A desktop coding agent with free access to frontier models.
TypeScript
import { sql } from "drizzle-orm"
import { boolean, index, int, json, mysqlTable, uniqueIndex, varchar, timestamp } from "drizzle-orm/mysql-core"
import { denTypeIdColumn } from "../columns"
/**
* Telemetry event vocabulary. Layer 1 answers "who is using AI" (activity
* pings); Layer 2 answers "how often" (session and task lifecycle). Event
* type strings are a wire contract with the desktop app and workers.
*/
export const TelemetryEventType = [
// Layer 1 — who is using AI
"user.active",
"session.active",
// Layer 2 — how often
"session.started",
"session.ended",
"task.started",
"task.completed",
"task.failed",
] as const
/**
* One telemetry event, org-scoped and attributed to a member. Rows carry only
* identifiers, timings, and outcomes — never user content. `session_id` is an
* opaque correlation id; `source` distinguishes desktop/web ("app") from
* remote runtimes ("worker").
*/
export const TelemetryEventTable = mysqlTable(
"telemetry_event",
{
id: denTypeIdColumn("telemetryEvent", "id").notNull().primaryKey(),
org_id: denTypeIdColumn("organization", "org_id").notNull(),
member_id: denTypeIdColumn("member", "member_id").notNull(),
event_type: varchar("event_type", { length: 64 }).notNull(),
event_timestamp: timestamp("event_timestamp", { fsp: 3 }).notNull(),
source: varchar("source", { length: 32 }),
session_id: varchar("session_id", { length: 128 }),
duration_ms: int("duration_ms"),
success: boolean("success"),
created_at: timestamp("created_at", { fsp: 3 }).notNull().defaultNow(),
},
(table) => [
index("telemetry_event_org_id_type_ts").on(table.org_id, table.event_type, table.event_timestamp),
index("telemetry_event_org_id_member_id").on(table.org_id, table.member_id),
index("telemetry_event_member_ts").on(table.member_id, table.event_timestamp),
index("telemetry_event_org_session_ts").on(table.org_id, table.session_id, table.event_timestamp),
],
)
/**
* Analytics dimensions attached to a session (for example project or model),
* upserted per (org, source, session, dimension type). Values are bounded
* slugs; labels are display text; metadata is a small JSON blob. Sessions are
* joined back to events per (org, session, source).
*/
export const TelemetrySessionDimensionTable = mysqlTable(
"telemetry_session_dimension",
{
id: denTypeIdColumn("telemetrySessionDimension", "id").notNull().primaryKey(),
org_id: denTypeIdColumn("organization", "org_id").notNull(),
session_id: varchar("session_id", { length: 128 }).notNull(),
source: varchar("source", { length: 32 }).notNull(),
dimension_type: varchar("dimension_type", { length: 64 }).notNull(),
dimension_value: varchar("dimension_value", { length: 128 }).notNull(),
dimension_label: varchar("dimension_label", { length: 255 }).notNull(),
metadata: json("metadata").$type<Record<string, unknown> | null>(),
created_at: timestamp("created_at", { fsp: 3 }).notNull().defaultNow(),
updated_at: timestamp("updated_at", { fsp: 3 })
.notNull()
.default(sql`CURRENT_TIMESTAMP(3) ON UPDATE CURRENT_TIMESTAMP(3)`),
last_seen_at: timestamp("last_seen_at", { fsp: 3 }).notNull().defaultNow(),
},
(table) => [
uniqueIndex("telemetry_session_dimension_org_source_session_type").on(
table.org_id,
table.source,
table.session_id,
table.dimension_type,
),
index("telemetry_session_dimension_filter").on(
table.org_id,
table.dimension_type,
table.dimension_value,
table.session_id,
),
],
)