User Guide

This guide reflects the current user-facing workflow in the app.

Signing In

Email/password sign-in uses Better Auth on every instance.
Google sign-in appears when configured for the instance.
Successful sign-in creates an authenticated workspace session.
With SMTP or Resend, signup shows a check-email confirmation before login. Verification links preserve the requested workspace destination.
Password changes require a linked email/password account. Google-only users do not see current-password controls. Recovery is unavailable with testing email configuration (EMAIL_PROVIDER=none). Password resets revoke sessions.

Main authenticated areas:

/recordings
/transcription (with bulk import at /transcription/import)
/contacts
/files
/upload
/settings
/support

The web app works on desktop/mobile browsers and as a PWA. Native Android work is archived.

Uploading Recordings

The primary upload flow lives on /upload.

Prepare the original recording files

For the organized bulk archive, use a supported dialer that saves original call recordings in storage visible to your phone’s file manager. Copy the files together without renaming them and keep a separate backup. On supported Samsung Phone setups, start at My Files → Internal storage → Recordings → Call. Xiaomi, OPPO, and vivo/iQOO OEM options depend on the exact firmware, region, file access, and filename layout; a recording button alone is insufficient.

Google Phone’s individual Share workflow is unsupported for this archive. Callery cannot read private dialer storage, import the dialer’s private call log, or reconstruct caller/date details missing from shared filenames. Switching to a supported OEM dialer affects future recordings; it does not migrate old ones. Check the phone compatibility guide and try a few original files before importing a whole collection.

Automatic caller and date recognition depends on the original filename; see the recognized layouts. An unfamiliar supported audio file can still upload, but its unknown caller and upload time do not provide the intended contact-and-date archive.

Transfer and resume uploads

Uploads use a resumable session lasting 24 hours. The queue shows byte progress, expiry and per-file status. Pause keeps completed parts; cancel discards the unfinished upload. Two files can transfer concurrently, with four part transfers across the queue. Uploads continue while navigating inside the app.

After a reload or browser restart, choose Select original file to continue. The browser stores session metadata and part fingerprints, not another copy of your audio. Completed parts are checked against the selected file before reuse. Offline uploads wait for connectivity; uncertain completion retries the same session rather than transferring the file again. Closing the browser stops transfers until it is reopened.

Bytes travel directly to storage when the bucket supports verified multipart uploads, or through bounded part requests to the app. Finalization validates the object, creates one recording and optionally queues auto-transcription. Bulk imports retain their processing hold through retries.

If the filename contains a recognizable contact name, phone number, or timestamp, the uploader uses that to pre-fill the recording context and storage key.

Current limits:

maximum upload size: 5 GiB
maximum transcription-eligible size: 2 GiB
supported formats: mp3, wav, m4a, mp4, webm

Files larger than the transcription limit still upload and remain playable.

Recordings Workspace

/recordings is the main archive view. Users can:

browse and paginate recordings
filter by contact and recording date, and sort by date or duration
search by text or semantic meaning
save custom views locally in the browser
apply the Recent week, Longest calls, and Needs transcription collections; Recent week uses the recording date
open a recording detail view
request manual transcription
generate signed share links

The main recording list excludes archived recordings. The data model and update action retain tags and star/archive fields, but the current web interface does not expose controls for editing or filtering them.

In AI search, Index saved transcripts makes older transcripts searchable by meaning without transcribing their audio again. Contact/date filters apply. Each confirmed request queues up to 100 recordings; use the queue status and repeat for the remaining library. Indexing uses the configured embedding provider, so provider charges may apply. Saved embedding responses are reused on retry; uncertain provider outcomes require administrator review. This also rebuilds indexes after an embedding-profile change.

Keyword fallback can paginate the full matching archive. Search by meaning ranks up to 200 conversations per query; narrow the contact/date filters when needed.

Recording detail state can include:

playback
transcript segments with speaker labels
summary
transliteration output when available
transcription provider metadata
sharing controls

Finding a moment in a call

In AI search, describe what you remember. Search covers transcript passages, including details absent from the summary. Contact and date filters also apply. Results show matching excerpts; open one to jump to its transcript segment and its audio time when timing is available. Search coverage indicates recordings still indexing or needing processing attention. A deployment without a ready vector index displays explicitly labelled keyword matches instead.

Within a recording, click a timestamp to seek and follow the highlighted current segment. Find in transcript supports previous/next matches in the displayed original or transliterated script. Playback position is remembered per signed-in account and recording; an explicit search result takes precedence. Public-share recipient positions are not saved. Transcripts remain read-only and display every speaker returned by the model as Speaker 1, Speaker 2 and so on. Missing timing is shown as unavailable.

Use Copy transcript or Download TXT to take a single transcript out of the viewer. Both include all loaded segments with speaker labels and available timestamps, even when Find highlights only a few matches. They follow the selected Original or Transliteration script, falling back to original text where a segment has no transliteration. No new processing or archive export job is started.

Transcription Workspace

Open Transcription → Bulk import to upload recordings without starting provider processing. On /transcription, select waiting recordings to start a batch or save it for later. Saved batches can be started, paused, and resumed explicitly; finishing one does not start another saved draft. Other batches already started remain eligible for processing. See transcription batches for processing limits, saved-transcript recovery, and uncertain provider outcomes.

Share the complete recording, including its audio, transcript, transliteration and summary when available. Recipients do not need an account.

Each link has an expiry: choose one hour, 24 hours (the default), seven days or a custom future date/time. Open Share to copy existing links, review their status and revoke them individually. Expiry and revocation deny subsequent page and audio requests; content already received or buffered cannot be withdrawn.

Public sharing is disabled in a company archive, including links issued before company mode was enabled. Administrators use Company recordings instead.

Contacts

/contacts is the relationship-focused view of the archive.

Users can:

search contacts
open a contact detail page
review linked recordings for a person or company
sync Google Contacts from Settings

Contact records track aggregate information such as total recordings and last recording date so the app can use contacts as an index into the archive. In Settings → Sources, Connect Google opens a separate Contacts authorization flow and starts the initial sync in the background. Once connected, Sync now waits for a refresh to finish. Disconnect removes the saved connection while keeping the last synced contacts until a future sync replaces them. Shared phone numbers appear as review items to resolve in Google Contacts. The sign-in identity and Contacts consent are separate, even when both use the same Google Cloud project.

Files

/files shows a storage-oriented view derived from recording keys.

Use it when you want to:

browse the archive like a folder tree
inspect recordings from their storage path
clean up intentionally through the file browser workflow

This is different from /recordings, which is optimized for retrieval and search. The file browser can also open a recording detail dialog or launch the single-item delete flow from the storage view. Company members cannot permanently delete recordings through this or any other view.

Settings

The current settings tabs are:

Profile: name, password changes, and account details
General: timezone
Sources: Google Contacts connection state
Processing: auto-transcription and its default provider
Integrations: named API tokens and upload/sync client access
Retention: usage review and automatic deletion policy
Export: archive export jobs and downloads
Notifications: activity digest and browser notification preferences
Billing: plan, usage, invoices, portal, and checkout when billing is enabled

Notable settings behavior:

deletion policy values are never, 3m, 6m, and 1y
export jobs create downloadable archive parts and an offline viewer
billing is hidden unless NEXT_PUBLIC_ENABLE_BILLING=true
billing controls appear under the Billing & Plans panel title when enabled

In a company archive, Processing and Retention show the administrator’s policy instead of personal editing controls. Integrations creates scoped API tokens for clients, displays each new secret once, and lets you rotate or revoke tokens separately. Connected OAuth apps are listed with their permissions and can be disconnected independently. AI search results link back to Callery for playback; opening a link still requires an authorized Callery session. The public website publishes integration setup and AI assistant connection instructions. Each app instance exposes its live OpenAPI reference at /api/v1/openapi.json; manage credentials in Settings → Integrations.

Closing a personal account

Use Profile → Delete account to permanently close a personal account. Export your archive first if you want to keep a copy. Confirmation requires typing your email address and a login from the last ten minutes; an older login requires signing in again and repeating confirmation. The last active administrator must make another user an administrator before closing their account.

Accepted closure removes access immediately and signs you out. If storage or active processing delays deletion, the confirmation page explains that cleanup will continue automatically. Existing charges and billing settlement still apply. Company accounts instead show guidance to contact an administrator, who can remove access while retaining the company archive.

Exports

Exports are designed to stay useful outside the app.

Each export job packages:

zipped archive parts stored in the instance’s private S3-compatible storage
recordings grouped by contact and date
JSON metadata
transcript text files
a viewer.html file for offline browsing

Only one pending or processing export job is allowed per user at a time.

One fresh full archive is available per 24 hours. If a storage transfer fails, Retry export resumes the same job once without starting a new quota window. Saved archive parts are reused only after verification. Retry is unavailable if the recordings changed, the archive expired, access was removed, or both generation attempts were used; the export panel explains the reason. Individual recording downloads remain available.

Archive parts are retained for seven days after completion. Download links last 15 minutes; reopen the export panel for fresh links.

Extract every archive part into the same folder, then open viewer.html. If the browser asks for a directory, select the extracted Contacts folder. An export contains your recording library; complete deployment backup and restore require the database, bucket, and configuration backups.

Notifications

The notification center can show:

transcription completion or failure
export completion
billing-related notices

The center initially loads the newest 20 notifications. Load older notifications opens more history; the unread badge counts all unread notifications, including those not loaded yet. Mark all read and Clear list apply to the whole notification history.

Recording activity emails are limited to optional weekly or monthly activity digests; individual uploads and transcriptions do not send processing emails. New accounts start with weekly digests, which can be disabled in Settings. Monthly digests cover the previous UTC calendar month. Weekly digests cover the seven days ending at the scheduled run time; future-dated recordings are excluded.

If VAPID keys are configured, users can also register browser push notifications. Expired browser subscriptions are removed after a permanent provider response; temporary delivery errors keep the subscription for future notifications. Administrator announcements appear separately as workspace notices.

Billing

When billing is enabled, users can:

review the active plan
see storage and transcription usage
enable or disable overage on an eligible recurring plan
start Stripe Checkout
open the Stripe customer portal
review invoice history

If billing is disabled for the deployment, these controls are not shown. Free-trial users cannot enable pay-per-usage overage. Lifetime owners manage or cancel their separately billed usage through the customer portal; they do not have the recurring-plan overage switch.

Company Archives

Company members see their own recordings and contacts. Administrators can retrieve employee recordings through /admin/recordings, including former employees’ archives. Removing access keeps those recordings; a timed company retention policy can still delete them. See small team setup.

Android archive

The native wrapper is archived. Upload recordings through the web app/PWA or a client using the integration API. Backend upload/sync APIs remain available; see Android archive.

Support

/support lets users submit a support request with:

name
email
subject
message

The request is stored in MongoDB. If support email delivery is configured, the app also sends the outbound email after the response is returned. The route works for signed-in and signed-out visitors, but it enforces origin checks, rate limits, and a honeypot field.