session activity identify stitching

This commit is contained in:
Francis Cao committed 2026-07-15 11:20:27 -07:00
1 parent 8a75208d5b
commit 7afc4ccf6b
15 files changed
+374 -23

No files matched your search

@@ -0,0 +1,12 @@
-- Create session_link
CREATE TABLE umami.session_link
(
website_id UUID,
session_id UUID,
distinct_id String,
created_at DateTime('UTC'),
INDEX idx_session_id session_id TYPE bloom_filter GRANULARITY 1
)
ENGINE = ReplacingMergeTree
ORDER BY (website_id, distinct_id, session_id)
SETTINGS index_granularity = 8192;
+13
View File
@@ -424,3 +424,16 @@ ENGINE = MergeTree
PARTITION BY toYYYYMM(created_at)
ORDER BY (website_id, url_path, event_type, created_at)
SETTINGS index_granularity = 8192;
-- Create session_link
CREATE TABLE umami.session_link
(
website_id UUID,
session_id UUID,
distinct_id String,
created_at DateTime('UTC'),
INDEX idx_session_id session_id TYPE bloom_filter GRANULARITY 1
)
ENGINE = ReplacingMergeTree
ORDER BY (website_id, distinct_id, session_id)
SETTINGS index_granularity = 8192;
@@ -0,0 +1,11 @@
-- CreateTable
CREATE TABLE "session_link" (
"website_id" UUID NOT NULL,
"session_id" UUID NOT NULL,
"distinct_id" VARCHAR(50) NOT NULL,
"created_at" TIMESTAMPTZ(6) DEFAULT CURRENT_TIMESTAMP,
CONSTRAINT "session_link_pkey" PRIMARY KEY ("website_id","distinct_id","session_id")
);
CREATE INDEX "session_link_website_id_session_id_idx" ON "session_link"("website_id", "session_id");
+11
View File
@@ -63,6 +63,17 @@ model Session {
@@map("session")
}
model SessionLink {
websiteId String @map("website_id") @db.Uuid
sessionId String @map("session_id") @db.Uuid
distinctId String @map("distinct_id") @db.VarChar(50)
createdAt DateTime? @default(now()) @map("created_at") @db.Timestamptz(6)
@@id([websiteId, distinctId, sessionId])
@@index([websiteId, sessionId])
@@map("session_link")
}
model Website {
id String @id() @map("website_id") @db.Uuid
name String @db.VarChar(100)
@@ -22,11 +22,13 @@ export function SessionActivity({
sessionId,
startDate,
endDate,
distinctId,
}: {
websiteId: string;
sessionId: string;
startDate: Date;
endDate: Date;
distinctId?: string;
}) {
const { t, labels } = useMessages();
const { formatTimezoneDate } = useTimezone();
@@ -35,6 +37,7 @@ export function SessionActivity({
sessionId,
startDate,
endDate,
distinctId,
);
const { isMobile } = useMobile();
let lastDay = null;
@@ -81,6 +81,7 @@ export function SessionProfile({
sessionId={sessionId}
startDate={data?.firstAt}
endDate={data?.lastAt}
distinctId={data?.distinctId}
/>
</TabPanel>
<TabPanel id="properties">
+34 -2
View File
@@ -11,13 +11,14 @@ import { parseRequest } from '@/lib/request';
import { badRequest, forbidden, json, serverError } from '@/lib/response';
import { anyObjectParam, urlOrPathParam } from '@/lib/schema';
import { safeDecodeURI, safeDecodeURIComponent } from '@/lib/url';
import { createSession, saveEvent, saveSessionData } from '@/queries/sql';
import { createSession, saveEvent, saveSessionData, saveSessionLink } from '@/queries/sql';
interface Cache {
websiteId: string;
sessionId: string;
visitId: string;
iat: number;
sessionLinkId?: string;
}
// Reject strings whose first character is a spreadsheet formula trigger to
@@ -128,6 +129,9 @@ export async function POST(request: Request) {
}
}
// Carried forward in the cache token so repeat identify calls skip the link write
let sessionLinkId = cache?.sessionLinkId;
// Client info
const { ip, userAgent, device, browser, os, country, region, city } = await getClientInfo(
request,
@@ -277,6 +281,34 @@ export async function POST(request: Request) {
twclid,
});
} else if (type === COLLECTION_TYPE.identify) {
if (websiteId && id) {
// Persist both memberships so stitched activity can be resolved from
// either the anonymous session or the identified session.
const anonymousSessionId = uuid(sourceId, ip, userAgent, sessionSalt);
const linkedSessionIds = [...new Set([anonymousSessionId, sessionId])];
const newLinkId = hash(...linkedSessionIds, id);
if (sessionLinkId !== newLinkId) {
// Best-effort: a link failure must not block the session data write below
try {
await Promise.all(
linkedSessionIds.map(linkedSessionId =>
saveSessionLink({
websiteId,
sessionId: linkedSessionId,
distinctId: id,
createdAt,
}),
),
);
sessionLinkId = newLinkId;
} catch (e) {
// eslint-disable-next-line no-console
console.error('Failed to save session link:', e);
}
}
}
if (data) {
await saveSessionData({
websiteId,
@@ -316,7 +348,7 @@ export async function POST(request: Request) {
}
const token = createToken(
{ websiteId, sessionId, visitId, iat, type: CACHE_TOKEN_TYPE },
{ websiteId, sessionId, visitId, iat, sessionLinkId, type: CACHE_TOKEN_TYPE },
secret(),
);
@@ -1,8 +1,15 @@
import { z } from 'zod';
import { FIELD_LENGTH } from '@/lib/constants';
import { uuid } from '@/lib/crypto';
import { getQueryFilters, parseRequest } from '@/lib/request';
import { json, unauthorized } from '@/lib/response';
import { canViewWebsiteSection } from '@/permissions';
import { getSessionActivity } from '@/queries/sql';
import {
getLinkedDistinctIds,
getLinkedSessionIds,
getSessionActivity,
getSessionActivityBounds,
} from '@/queries/sql';
export async function GET(
request: Request,
@@ -11,6 +18,7 @@ export async function GET(
const schema = z.object({
startAt: z.coerce.number().int(),
endAt: z.coerce.number().int(),
distinctId: z.string().max(FIELD_LENGTH.distinctId).optional(),
});
const { auth, query, error } = await parseRequest(request, schema);
@@ -22,19 +30,38 @@ export async function GET(
const { websiteId, sessionId } = await params;
if (
!(await canViewWebsiteSection(auth, websiteId, [
'sessions',
'events',
'realtime',
'revenue',
]))
!(await canViewWebsiteSection(auth, websiteId, ['sessions', 'events', 'realtime', 'revenue']))
) {
return unauthorized();
}
const filters = await getQueryFilters(query, websiteId);
let sessionIds = [sessionId];
let startAt = query.startAt;
let endAt = query.endAt;
const distinctIds = query.distinctId
? [query.distinctId]
: await getLinkedDistinctIds(websiteId, sessionId);
const data = await getSessionActivity(websiteId, sessionId, filters);
if (distinctIds.length) {
const links = await Promise.all(distinctIds.map(distinctId => getLinkedSessionIds(websiteId, distinctId)));
const linkedIds = links.flatMap(group => group.map(link => link.sessionId));
const identifiedIds = distinctIds.map(distinctId => uuid(websiteId, distinctId));
sessionIds = Array.from(new Set([sessionId, ...identifiedIds, ...linkedIds]));
if (sessionIds.length > 1) {
const bounds = await getSessionActivityBounds(websiteId, sessionIds);
if (bounds?.firstAt && bounds?.lastAt) {
startAt = Math.min(startAt, +new Date(bounds.firstAt));
endAt = Math.max(endAt, +new Date(bounds.lastAt));
}
}
}
const filters = await getQueryFilters({ ...query, startAt, endAt }, websiteId);
const data = await getSessionActivity(websiteId, sessionIds, filters);
return json(data);
}
@@ -5,15 +5,17 @@ export function useSessionActivityQuery(
sessionId: string,
startDate: Date,
endDate: Date,
distinctId?: string,
) {
const { get, useQuery } = useApi();
return useQuery({
queryKey: ['session:activity', { websiteId, sessionId, startDate, endDate }],
queryKey: ['session:activity', { websiteId, sessionId, startDate, endDate, distinctId }],
queryFn: () => {
return get(`/websites/${websiteId}/sessions/${sessionId}/activity`, {
startAt: +new Date(startDate),
endAt: +new Date(endDate),
...(distinctId && { distinctId }),
});
},
enabled: Boolean(websiteId && sessionId && startDate && endDate),
+4
View File
@@ -43,7 +43,10 @@ export * from './reports/getJourney';
export * from './reports/getRetention';
export * from './reports/getUTM';
export * from './sessions/createSession';
export * from './sessions/getLinkedDistinctIds';
export * from './sessions/getLinkedSessionIds';
export * from './sessions/getSessionActivity';
export * from './sessions/getSessionActivityBounds';
export * from './sessions/getSessionData';
export * from './sessions/getSessionDataActivityStats';
export * from './sessions/getSessionDataArraySeries';
@@ -61,3 +64,4 @@ export * from './sessions/getWebsiteSession';
export * from './sessions/getWebsiteSessionStats';
export * from './sessions/getWebsiteSessions';
export * from './sessions/saveSessionData';
export * from './sessions/saveSessionLink';
@@ -0,0 +1,44 @@
import clickhouse from '@/lib/clickhouse';
import { CLICKHOUSE, PRISMA, runQuery } from '@/lib/db';
import prisma from '@/lib/prisma';
const FUNCTION_NAME = 'getLinkedDistinctIds';
export async function getLinkedDistinctIds(
...args: [websiteId: string, sessionId: string]
): Promise<string[]> {
return runQuery({
[PRISMA]: () => relationalQuery(...args),
[CLICKHOUSE]: () => clickhouseQuery(...args),
});
}
async function relationalQuery(websiteId: string, sessionId: string): Promise<string[]> {
const { rawQuery } = prisma;
return rawQuery(
`
select distinct distinct_id as "distinctId"
from session_link
where website_id = {{websiteId::uuid}}
and session_id = {{sessionId::uuid}}
`,
{ websiteId, sessionId },
FUNCTION_NAME,
).then(result => (result as { distinctId: string }[]).map(({ distinctId }) => distinctId));
}
async function clickhouseQuery(websiteId: string, sessionId: string): Promise<string[]> {
const { rawQuery } = clickhouse;
return rawQuery(
`
select distinct distinct_id as distinctId
from session_link
where website_id = {websiteId:UUID}
and session_id = {sessionId:UUID}
`,
{ websiteId, sessionId },
FUNCTION_NAME,
).then(result => (result as { distinctId: string }[]).map(({ distinctId }) => distinctId));
}
@@ -0,0 +1,50 @@
import clickhouse from '@/lib/clickhouse';
import { CLICKHOUSE, PRISMA, runQuery } from '@/lib/db';
import prisma from '@/lib/prisma';
const FUNCTION_NAME = 'getLinkedSessionIds';
export async function getLinkedSessionIds(
...args: [websiteId: string, distinctId: string]
): Promise<{ sessionId: string; createdAt: string }[]> {
return runQuery({
[PRISMA]: () => relationalQuery(...args),
[CLICKHOUSE]: () => clickhouseQuery(...args),
});
}
async function relationalQuery(websiteId: string, distinctId: string) {
const { rawQuery } = prisma;
return rawQuery(
`
select
session_id as "sessionId",
min(created_at) as "createdAt"
from session_link
where website_id = {{websiteId::uuid}}
and distinct_id = {{distinctId}}
group by session_id
`,
{ websiteId, distinctId },
FUNCTION_NAME,
);
}
async function clickhouseQuery(websiteId: string, distinctId: string) {
const { rawQuery } = clickhouse;
return rawQuery(
`
select
session_id as sessionId,
min(created_at) as createdAt
from session_link
where website_id = {websiteId:UUID}
and distinct_id = {distinctId:String}
group by session_id
`,
{ websiteId, distinctId },
FUNCTION_NAME,
);
}
+11 -11
View File
@@ -7,7 +7,7 @@ import type { QueryFilters } from '@/lib/types';
const FUNCTION_NAME = 'getSessionActivity';
export async function getSessionActivity(
...args: [websiteId: string, sessionId: string, filters: QueryFilters]
...args: [websiteId: string, sessionIds: string[], filters: QueryFilters]
) {
return runQuery({
[PRISMA]: () => relationalQuery(...args),
@@ -15,7 +15,7 @@ export async function getSessionActivity(
});
}
async function relationalQuery(websiteId: string, sessionId: string, filters: QueryFilters) {
async function relationalQuery(websiteId: string, sessionIds: string[], filters: QueryFilters) {
const { rawQuery } = prisma;
const { startDate, endDate } = filters;
@@ -37,18 +37,18 @@ async function relationalQuery(websiteId: string, sessionId: string, filters: Qu
and created_at between {{startDate}} and {{endDate}}) AS "hasData"
from website_event
where website_id = {{websiteId::uuid}}
and session_id = {{sessionId::uuid}}
and session_id = any({{sessionIds}}::uuid[])
and event_type != ${EVENT_TYPE.performance}
and created_at between {{startDate}} and {{endDate}}
order by created_at desc
limit 500
`,
{ websiteId, sessionId, startDate, endDate },
{ websiteId, sessionIds, startDate, endDate },
FUNCTION_NAME,
);
}
async function clickhouseQuery(websiteId: string, sessionId: string, filters: QueryFilters) {
async function clickhouseQuery(websiteId: string, sessionIds: string[], filters: QueryFilters) {
const { rawQuery } = clickhouse;
const { startDate, endDate } = filters;
@@ -64,20 +64,20 @@ async function clickhouseQuery(websiteId: string, sessionId: string, filters: Qu
event_name as eventName,
visit_id as visitId,
hostname,
event_id IN (select event_id
from event_data
where website_id = {websiteId:UUID}
and session_id = {sessionId:UUID}
event_id IN (select event_id
from event_data
where website_id = {websiteId:UUID}
and session_id in {sessionIds:Array(UUID)}
and created_at between {startDate:DateTime64} and {endDate:DateTime64}) AS hasData
from website_event
where website_id = {websiteId:UUID}
and session_id = {sessionId:UUID}
and session_id in {sessionIds:Array(UUID)}
and event_type != ${EVENT_TYPE.performance}
and created_at between {startDate:DateTime64} and {endDate:DateTime64}
order by created_at desc
limit 500
`,
{ websiteId, sessionId, startDate, endDate },
{ websiteId, sessionIds, startDate, endDate },
FUNCTION_NAME,
);
}
@@ -0,0 +1,62 @@
import clickhouse from '@/lib/clickhouse';
import { EVENT_TYPE } from '@/lib/constants';
import { CLICKHOUSE, PRISMA, runQuery } from '@/lib/db';
import prisma from '@/lib/prisma';
const FUNCTION_NAME = 'getSessionActivityBounds';
interface SessionActivityBounds {
firstAt: string | Date;
lastAt: string | Date;
}
export async function getSessionActivityBounds(
...args: [websiteId: string, sessionIds: string[]]
): Promise<SessionActivityBounds | null> {
return runQuery({
[PRISMA]: () => relationalQuery(...args),
[CLICKHOUSE]: () => clickhouseQuery(...args),
});
}
async function relationalQuery(
websiteId: string,
sessionIds: string[],
): Promise<SessionActivityBounds | null> {
const { rawQuery } = prisma;
return rawQuery(
`
select
min(created_at) as "firstAt",
max(created_at) as "lastAt"
from website_event
where website_id = {{websiteId::uuid}}
and session_id = any({{sessionIds}}::uuid[])
and event_type != ${EVENT_TYPE.performance}
`,
{ websiteId, sessionIds },
FUNCTION_NAME,
).then(result => (result?.[0]?.firstAt && result?.[0]?.lastAt ? result[0] : null));
}
async function clickhouseQuery(
websiteId: string,
sessionIds: string[],
): Promise<SessionActivityBounds | null> {
const { rawQuery, getDateStringSQL } = clickhouse;
return rawQuery(
`
select
${getDateStringSQL('min(created_at)')} as firstAt,
${getDateStringSQL('max(created_at)')} as lastAt
from website_event
where website_id = {websiteId:UUID}
and session_id in {sessionIds:Array(UUID)}
and event_type != ${EVENT_TYPE.performance}
`,
{ websiteId, sessionIds },
FUNCTION_NAME,
).then(result => (result?.[0]?.firstAt && result?.[0]?.lastAt ? result[0] : null));
}
@@ -0,0 +1,79 @@
import clickhouse from '@/lib/clickhouse';
import { FIELD_LENGTH } from '@/lib/constants';
import { CLICKHOUSE, PRISMA, runQuery } from '@/lib/db';
import { truncateString } from '@/lib/format';
import kafka from '@/lib/kafka';
import prisma from '@/lib/prisma';
const FUNCTION_NAME = 'saveSessionLink';
export interface SaveSessionLinkArgs {
websiteId: string;
sessionId: string;
distinctId: string;
createdAt?: Date;
}
export async function saveSessionLink(data: SaveSessionLinkArgs) {
return runQuery({
[PRISMA]: () => relationalQuery(data),
[CLICKHOUSE]: () => clickhouseQuery(data),
});
}
async function relationalQuery({
websiteId,
sessionId,
distinctId,
createdAt,
}: SaveSessionLinkArgs) {
const { writeRawQuery } = prisma;
await writeRawQuery(
`
insert into session_link (
website_id,
session_id,
distinct_id,
created_at
)
values (
{{websiteId}},
{{sessionId}},
{{distinctId}},
{{createdAt}}
)
on conflict (website_id, distinct_id, session_id) do nothing
`,
{
websiteId,
sessionId,
distinctId: truncateString(distinctId, FIELD_LENGTH.distinctId),
createdAt,
},
FUNCTION_NAME,
);
}
async function clickhouseQuery({
websiteId,
sessionId,
distinctId,
createdAt,
}: SaveSessionLinkArgs) {
const { insert, getUTCString } = clickhouse;
const { sendMessage } = kafka;
const message = {
website_id: websiteId,
session_id: sessionId,
distinct_id: truncateString(distinctId, FIELD_LENGTH.distinctId),
created_at: getUTCString(createdAt),
};
if (kafka.enabled) {
await sendMessage('session_link', [message]);
} else {
await insert('session_link', [message]);
}
}