:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import { MOCK_DEFAULT_ORGANIZATION } from 'lib/api.mock'
import { expectLogic } from 'kea-test-utils'
import { SIDE_PANEL_PANEL_ID } from '~/layout/ai/aiPanelIds'
import { sidePanelStateLogic } from '~/layout/navigation-3000/sidepanel/sidePanelStateLogic'
import { useMocks } from '~/mocks/jest'
import { initKeaTests } from '~/test/init'
import { SidePanelTab } from '~/types'
import { taskTrackerSceneLogic } from 'products/posthog_ai/frontend/scenes/TaskTracker/taskTrackerSceneLogic'
import { maxLogic } from './maxLogic'
import { phaiSidePanelComposerSeedLogic } from './phaiSidePanelComposerSeedLogic'
describe('phaiSidePanelComposerSeedLogic', () => {
let seedLogic: ReturnType<typeof phaiSidePanelComposerSeedLogic.build>
let trackerLogic: ReturnType<typeof taskTrackerSceneLogic.build>
let createCount: number
let runCount: number
// No hook calls in here (rules-of-hooks) — mocks are set up in beforeEach / the test bodies.
const mountPanelLogics = (): void => {
sidePanelStateLogic.mount()
trackerLogic = taskTrackerSceneLogic({ panelId: 'max-side-panel' })
trackerLogic.mount()
seedLogic = phaiSidePanelComposerSeedLogic({ panelId: 'max-side-panel' })
seedLogic.mount()
}
beforeEach(() => {
createCount = 0
runCount = 0
useMocks({
get: {
'/api/projects/:team/tasks/': { results: [], count: 0 },
'/api/projects/:team/tasks/repositories/': { repositories: [] },
'/api/projects/:team/integrations/': { results: [] },
},
post: {
'/api/projects/:team/tasks/': async ({ request }) => {
createCount++
return [200, { id: 'new-task', ...((await request.json()) as Record<string, any>) }]
},
'/api/projects/:team/tasks/:id/run/': () => {
runCount++
return [200, { id: 'new-task', latest_run: { id: 'run-1' } }]
},
},
})
})
afterEach(() => {
seedLogic?.unmount()
trackerLogic?.unmount()
})
// The panel swaps to the legacy chat when the server refuses a 2.0 send, which unmounts the runner and its
// restored draft. The refused message must become the legacy chat's draft, or the user loses it.
it('gives a send the server refuses to the legacy chat as its draft', async () => {
useMocks({
post: {
'/api/projects/:team/tasks/': () => [
403,
{ type: 'permission_denied', code: 'task_unavailable', error: "This task isn't available to you." },
],
},
})
initKeaTests()
const legacyChat = maxLogic({ panelId: SIDE_PANEL_PANEL_ID })
legacyChat.mount()
try {
mountPanelLogics()
trackerLogic.actions.setNewTaskData({ description: 'Explain the example chart' })
await expectLogic(trackerLogic, () => trackerLogic.actions.submitNewTask()).toFinishAllListeners()
expect(legacyChat.values.question).toBe('Explain the example chart')
} finally {
legacyChat.unmount()
}
})
// The org-level AI data-processing consent gate: an auto-run CTA (`!`-prefixed prompt) must never start
// an agent run before the org approved AI data processing — the legacy path blocks this in `askMax`, and
// dropping the `dataProcessingAccepted` check in the seed bridge would silently ship unconsented sandbox
// runs (the tasks backend has no server-side consent check).
it('holds an auto-run seed as prefill and sends no task create/run when consent is missing', async () => {
useMocks({
get: {
// The org bootstrapped via app context below gets re-fetched on organizationLogic mount —
// keep the API in agreement so a late load can't flip the consent value mid-test.
'/api/organizations/@current/': () => [
200,
{ ...MOCK_DEFAULT_ORGANIZATION, is_ai_data_processing_approved: false },
],
},
})
initKeaTests(true, undefined, undefined, {
...MOCK_DEFAULT_ORGANIZATION,
is_ai_data_processing_approved: false,
})
mountPanelLogics()
sidePanelStateLogic.actions.openSidePanel(SidePanelTab.Max, '!Foo bar')
await expectLogic(trackerLogic).toFinishAllListeners()
expect(createCount).toBe(0)
expect(runCount).toBe(0)
// The prompt is retained as a prefill so consent approval doesn't lose it; nothing was submitted.
expect(trackerLogic.values.newTaskData.description).toBe('Foo bar')
expect(trackerLogic.values.activeCreation).toBeNull()
})
// Control for the test above: with consent accepted the same CTA must create and run the task — proving
// the no-request assertion holds because of the gate, not because the bridge stopped forwarding seeds
// altogether. The org is passed explicitly since app context persists across tests in this file.
it('auto-submits an auto-run seed when consent is accepted', async () => {
initKeaTests(true, undefined, undefined, {
...MOCK_DEFAULT_ORGANIZATION,
is_ai_data_processing_approved: true,
})
mountPanelLogics()
sidePanelStateLogic.actions.openSidePanel(SidePanelTab.Max, '!Foo bar')
await expectLogic(trackerLogic).toFinishAllListeners()
expect(createCount).toBe(1)
expect(runCount).toBe(1)
expect(trackerLogic.values.activeCreation).toMatchObject({ taskId: 'new-task', runId: 'run-1' })
})
})