utc>,
    pub first_ended_at: Option<DateTime<Utc>>,
    pub last_ended_at: Option<DateTime<Utc>>,
    pub last_activity: DateTime<Utc>,
    pub revival_count: u32,
    pub status: Option<SessionStatus>,
    pub status_updated_at: Option<DateTime<Utc>>,
    pub pid: Option<i32>,
    pub process_start_time: Option<i64>,
    pub cwd: Option<String>,
    pub local_git_dir: Option<String>,
    pub git_branch: Option<String>,
    pub git_worktree_path: Option<String>,
    pub github_repo: Option<String>,
    pub github_issue: Option<i32>,
    pub github_pr: Option<i32>,
    pub git_additions: Option<i32>,
    pub git_deletions: Option<i32>,
    pub first_user_message: Option<String>,
    pub last_user_message: Option<String>,
    pub model: Option<String>,
    pub compaction_count: u32,
    pub message_count: u32,
    pub tool_call_count: u32,
    pub api_request_count: u32,
    pub api_error_count: u32,
    pub tokens_input: i64,
    pub tokens_output: i64,
    pub tokens_cache_read: i64,
    pub tokens_cache_write: i64,
    pub cost_usd: Option<f64>,
    pub context: Option<i64>,
    pub n_hook_events: u32,
    pub n_otel_events: u32,
    pub transcript_available: bool,
    pub transcript_path: Option<String>,
}

A Claude Code session with aggregated metadata.

Sessions are derived from events but stored separately for efficient querying. All aggregate fields are updated incrementally when events are inserted.

Schema v2

This struct implements Sessions Schema v2 which uses unified columns with explicit source tracking instead of source-specific columns (e.g., tokens_input instead of tokens_input_otel and tokens_input_transcript).

Source priority is: transcript > otel > hook

• Transcript is preferred as the authoritative source (includes is_sidechain info)
• OTEL is used as fallback for token/API data
• Hook provides real-time status and prompts

Session Revival

Sessions can be revived after ending. The schema tracks:

• first_started_at / last_started_at: Original and most recent start times
• first_ended_at / last_ended_at: Original and most recent end times
• revival_count: Number of times the session was revived

Fields

session_id: String

Unique session identifier

machine_id: String

Machine identifier for multi-machine sync

framework: Option<String>

Source framework (claude_code, gemini, codex, etc.)

first_started_at: DateTime<Utc>

Original session start time (immutable)

last_started_at: DateTime<Utc>

Most recent start/revival time

first_ended_at: Option<DateTime<Utc>>

First time session ended (immutable historical record)

last_ended_at: Option<DateTime<Utc>>

Most recent end time (None = running)

last_activity: DateTime<Utc>

Timestamp of most recent activity

revival_count: u32

Number of times the session was revived after ending

status: Option<SessionStatus>

Current session status (idle, busy, wait, done)

status_updated_at: Option<DateTime<Utc>>

Timestamp when status was last updated (prevents stale updates)

pid: Option<i32>

Process ID (NULL when process ended, can change on revival)

process_start_time: Option<i64>

Process start time in seconds (Unix timestamp). Used for deterministic PID identity validation - (pid, start_time) is globally unique. This prevents false positives from PID reuse and handles late mi6 installation.

cwd: Option<String>

Working directory

local_git_dir: Option<String>

Absolute path to .git directory

git_branch: Option<String>

Current git branch name

git_worktree_path: Option<String>

Worktree root if in a worktree

github_repo: Option<String>

Repository in owner/repo format (e.g., “paradigmxyz/mi6”)

github_issue: Option<i32>

GitHub issue number from gh commands (most recent)

github_pr: Option<i32>

GitHub PR number from gh commands (most recent)

git_additions: Option<i32>

Lines added (from git diff vs main/master or gh pr view)

git_deletions: Option<i32>

Lines deleted (from git diff vs main/master or gh pr view)

first_user_message: Option<String>

First user prompt (transcript can correct hook’s initial value)

last_user_message: Option<String>

Most recent user message text

model: Option<String>

Model used in most recent API request

compaction_count: u32

Number of context compactions

message_count: u32

Total user message count

tool_call_count: u32

Total tool call count

api_request_count: u32

Total API request count

api_error_count: u32

Total failed API request count

tokens_input: i64

Total input tokens

tokens_output: i64

Total output tokens

tokens_cache_read: i64

Total cache read tokens

tokens_cache_write: i64

Total cache write tokens

cost_usd: Option<f64>

Total cost in USD

context: Option<i64>

Current context window usage (raw token count)

n_hook_events: u32

Cumulative count of hook events (enables priority logic)

n_otel_events: u32

Cumulative count of OTEL events (enables priority logic)

transcript_available: bool

Whether transcript has been successfully parsed

transcript_path: Option<String>

Full absolute path to transcript file

Implementations

impl Session

pub fn is_active(&self) -> bool

Check if the session is still active (no end time recorded)

pub fn is_process_running(&self) -> bool

Check if the session’s process is still running.

Returns true if the session has a PID and that process is still alive. Returns false if there’s no PID or the process is not running.

pub fn total_tokens(&self) -> i64

Total tokens across all categories

pub fn duration(&self) -> Duration

Duration of the session (from first start to last end or last activity)

pub fn metrics_source(&self) -> &'static str

Determine which source currently owns the metrics.

Priority: transcript > otel > hook

pub fn should_otel_update_tokens(&self) -> bool

Whether OTEL events should update token metrics.

OTEL can update tokens only when transcript is not available.

pub fn should_hook_update_counts(&self) -> bool

Whether hook events should update message/tool counts.

Hooks can update counts only when transcript is not available.

pub fn uses_shared_process(&self) -> bool

Check if this session uses a shared process (one process for many sessions).

For frameworks like Cursor IDE, a single long-running process hosts multiple sessions. PID-based liveness detection doesn’t work reliably for these because the process stays alive even after individual sessions end.

Use this to skip PID timestamp validation for shared-process frameworks.

pub fn is_amp(&self) -> bool

Check if this is an Amp session.

Amp sessions have limited data availability compared to other frameworks because Amp doesn’t support hooks. Features like tool calls, API request counts, compactions, and OTEL events are not tracked for Amp sessions.

Trait Implementations

impl Clone for Session

fn clone(&self) -> Session

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Session

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

Formats the value using the given formatter.

impl Default for Session

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for Session

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for Session

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

where __S: Serializer,

Serialize this value into the given Serde serializer.