Release notes

plycdn ships two product lines, each with its own version. Content Search (the packages @plycdn/content-search, @plycdn/content-search-react, Plycdn.ContentSearch.AspNetCore, @plycdn/content-search-node and plycdn-content-search) moves in step with the service. Storage & video (Plycdn.AspNetCore.Storage, .Video, .Storage.S3, .Storage.AzureBlob, @plycdn/angular, @plycdn/react, @plycdn/node and plycdn) has its own line, and its packages always move together. The guides on this site describe the current release of both; how to move between releases is on Upgrading, together with moving to plycdn.com.

1Coming in the next release: bills you can check

Existing integrations keep working; every field below is new and optional to read.

  • A GB is 1,000,000,000 bytes everywhere: on the bill, in the dashboard and in your storage limits, from November 2026. GB quantities are written with two decimals (1,675.42 GB), and the statement CSV gives each in bytes too. Your bill and running charges.
  • A rate on every line. A line priced in bands shows the rate you are on now and each band it went through (currentRate, appliedRates).
  • Provisional figures. Delivery and storage-API use that has not settled yet is shown beside its line as an estimate (provisional, provisionalTotal), never in the total.
  • Running charges. Billing & usage → Costs shows this month's charges day by day; from your server, GetChargesAsync / getCharges / get_charges (GET /usage/charges), in both product lines; and your AI assistant can read them (get_charges).
  • A daily usage email, for each member who turns it on under Account & security → Profile & organization.

2New in Content Search 1.47.0 and Storage & video 1.38.0

Existing integrations keep working; a route table you saved before is served exactly as it was.

  • Redirect a path. A rule on a shared address can now send a folder, or the whole address, to another address: permanent (301) or temporary (302), keeping the rest of the path and the query if you wish. Domains & routing → Routing → the address → Rules → Add rule → Redirect. Redirect a path.
  • Share traffic between buckets, with backups. A rule can name up to five buckets, each with a share of the viewers. A viewer stays on one bucket, and when it cannot answer the next one does, so a bucket that is down does not take the address down. Share traffic between buckets.
  • Packages. The server packages build both kinds of rule and sign links on a shared rule with the first bucket's key automatically. New error codes: route_redirect_invalid, route_redirect_loop, route_targets_invalid, route_bucket_and_targets.
  • A refreshed dashboard. Every list now filters, sorts and pages the same way, with the filters kept in the page address so a view can be shared; new buckets, moves from another cloud and custom domains are set up step by step; and deleting a domain or revoking a key asks you to type its name first. Every page, link and action you used before is still there.

3New in Content Search 1.46.0 and Storage & video 1.37.0

Existing integrations keep working; nothing changes for your API keys or your existing addresses.

  • Course packages and other web pages on your own domain. A private bucket can now show web pages in the browser on its custom domains instead of downloading them, so SCORM and xAPI packages play from your own address as they did on your old host. Off by default; the bucket's plycdn address always downloads pages. Storage buckets → the bucket → Settings → Where files can be shown, or delivery.webPagesOnOwnDomains (WebPagesOnOwnDomains in .NET). Courses and other web pages.
  • Moves from another cloud are checked as they copy. Each file's checksum is taken while it is copied, so a copied file is verified at once instead of being read back afterwards.
  • Custom domains switch sooner. When a domain's certificate is ready before its CNAME, the domain shows Certificate issued — waiting for your CNAME and goes live within about a minute of the address switch.

4New in Content Search 1.45.0 and Storage & video 1.36.0

Existing integrations keep working; nothing changes for your API keys or your existing addresses.

  • Serve several buckets from one address. A custom domain can now carry a route table: each path prefix (such as courses/ or org-content/videos/) is served from a bucket of your organization or one of its sub-organizations, at most one video library answers the player's paths, and the domain's own bucket stays the default. Up to 50 routes; changes take effect within about a minute, with no DNS or certificate change, and any earlier table can be restored in one step. Delivery → Custom domains → the domain → Routes, or GET/PUT /domains/{id}/routes and POST /domains/{id}/routes/restore. Links signed for a shared address cover the full path the viewer sees: the server packages add pathPrefix (PathPrefix in .NET) to signing, and the route-table calls (GetDomainRoutesAsync, SetDomainRoutesAsync, RestoreDomainRoutesAsync; getDomainRoutes…; get_domain_routes…). New error codes: route_prefix_invalid, route_prefix_duplicate, route_prefix_reserved, route_too_many, route_video_twice, route_target_not_found, route_target_scope, route_version_not_found and routes_need_private_address. Serve several buckets from one address.
  • Faster moves from another cloud. Copies reuse their connection to the old storage and no longer wait on one another while they register, and Files copied at once now goes up to 256 (was 64), with more servers joining automatically for a large move, so a large move copies several times faster.

5New in Content Search 1.44.0 and Storage & video 1.35.0

Existing integrations keep working; nothing changes for your API keys or your moves.

  • Old links through a gateway or proxy. When you move from another cloud and your application's old links reach the storage through a gateway that puts a fixed path in front of the file (such as /org-content/videos/...), list that path under Paths in front of old links (Storage buckets → the bucket → Move from another cloud), or send legacyPathPrefixes when you set up or change the move. Up to 10 paths; links that start with one are served as the file that follows it, and other old links work as before. The move shows the list as legacyLinks.pathPrefixes. The server packages add LegacyPathPrefixes / legacyPathPrefixes to the move's create and update models. Old links through a gateway or proxy.

6New in Content Search 1.43.0 and Storage & video 1.34.0

Existing integrations keep working; nothing changes for your API keys. Buckets you already have stay where they are.

  • Choose where a bucket lives. Mumbai, Hyderabad or Ashburn (US East), with more to follow. Processing (encoding, captions, search) happens in India for every location, the storage price is the same everywhere, and strict residency is offered for India locations only (strict_residency_not_available otherwise). Each location in GET /regions now says storageOnly, processingCountry and drTargets. Choose where files live.
  • Keep a disaster-recovery copy (Growth and Enterprise). Turn on a copy of a bucket in a second location under Storage buckets → the bucket → Disaster recovery, or with PUT /buckets/{id}/replication. Deleting a file deletes it from the copy too. Billed as DR copy storage and DR data transfer. New error codes: replication_not_in_plan, replication_region_not_available, replication_same_region, replication_residency_conflict, replication_already_on, replication_busy and replication_not_found. The server packages add GetBucketReplicationAsync, EnableBucketReplicationAsync and DisableBucketReplicationAsync, getBucketReplication, enableBucketReplication and disableBucketReplication, and get_bucket_replication, enable_bucket_replication and disable_bucket_replication. Keep a disaster-recovery copy.

7New in Content Search 1.42.0 and Storage & video 1.33.0

Existing integrations keep working; nothing changes for your API keys.

  • Connect your AI assistant by signing in. Add https://mcp.plycdn.com/mcp to Claude, ChatGPT, Cursor, VS Code or any assistant that supports sign-in, choose Connect, sign in to plycdn, then pick the organization and the access (Standard or Admin) and approve. Nothing to copy or paste. See and disconnect your assistants under Developers → AI assistants, in Connected assistants. Searching the docs still needs no sign-in at all, and connecting with an API key keeps working for assistants without sign-in and for CI. Connect with sign-in.
  • Your audit trail names the assistant. Changes an assistant makes show as, for example, AI assistant "Claude", instead of "An API key".
  • New error code dashboard_or_api_key_required (403). A signed-in assistant cannot make observability query keys; make them in the dashboard or with an API key. The server packages know the code from this release.

8Service update, 4 November 2026 (no package change)

Existing integrations keep working; nothing changes for the packages or your API keys.

  • A query stopped part-way says what it read. query_timeout and query_too_expensive refusals now carry stats (readBytes, readRows, elapsedMs), as a successful answer does, so you can count every read that is billed. What a query read.
  • What "Data read" counts, stated in full. Nothing changes in how it is measured. A read that stops part-way (out of time, over a limit, failed or cancelled) counts for what it scanned, as before; a query refused before it starts counts nothing. Every GB on the observability meters and in Volumes is 1,073,741,824 bytes. Volumes, and what is billed.
  • No hourly read limit on Growth and Enterprise. Organizations on Growth or Enterprise, and contracts made from them, may now read as much of their logs, metrics and traces as they need in an hour: read_quota_exceeded no longer comes back for them. Starter keeps its 1 TB an hour. Every other read limit stays: 30 seconds and 50 GB for each statement, eight at a time. The reference answer's limits gains hourlyReadLimited, which says whether the hourly allowance applies to your organization; the packages' types carry it from their next release. Read limits.
  • Longer queries and more of them at once, where your account needs it. We can now raise an account's (or one organization's) query ceiling up to 120 seconds a statement and its statements at once up to 64; ask us. A query may then ask for up to its ceiling with maxExecutionMs (up to 120,000), and the reference answer's limits.queryMaxExecutionMs and limits.concurrentStatements show your own values. Nothing changes without a raise: 30 seconds, eight at a time, and a maxExecutionMs above 30,000 is still validation_error (422). Read limits.
  • Exports, live tails and long queries through the API address are no longer cut at 30 seconds. A download runs for up to its 10 minutes, a tail for its 5. A transfer that is cut part-way (a release on our side, say) always ends as a broken transfer, never as a short file that looks whole: send the export again, or resume the tail from its last cursor.

9New in Content Search 1.41.0 and Storage & video 1.32.0

Existing integrations keep working; nothing changes for the packages or your API keys.

  • Query keys. Give a customer, a team or a partner a read-only key for your logs, metrics and traces, with an optional row filter the platform enforces on every query, export, live tail and trace view, however the statement is written. Filter on a resource attribute, a record attribute, a record attribute else the resource attribute, or the service; match all or any of your rules, with in or not in. You can have up to 200 keys per organization. Make them under Observability → Query keys or from code (createQueryKey, listQueryKeys, revokeQueryKey in the three server packages). A filtered key sees time only in the read statistics, so the numbers cannot reveal rows it may not read. The guide Give a customer, a team or a partner access to part of your logs goes through it step by step.
  • Archive status and daily manifest. Observability → Archive shows, for each day and kind of data, whether it was written (with its size, rows and SHA-256), empty or failed (with the reason). Each day also gets a manifest file in your archive bucket listing what was written. The same status is one call in the server packages (getObservabilityArchives).
  • A change. DESCRIBE of the log, span and metric tables keeps working. The merge() and loop() table functions are no longer available in observability queries, and queries can no longer read internal storage columns (_part, _part_offset and the like); only your data's own columns are readable.

10New in Content Search 1.40.0 and Storage & video 1.31.0

Existing integrations keep working; nothing changes for the packages or your API keys.

  • Every move in one place. Moving many containers? Storage → Moves from other clouds lists every bucket's move with its state, files copied, waiting, failed and cannot move, when an old link last arrived and when its read-only key expires. Narrow it by state, by Needs a look (files failed, files that cannot move, old links still arriving, key expiring soon) or by bucket name; the filters stay in the page address. The same list is one call in the three server packages (ListMigrationsAsync / listMigrations / list_migrations), with totals for your whole account, so a script can start and watch hundreds of moves. The guide adds moving many containers step by step, and what to do for containers your viewers reach at the storage account's own address. Your AI assistant can read the list too.
  • Fixed: the Old links panel of a move could fail to load.

11New in Content Search 1.39.0 and Storage & video 1.30.0

Existing integrations keep working; nothing changes for the packages or your API keys.

  • Move from another cloud, with no downtime. Point a file bucket at your Azure Blob Storage container, S3-compatible bucket or web server with a read-only key, add your address, and switch your DNS record when you are ready. plycdn serves a file from your old storage the first time it is asked for, keeps it at the same path, and copies the rest in the background from the day after you start serving. On a private bucket with plycdn links, Azure-signed links keep working, each checked with Azure, until you switch them off; the Old links panel shows which still arrive and from which sites. There, a third DNS record (_acme-challenge) makes your certificate ready before the switch. In the dashboard: Storage → Storage buckets → a bucket → Move from another cloud. The calls (StartMigrationAsync / startMigration / start_migration, the check, pause, resume, switching off old links and closing, and the move's keys and link usage) are in the three server packages, and the domain answer's records carry a purpose on the new record. New errors: migration_not_found, migration_already_running, migration_state_conflict, migration_not_closable, migration_source_unreachable, migration_credential_refused, migration_credential_too_broad, old_links_need_private_bucket and bucket_not_supported. Storage and delivery are billed as usual; copying in is not charged as upload. Step by step. Your AI assistant can read a move's progress, its keys and its link usage.

12New in Content Search 1.38.0 and Storage & video 1.29.0

Existing integrations keep working; nothing changes for the packages or your API keys.

  • Know what to store, and play on your own domain. Every video now carries playbackPaths: the fixed paths it plays from (/v/<videoId>/master.m3u8, and its poster, scrubbing previews and download once it has them). Keep the video's id in your application and sign a link for each viewer. Add a hostname of yours to a video library as a custom domain and sign on it: hostname on POST /sign and POST /playback-sessions, the signers' hostname option (PlycdnVideoSignOptions in .NET), and the players' hostname input. Every address in the answer is then on your hostname. In the dashboard, a video's Playback panel shows its ids, paths and address on each hostname, with Copy test link, and a library's Settings shows its custom domains. Your AI assistant can make a video's test link too. Step by step.

13New in Content Search 1.37.0 and Storage & video 1.28.0

Existing integrations keep working; nothing changes for the packages or your API keys.

  • Sign in to the dashboard with a passkey. Use your fingerprint, face, screen lock, phone or a security key instead of typing your password. Add one in Account → Sign-in & security, then choose it from the email field on the sign-in page or with Sign in with a passkey. Your password still works. You get an email when a passkey is added or removed, and when your password is changed after a reset. Changes appear in Change history. Passkeys are for people in the dashboard: no API key or package reaches them. Step by step.
  • Show any file in your application. <plycdn-file-viewer> (Angular, @plycdn/angular/storage) and <PlycdnFileViewer> with useFileViewer (React, @plycdn/react/storage) show one of your stored files in place: video and audio play, images show, and PDFs, Word, Excel, CSV and text open on the page, with a library video playing in the plycdn player. Content search gets <plycdn-content-file-viewer> and <PlycdnContentFileViewer> with useContentFileViewer, which open a result's file at the match (the page, row, line or paragraph) and also work for files that were never indexed. Anything else, PowerPoint and ZIP files included, ends on a file card with Download. HTML, XML and code are shown as text, never rendered. In the dashboard, Storage → Storage buckets → (a bucket) → Files → Preview and Content search → Library → (a file) → Preview use it. Step by step.
  • Large files open at once. Text, CSV, Word, Excel and PDF files are read a part at a time as the person scrolls, so a 2 GB log or a 300 MB spreadsheet opens, and the view ends with The rest of this file isn't shown here. and Download at its limit. A PDF saved "optimised for the web" opens page by page; one built with a flat page index is read almost whole. A workbook shows its sheets as tabs, and a search result opens the sheet that matched.
  • Fixed: a PowerPoint or ZIP file in <plycdn-content-document> now shows a file card with Download instead of failing, and a Word or Excel file that inflates far beyond its size is refused instead of exhausting the browser. A request for a document no longer drops the page's referrer.
  • [knownPdf] on <plycdn-content-document>. An optional input: true for a PDF the service indexed, so the browser's own viewer stays the fallback on storage that doesn't allow cross-origin reads. <plycdn-content-playback> sets it from the result, so its PDFs open as before.
  • New viewer words in ContentLabels. ContentLabels gains optional viewer words (restNotShown, loadedPart, pictureNotShown, the card sentences and the type words), so a label set written before the viewer still compiles.
  • Pages on your site can read part of a file. Every bucket that allows an origin now also allows the range request header and exposes content-range, accept-ranges and content-length, with nothing to set. Viewers on your pages load only the part of a file they show. Your own CORS settings are unchanged and read back as you set them. CORS.
  • <plycdn-content-document> (Angular): undocumented members removed. Its inputs, (unavailable) and labels are unchanged, and zoom(), fitWidth(), goToMatch(), mode, pageCount, currentPage, matchPage and zoomPercent are kept. The internal members loading, lines, sheets, blocks, frame, marked, trackPart, isHit, isRow, isBlock and pdfPages are gone, and mode now reads 'pdf-frame' where it read 'pdf'. If your code reached into the component for any of these, read the file's state from (stateChange) on <plycdn-content-file-viewer> instead.

14New in Content Search 1.36.0 and Storage & video 1.27.0

Existing integrations keep working. One answer changes: a query refused because your organization already had eight statements running, or had used its hourly read allowance, was answered query_too_expensive (400, not retryable); it is now query_busy or read_quota_exceeded (429, retryable, with Retry-After). If your code matched the old code to give up, match the new ones to wait and send again. The Node.js and Python packages wait for a short Retry-After and send again for you; a long one (the hourly allowance) is raised at once with retryAfterMs / retry_after set, so you decide. The .NET package throws, and you wait RetryAfter. Earlier Node.js and Python packages wait up to about two minutes before raising read_quota_exceeded: if that matters for interactive calls, set maxRetries (Node.js) or max_retries (Python) to 0. Observability's Volumes also steps down on the day you move to 1.36.0: see the first item below.

  • Volumes counts what is billed, and shows what arrived. The tag we add to each row to know which organization it belongs to is no longer counted in the stored size. Days before the release are not changed and nothing is credited for them, so the daily figure steps down on the release day. Each day also carries receivedGb, the size of your requests as you sent them (not billed), and a worked example of why stored is larger than received: a span of about 120 bytes as sent is about 450 bytes stored. In the dashboard, Observability → Volumes shows Received, Stored, Kept and Read. Volumes, and what is billed.
  • Today is visible, and a day says when it is final. Ask for from and to as today to see what has arrived; each day has complete, which is true once every figure has counted it. Today never is.
  • Refused and removed data is counted. Each day lists rejected by signal and reason (older_than_retention, ingest_invalid, ingest_too_large, not_stored), with a table in Observability → Volumes that you can narrow by signal. Records stamped older than your retention that arrive are now included in the day they arrived. A request is answered once it is safely queued, so there is no answer per record: why.
  • Check a query before you run it. POST /observability/validate (ValidateTelemetryQueryAsync, validateTelemetryQuery, validate_telemetry_query) tells you whether a SQL or KQL query would run and what it would return, with the line and column of a mistake, without reading your data. Step by step.
  • A reference your tools can read. GET /observability/reference (GetTelemetryReferenceAsync, getTelemetryReference, get_telemetry_reference) lists every KQL operator, function and aggregation with an example, the tables and columns, the query errors and the limits. The same lists are in the guide's KQL reference.
  • See what a query read. Every query, check, trace and live-tail answer carries stats (readBytes, readRows, elapsedMs). A download answers with X-Query-Id; FindSlowQueryAsync, the queryId option of getSlowQueries and the query_id argument of get_slow_queries look it up, and every Slow queries item carries its queryId. What a query read.
  • Give a query less time. maxExecutionMs (1,000 to 30,000; 1,000 to 600,000 for an export) on a query or an export, or SETTINGS max_execution_time in SQL. Time limits for a query.
  • The read limits are written down, and they hold across all our servers. Eight statements, four live tails and two exports at a time for your organization, wherever they run; 30 seconds for a query and 10 minutes for an export; 100,000 rows and 64 MiB for a query answer; 50 GB for one statement and 1 TB an hour. Read limits. A call made too often through the plycdn address now answers { "error", "code": "rate_limited", "retryable": true }.
  • A trace lookup that is too slow answers query_timeout instead of a gateway timeout, and a query that overruns its limit is answered query_timeout by us.
  • Compression is documented. OpenTelemetry accepts gzip and zstd over HTTP and gRPC, JSON lines gzip; the 5 MB limit is on the bytes as sent, and a request may be up to 20 MB once decompressed. Every row has _ingested, the moment we stored it. Sending compressed data.
  • Usage figures no longer all fail together. If one of the three observability figures cannot be worked out on a day, the other two are still recorded.
  • Step by step. New sections of Observability, step by step: check a query, keep it within a time limit, see what it read, and check what you sent and what was refused.

15New in Content Search 1.35.0 and Storage & video 1.26.0

Existing integrations keep working without a change: the watermark and logo are off until you switch them on.

  • Viewer watermark. Draw each viewer's name, email, phone, your ID for them, their address and the date and time over your videos, so a screen recording or a phone filming the screen shows who it came from. Choose what it shows, in what order, its opacity, size, colour, how often it moves and where, under Account → Branding → Watermark, for every library, or give one library its own under its Protection tab. Your server names the viewer with PlaybackWatermarkFor, playbackWatermarkFor or playback_watermark_for; the React and Angular players draw it with no change to your pages. Show draws it in plycdn's players; Require plays the library's videos only in players that draw it, with 10-minute links renewed as the video plays, and stops playback when the watermark is removed or hidden. Included in every plan. See Viewer watermark and your logo.
  • Your logo on every video. Under Account → Branding → Player logo: a corner, a size, an opacity and a distance from the edges. It is drawn by the player over the picture, never part of the video file and never a link. A library can leave it off.
  • Branding has its own place. Account → Branding holds the player logo, the watermark and the watch page's look, for every account, with or without live streaming; Live streaming → Watch page keeps your watch domains. Live streams and share links show the logo and a watermark of your fixed text and the date and time.
  • For players of your own. PlycdnWatermark.Render, render and render give what to draw from a playback session, with the check-in and the renewal; PlayerMark draws and defends the marks in the browser. New codes: watermark_required, watermark_proof_invalid and watermark_tampered; signing_profile.changed gains the reason watermark_changed.
  • Plan ad breaks, list them and cancel them. A live stream's Overview has an Ad breaks panel: insert a break now or plan one up to 24 hours ahead with your own break id, see the stream's breaks, and cancel a planned one. From code, GET and DELETE /live-inputs/{id}/breaks (ListAdBreaksAsync, CancelAdBreakAsync, listAdBreaks, cancelAdBreak, list_ad_breaks, cancel_ad_break), with the new codes ad_break_not_cancellable and ad_break_not_found. See Ad breaks.
  • Live settings follow your plan. Ads, captions and backup ingest in a stream's Settings are offered when your plan includes them and say how to ask when it doesn't. Content protection is a choice of your account's protection setups (GET /live/protection-setups, ListProtectionSetupsAsync, listProtectionSetups, list_protection_setups), and Use the source picture when it is H.264 is a switch under Quality. Step by step: Live captions, Content protection, Passthrough and Backup ingest.
  • Your plan, in one place. Account → Usage → Plan lists what your plan includes, what is available on request, its limits, your live streaming limits, and the watermark and player logo that every plan includes. GET /plan carries the live safety caps in limits, and live_backup_ingest or live_redundant_transcoding as false when your plan leaves that setting out; the server packages read them by name: ContentPlan.Limit(PlanLimits.LiveConcurrentStreams), planLimit(plan, PLAN_LIMITS.LiveConcurrentStreams), plan_limit(plan, "live_concurrent_streams").
  • Easier to find. Every report and settings screen has a How it works link to its guide; Delivery analytics, the breakdown, Search analytics, Usage and Library health show the filters in use as chips you can clear; Library health can be narrowed by kind of file and what is wrong; the libraries list marks serial playback; and the dashboard's search finds ads, captions, protection, serial playback, country and site rules by the words you use. New guides: Your watch page and Restrict where files play.
  • AI assistants. The assistant's discover now prefers operations that match every word you ask about, finds them by the words customers use (ads, watermark, logo, serial, geo, share link), and lists the operations it can't run for you, with the reason.

16New in Content Search 1.34.0 and Storage & video 1.25.0

Existing integrations keep working without a change.

  • Signing settings are kept until they change. The server packages fetch a bucket's or library's signing settings once and keep them, instead of fetching them again every 10 minutes. They check at most once an hour, and that check downloads nothing when nothing changed, so your servers call plycdn less, and after a key rotation new links are signed with the new key without a restart. A rotation made from the same server applies at once (in Python, through the same client object); SigningProfileCacheFor, signingProfileCacheMs and signing_profile_cache now set how often the check runs (1 hour by default, at least 5 minutes). See When the signing settings change.
  • New webhook event: Signing settings changed (signing_profile.changed). It is sent when a bucket's or library's signing key, domains, link settings or serial playback change. Give the packages' webhook receivers your signer and a change made in the dashboard or from another server applies at once (in ASP.NET Core, Plycdn.AspNetCore.Video's verifier does it with nothing to add; with the Storage package alone, call HandleWebhook); with an event you read yourself, call HandleWebhook, handleWebhook or handle_webhook. See Keep your signers current.
  • Links to stored files are signed in your server. The browser sign route (api/plycdn/v1/sign) signs a link to a stored file (objectId) of a private files bucket without calling plycdn. Links by file name, videos and live inputs work as before.
  • Refused links refresh the settings. When a link is refused, the React and Angular players ask your backend again with refused: true, and the sign route refreshes that library's signing settings (at most 3 times in 10 minutes). PlycdnPlaybackService.link takes it as an optional fourth argument, { refused }. The players send it only to a sign route that says it takes it (the server packages answer every sign with Plycdn-Sign-Refused: accepted), so a sign route of your own that passes the browser's request on to plycdn keeps working unchanged.
  • Every ID in full, with Copy. The organization ID is at the top of every dashboard page, next to its name, and every bucket, video library, video, live input, credential, webhook and domain shows its ID in its list and on its page, with a Copy button. The code sample on a bucket's Signing tab has your IDs in it. See Find your IDs.

17New in Content Search 1.33.0 and Storage & video 1.24.0

Existing integrations keep working without a change.

  • See where storage requests come from. The request log (Storage → Request log) has a Client IP column: the address each request came from. Type an address in Client IP to see only its requests, or open a request and choose Only this address. Requests logged before this release show no address. See Move your files to plycdn.
  • From your code: each item of GET /storage/requests has clientIp (an IPv4 or IPv6 address, or null), and clientIp narrows the log to one address. The server packages have it on StorageRequestQuery.ClientIp and StorageRequest.ClientIp (clientIp in Node.js; client_ip in Python).
  • Logthread. If you send your storage request log to Logthread, each record now carries the address as client.address, and fields with no value are left out rather than sent empty.
  • Delivery by region. Delivery from North America and the United Kingdom is now always counted in its own region.

18New in Content Search 1.32.0 and Storage & video 1.23.0

Existing integrations keep working without a change.

  • Send plycdn activity to Logthread. Connect a Logthread workspace in Developers → Integrations, then choose what to send: your storage request log, video and live events, and daily delivery totals. You search them in Logthread next to your applications' own logs, chart them and get alerts. Each organization and sub-organization connects its own workspace. See Send plycdn activity to Logthread.
  • From your code: GET, PUT and DELETE /integrations/logthread with an admin key. The server packages have GetLogthreadConnectionAsync, ConnectLogthreadAsync and DisconnectLogthreadAsync (getLogthreadConnection, connectLogthread, disconnectLogthread in Node.js; get_logthread_connection, connect_logthread, disconnect_logthread in Python).
  • New error codes. logthread_key_rejected (422), logthread_unreachable (503) and logthread_connect_failed (422).

19Docs update (no package change)

  • Observability: for your applications' logs, metrics and traces we now recommend Logthread, our observability product, billed on its own. plycdn's built-in observability stays available; its pages say so at the top.

20New in Content Search 1.31.0 and Storage & video 1.22.0

Existing integrations keep working without a change; every key you already have keeps reaching your whole account.

  • API keys for one organization. When you create a key in Developers → API keys, choose its Reach: only the organization you have selected (your main organization or one sub-organization), or the whole account. A key for one organization needs no X-Organization-Id and can't see or change the others. New keys start as "Only" the selected organization. See Give a key to one organization only.
  • A new error code. organization_out_of_key_reach (403) means the key can't reach the organization you named.
  • An account summary. GET /account says which organization a key acts for, which organizations it can reach, your regions with their endpoints, and the products in your plan. The server packages have GetAccountAsync, getAccount and get_account.
  • Connect your AI assistant. plycdn now has an MCP server at https://mcp.plycdn.com/mcp. Claude Code, Cursor, Windsurf, VS Code and Claude Desktop can search the docs, read your buckets, usage and request log, check that your integration works, and, when you agree, create a bucket, upload a check file or make a signed link. In the dashboard: Developers → AI assistants. See Connect your AI assistant. A new code, too_many_tool_calls, means one request asked for more than five tools at once.
  • Install the SDKs with npm, NuGet and pip. Every package is now on the plycdn feeds: npm.plycdn.com, nuget.plycdn.com and pypi.plycdn.com, so npm install @plycdn/node, dotnet add package Plycdn.AspNetCore.Storage and pip install plycdn --index-url https://pypi.plycdn.com/simple/ work and updates arrive the usual way. See Install the SDKs. Downloads stay in Developers → SDKs.
  • llms.txt. AI tools that read https://plycdn.com/llms.txt find our docs without the connector.
  • Every package now states its terms: for use with a plycdn account under the plycdn Terms of Use.

21New in Content Search 1.30.1

No change to the packages' calls and nothing to edit in your integration; the storage and video packages stay at 1.21.0.

  • A \ in a file name is now treated as a folder separator, so Windows paths upload as they are.

22New in Content Search 1.30.0 and Storage & video 1.21.0

Existing integrations keep working without a change.

  • A request log for your storage endpoints. Every request your tools and applications make to the S3 and Azure endpoints is recorded, including the ones plycdn refused, so you can see why a copy or a download failed: the time, the operation, bucket and file, the result and error code, bytes, duration and request ID. Secrets, signatures, SAS tokens, query strings and file contents are never recorded. It is kept for 30 days (plycdn can extend it for your account up to 365 days, for example while you migrate) and is not billed. In the dashboard it is Storage → Request log, and a Requests action on buckets, video libraries and storage credentials opens it already narrowed. Over the API it is GET /storage/requests; the server packages have GetStorageRequestsAsync, getStorageRequests and get_storage_requests, and calls that follow every page. See Check that requests reach plycdn.
  • A new error code. request_log_unavailable (503, retryable) means the log could not answer just now.

23New in Content Search 1.29.1

No change to the packages' calls and nothing to edit in your integration: install the Content Search packages at 1.29.1 together. In the service, video processing scales out further, so large batches of uploaded videos become ready sooner. The storage and video packages stay at 1.20.0.

24New in Content Search 1.29.0 and Storage & video 1.20.0

Existing integrations keep working without a change.

  • Watch time per viewer. Name the viewer on your server with your own id (a learner, an employee, a member, a patient, a customer) and plycdn keeps how long they really watched, in every video library, not only those with serial playback. Pauses don't count; watching again does. Anonymous viewers are not tracked. See Watch time per viewer, step by step.
  • A report per viewer and per video. GET /libraries/{libraryId}/watch-time answers totals and a page of viewers or videos for a date range, one viewer's videos or one video's viewers, and a CSV download. In the dashboard it is Video streaming → Watch time, and each library's Watch time tab. It is in every plan. The server packages have GetWatchTimeReportAsync, getWatchTimeReport and get_watch_time_report, and DownloadWatchTimeCsvAsync, downloadWatchTimeCsv and download_watch_time_csv.
  • Progress carries it too. A viewer's progress answers add watchTimeSeconds and firstWatchedAt, a started session adds progressIntervalSeconds and watchTimeSeconds, and the players' onProgress and (progress) events add played. A video's Viewer progress panel in the dashboard now shows in every library.
  • Ordinary libraries save every two minutes. When your server names the viewer, the sign route of the server packages now starts a session in any library, so watch time needs no code beyond naming the viewer. Libraries with serial playback still save every minute.
  • Step-by-step guides. A new group of guides takes each task from start to finish, with where to click in the dashboard, the code for every server package and the players, how to check it works, and what to do when it doesn't: serial playback, watch time, live streaming, playing video, delivering files and video, adding search, webhooks, your team, keys and customers, moving your library and observability.
  • Live streaming has its own place in the dashboard. Live streaming → Live streams lists every stream of every library; New live stream makes one in a new library made for it or in a library you have. Libraries that hold streams show Live, searchable ones Searchable, and the New library dialog can make a library searchable. The Overview has a Get started row, and each feature's page links to its guide with How it works.

25New in Content Search 1.28.0 and Storage & video 1.19.0

Existing integrations keep working without a change.

  • Query your telemetry in KQL. Queries and exports take dialect: "kql" to be written in Kusto Query Language, over the same tables, with the same limits and the same isolation as SQL.
  • Your slow-query log. GET /observability/slow-queries lists your organization's slowest recent queries with their duration, rows and bytes read, and the dashboard shows them under Observability → Slow queries.

26New in Content Search 1.27.4 and Storage & video 1.18.1

Existing integrations keep working without a change, with one exception noted below for live captions.

  • Live captions are made in India. Captions for live streams are now made by the same speech service as recordings, in India, in English and the 22 scheduled languages of India. A caption language outside that list is refused.
  • Deleting removes more. A deleted video's title and labels, and a deleted file's labels and stored headers, are removed with it. Ids, sizes, dates and the audit record stay for billing and security.
  • Uploads ride out a busy moment. The resumable uploader in the storage and video packages asks again when an upload is still being completed, instead of giving up.
  • Clearing a label works from Node.js. Setting a metadata value to null in updateObject now removes that key.
  • The dashboard works everywhere. Posters, search-result playback and recording links play on libraries with serial playback or address binding; the dashboard preview plays on every library; editing a sub-organization no longer makes a copy of it; and more actions explain exactly what to fix when they cannot go ahead.
  • Billing is worked out more precisely. Late usage of a closed month, prices shown with all their decimals, and a search whose answer could not be written is not charged.

27New in Content Search 1.27.3

Existing integrations keep working without a change.

  • Missing files answer with our own short error. A request for a file that is not in a bucket now answers {"code": "not_found"} with the same status as before (404), on the bucket's address and on custom domains alike.
  • Custom domain certificates are ready sooner. Right after your DNS records are found, the certificate is asked for again within minutes rather than every half hour while it is being issued. "Check now" on a custom domain works again.
  • Watch page and share-link settings save. Saving the watch page (including the logo size) and share links with their own title, description, layout or messages no longer fails.

28New in Content Search 1.27.2

Existing integrations keep working without a change.

  • Processing capacity grows reliably when many files arrive at once. When uploads queue up, more processing capacity is now added automatically and dependably, and released again once the queue is clear, so large batches finish sooner.

29New in Content Search 1.27.1

Existing integrations keep working without a change.

  • Long syslog messages are kept whole even when the network pauses. Continuation lines of a multi-line message that arrive after a short pause are now stored as further parts of the same message instead of being lost.

30New in Content Search 1.27.0 and Storage & video 1.18.0

Existing integrations keep working without a change.

  • Observability volumes per signal. Each day in GET /observability/volumes (and in the Storage & video packages and on Observability → Volumes) now also shows logs, metrics and traces separately: records, data received and data kept. The totals are unchanged.
  • More of what you send is kept as you meant it. Prometheus series named *_total sent without a type are kept as counters, so rates work on them. Records sent without a time are kept with the time they arrived. JSON lines keep levels, times and fields we could not use as they were, as attributes. Syslog messages that run over several lines are kept as one message (very long ones in numbered parts). Identical records sent separately are all kept.
  • Cleaner stored data. Stored telemetry now carries only one reserved attribute, plycdn.org. The guide lists what is kept, changed or not stored on every way of sending, including how to send Prometheus native histograms.
  • A full method reference. Method reference lists every call with its name in each package.

31New in Content Search 1.26.0 and Storage & video 1.17.0

Existing integrations keep working without a change.

  • No limits on how much you send. Logs, metrics and traces are no longer limited per ingest key or per address. Send as fast as your systems produce data. If we ever fall behind, your data arrives later than usual, but it is never lost.
  • Every accepted record is stored safely first. An answer of success (or a Fluent forward acknowledgement) now means your data is already held safely; a 503 observability_unavailable means send it again, which OpenTelemetry exporters and Fluent Bit do on their own. Syslog is paced instead of dropped: when we are briefly behind, we read more slowly and your agent keeps the messages until we catch up. Fluent forward no longer limits connections per key, so many hosts can share one key.
  • 429 rate_limited on sending now only answers many attempts from one address with a missing or wrong key.
  • Clearer dashboard figures. The overview counts each file once, by its latest processing: a file that failed before and is ready now no longer shows as needing attention. A file whose latest re-index failed but whose previous version is still searchable shows as ready, with a note. Sizes on Storage and Usage use the same units as your limits and invoices, and Observability → Volumes shows data read in GB.
  • New field previousVersionSearchable on attention items and on a file's latest job, in the API and in the server SDKs.

32New in Content Search 1.25.0 and Storage & video 1.16.0

Existing integrations keep working without a change.

  • Manage your account from the Storage & video packages. Sub-organizations (list, create, update, deactivate), the account audit, quotas and monthly statements are now in plycdn (Python, plycdn.account), @plycdn/node (@plycdn/node/account) and Plycdn.AspNetCore.Storage (PlycdnOrganizationApi). If you use observability, storage or video only, you no longer need a Content Search package to create a sub-organization for each of your customers. The Content Search packages keep the same calls under the same names.
  • Syslog over the rate is dropped. Syslog cannot be told to slow down, so messages over a key's 1,000 a second are dropped rather than refused; keep each key's syslog under that, or send through Fluent forward or OpenTelemetry.
  • Long documents are read with their layout. A document of several hundred pages is now read as it is laid out, like a short one.

33New in Content Search 1.24.0 and Storage & video 1.15.0

Existing integrations keep working without a change. Nothing needs switching on.

  • Documents are read as they are laid out, in every language. PDFs, PowerPoint decks and Word documents are read in reading order across columns, with page headers and footers left out. Passages follow the document's own structure, which is recognised for you: clauses and definitions in a policy or contract, sections in a report, a question with its answer, a slide.
  • Tables stay tables. Each table row is a result of its own, and it names the row and the cell that matched (location.cells, location.matchedCellId); the Angular document viewer marks both on the page. This covers PowerPoint and Word tables too.
  • Scanned pages are read. A page is listed in warnings as ocr_required:page:N only if it could not be read either, and a PDF fails with ocr_required only if no page could be.
  • New location fields, all optional: type, headingPath, pageEnd, bbox, tableId, cells, matchedCellId. page, charStart and charEnd are unchanged. Documents indexed before this keep their passages until they are indexed again; ask us to read a library or a list of documents again.

34New in Content Search 1.23.0 and Storage & video 1.14.0

Existing integrations keep working without a change.

  • Observability is available on every plan. Keep the logs, metrics and traces of your own applications in plycdn, search them with SQL, download them, follow your logs live and open one trace. Your first log line can be in place in about five minutes: the quick start uses the dashboard (Observability → Settings, then Ingest keys, then Volumes) or a terminal. Starter keeps data for up to 30 days, Growth for 90 and Enterprise for 365; you choose 7, 15, 30, 90, 180 or 365 days within that, and 30 unless you say otherwise.
  • Send with what you already run. OpenTelemetry (HTTP on port 4318, gRPC on 4317), Prometheus remote write 2.0 (Prometheus 3.0 or later), syslog over TLS (rsyslog, syslog-ng; port 6514), the forward output of Fluent Bit and Fluentd (port 24224), and plain JSON lines. Applications in .NET, Node.js and Python can be instrumented with environment variables and no code change: OpenTelemetry with no code change. Sending uses an ingest key (plyt_…), which can send data and nothing else.
  • In your region, for the time you choose. You choose the region once; data stays there. An optional nightly Parquet copy is written into a bucket of your own.
  • One organization per customer of yours. Each sub-organization has its own region, retention, ingest keys and rows: One sub-organization per customer of yours.
  • In the dashboard. An Observability section with Settings (switch on, retention, archive, and a copy-ready "Send your first data" panel for curl, OpenTelemetry, Prometheus, Fluent Bit and rsyslog), Ingest keys (a test command with your new key) and Volumes (what arrived, what is kept, what was read). Data received, kept and read are counted and shown on your usage and statement; the first 50 GB of data received each month is included.
  • In the packages. PlycdnObservabilityApi in .NET, Node.js (@plycdn/node/observability) and Python (plycdn.observability, with an async twin) cover every call: settings, ingest keys, QueryTelemetryAsync (queryTelemetry, query_telemetry), ExportTelemetryAsync, TailLogsAsync (tailLogs, tail_logs), GetTraceAsync and the volumes. A download or a live tail cut part-way raises an error rather than ending as if it were whole.
  • Odia on screen. Text shown on screen in videos and slides is now read in Odia too, alongside the other languages.
  • New codes. query_invalid, query_too_expensive, query_timeout, trace_not_found, tail_busy, observability_not_enabled, observability_region_fixed, observability_unavailable, region_not_available, ingest_key_limit_reached, ingest_key_not_found, ingest_key_invalid, ingest_wrong_region, ingest_invalid, ingest_too_large and archive_bucket_invalid are in the API reference.

35New in Content Search 1.22.0 and Storage & video 1.13.0

Existing integrations keep working without a change.

  • Many videos at once. A video library takes files from S3 and Azure tools (rclone, the AWS CLI, AzCopy, the Azure CLI) with a storage credential, into sources/ and folders inside it, and the server packages upload a whole folder with one call that resumes and skips what is already there: PlycdnVideoUploader with UploadVideoAsync and UploadVideosAsync (uploadVideo, uploadVideos in Node.js; upload_video, upload_videos in Python). Each file becomes one video, titled from its title metadata or its file name. A key a video already came from is refused with video_source_exists, so nothing is processed twice. GET /libraries/{id}/videos?sourceKey= and GetVideoBySourceKeyAsync (getVideoBySourceKey, get_video_by_source_key) find the video a file became; ContinueUploadAsync (continueUpload, continue_upload) sends an upload session you started, such as a created video's. On the dashboard: Bulk upload on each library. See Moving many videos in.
  • Download the packages from your dashboard. Developers → SDKs has the latest release of each product line, where each language is its own download with its SHA-256, for anyone signed in to your organization.
  • A clearer dashboard. The sidebar's sections open and close, it can show icons only ([), and on a phone it is a menu.

36New in Content Search 1.21.1 and Storage & video 1.12.1

Existing integrations keep working without a change. Install the packages at the new versions together.

  • Smaller videos. Videos encoded from now on take much less space and fewer bytes to stream at the same picture quality. Lectures and screen recordings typically shrink by most of their size, and talking heads by about a third to a half; fast-moving video stays as it was. A quality is never made larger than the uploaded file itself carries. Videos already encoded keep their files until they are encoded again.
  • Players wait for play. A player no longer downloads any video until the viewer presses play (or autoplay starts it); the poster and the length still show. Pages with many players use far less.
  • A failed re-encode loses nothing. A video that is encoded again and fails keeps playing what it had and keeps its original, and is not charged. errorCode says what went wrong. See Encoding a video again.

37New in Content Search 1.21.0 and Storage & video 1.12.0

Existing integrations keep working without a change.

  • Serial playback. Turn it on for a video library (Growth and Enterprise) and no viewer can reach a moment of a video sooner than that much time after they started watching, across every visit with the same resumeId, whatever runs in their browser. Time paused or unattended counts as time watched, so pair it with attention monitoring if you need proof that someone was watching. Your server names each viewer (PlaybackResumeIdFor, playbackResumeIdFor, playback_resume_id_for); the players need no change. POST /playback-sessions (StartPlaybackAsync, startPlayback, start_playback) and each viewer's progress to read, list and start over (GetPlaybackProgressAsync, ListPlaybackProgressAsync, ResetPlaybackProgressAsync and their Node.js and Python twins). On the dashboard: the library's Serial playback switch and each video's Viewer progress. A library made before plycdn links can't turn it on yet: contact support and we will move it. New answers: serial_playback_requires_session, serial_playback_needs_plycdn_links, serial_playback_unavailable, serial_playback_busy, progress_not_found and playback_session_expired. See Serial playback.
  • Where a player is. Every player reports its position about once a second (onProgress in React, (progress) in Angular), in every library. These positions are for convenience; for anything you enforce, use the watchedSeconds plycdn keeps.

38New in Content Search 1.20.0

Existing integrations keep working without a change. Most additions below are things you send when you want them; the exception is answers, which now draw on every document in their scope (the "Answers across documents" bullet).

  • Proof of deletion. After a file's deletion finishes, GET /deletion-receipts returns a receipt signed by plycdn: what was removed, when, and the date its last copies leave our encrypted backups. Anyone can check it against our published keys, without your account; a key we retire carries a validUntil and a receipt signed after it is not valid. ListDeletionReceiptsAsync, listDeletionReceipts, list_deletion_receipts; the dashboard's Audit log lists them. See Proof of deletion.
  • Limits for each sub-organization. Your account can cap each sub-organization's searches and written answers a month and the content it keeps searchable, from code (SetOrganizationQuotasAsync and its Node.js and Python twins, with an admin key) or the Sub-organizations page. Over a limit only that sub-organization's new work is refused, with the existing quota_exceeded; a sub-organization sees its own limits in organizationQuotas on GET /usage/quotas. See Limits for each sub-organization.
  • Passages around a result. context on a search returns the passages before and after each result, and GET /assets/{id}/passages?around= returns them, or the result's whole section, on their own. Every result carries passageId to ask with.
  • A file's indexed text. GET /assets/{id}/passages reads it back exactly as indexed, in order, a page at a time.
  • Quotes checked word for word. "quotes": true on an answer quotes each claim's passage and checks every quote against it: verified, and each citation's quotes.
  • A time limit. timeoutMs on a search or an answer returns the best result in time, marked partial, instead of failing.
  • Ranking for one request. ranking.keywordWeight balances wording against meaning; ranking.rerank: "off" is faster.
  • Answers across documents. An answer draws on every document in its scope that has a relevant passage, offers each document's best passage first, and says which document each fact comes from.
  • New in every Content Search package: GetPassagesAsync, getPassages, get_passages, passages(), usePassages, and the new fields. One new error code: 409 content_not_ready.

39Service update: answers cite only what they list

Every number in an answer now refers to one of the citations returned with it; a reference to anything else is removed before the answer reaches you, in the JSON answer and in a streamed one alike. No package changes and there is nothing to edit.

40New in Content Search 1.19.0 and Storage & video 1.11.0

Nothing changes for existing integrations, and there is nothing to edit. Two new server packages join each line, so a backend written in Node.js or Python has what an ASP.NET Core backend has:

  • Content Search: @plycdn/content-search-node (Node.js 18 and later) and plycdn-content-search (Python 3.9 and later).
  • Storage & video: @plycdn/node (Node.js 18 and later) and plycdn (Python 3.9 and later).

Each carries every server call of its .NET counterpart, under the same names in the usual style of the language (registerAsset in Node.js, register_asset in Python); the routes your Angular and React frontends call, with the same rules (every route is refused until your code allows it, and the organization always comes from your signed-in user); the source-access callback for Content Search; webhook verification for video; signed links and playback links made on your server; server uploads that resume; and clients for the S3 and Azure tools. Retries and timeouts are built in: reads are repeated after a temporary failure, and a call that creates something is repeated only with an idempotency key, so nothing is created twice. Answers can be read as they are written. The Node.js packages work from CommonJS and ES modules and carry TypeScript types. The Python packages come as a synchronous and an asynchronous form with the same methods, and the routes are for your FastAPI application. The guides are plycdn for Node.js and plycdn for Python.

41Updates since 1.18.0

No package version moved and there is nothing to install or edit.

Faster search. Searches and written answers return sooner, and searches in large libraries no longer slow down as the library grows. Results are the same as before.

Documentation. The guides were corrected and extended. PDFs are described page by page and in passages of about 200 words, and a PDF with some image-only pages is described as it behaves. Search is described as meaning and wording together, with a re-ranking step. Answers are written in the language of the question. Restricting a file is Enterprise while sending principals works on every plan. Registering the same file again, sub-organizations as separate tenants, and the dashboard's Search page are explained. The OpenAPI file is published at /docs/openapi.json, and Evaluating plycdn lists each capability as yes, partly or no.

42New in Content Search 1.18.0 and Storage & video 1.10.0

No breaking change and nothing to edit: nothing new to call. Install the new package versions together, as before.

Identical videos are recognised. A video in a video library identical to one uploaded from 1.18.0 on and already searchable in your account - the same original, or the same MP4 download - is made searchable from that work, without being transcribed again. We recognise it by the checksum we take when it is uploaded; there is nothing for you to send (Make a video library searchable).

Large recordings without a checksum are compared by size. A recording registered by address without contentSha256, of exactly the size of one already indexed in your account with the same options, is read in full to compare, and an identical one is not transcribed again. It is charged as any identical copy is (Duplicate content).

  • Service: a video with several qualities finishes encoding sooner. A video whose upload is replaced by another upload to the same name before encoding finishes now ends with video_source_missing (upload it again), instead of being encoded from the newer file.
  • .NET, Angular, React: no code change; versions 1.18.0 and 1.10.0.
  • Dashboard: no change.

43New in Content Search 1.17.0 and Storage & video 1.9.0

No breaking change and nothing to edit: nothing new to call. Install the new package versions together, as before.

The same file twice is ready sooner. A file we recognise as identical, with the same options, to one already indexed anywhere in your account - another upload, another bucket, another sub-organization - is made searchable from that work: it is not transcribed or read for on-screen text again, it is ready once it is recognised (within seconds, or the time it takes to fetch a file we read in full first), and it is not charged again (/usage/assets shows it with no hours transcribed, 0 billed and duplicateOf; bytes fetched to recognise it show as fetched). A recording or video in a bucket made searchable is recognised by the checksum we verified when it was stored, and is not read again; videos in a video library are recognised only when read in full. A correction you make to one file's transcript stays with that file (Duplicate content).

Captioned recordings are searchable sooner. When captions were on for a stream in one language from start to end, its recording is made searchable from those captions: ready soon after the recording is, with passages timed to the caption line (Searchable recordings).

  • Fixes in this release: the transcript timing of a recording after a gap in the stream (a reconnect) is now exact, where it started early by the length of the gap (recordings finished after the update; earlier ones keep their timing); encoding a video again now makes every quality with the current settings (the higher qualities kept the earlier encode); a transcript correction keeps the file's warnings; re-indexing a file that never finished processing processes it afresh.
  • Service: an Idempotency-Key is remembered for 30 days.
  • .NET, Angular, React: no code change; versions 1.17.0 and 1.9.0.
  • Dashboard: no change.

44New in Content Search 1.16.0 and Storage & video 1.8.0

No breaking change. If your site sends a content security policy and you make a bucket or library searchable, allow its delivery address there, and allow your site in the bucket's cross-origin settings to show documents in place (Your own plycdn storage). Install the new package versions together, as before.

Your own storage, searchable with one switch. Turn on search for a bucket or a video library and its files, uploaded videos and live recordings become searchable by what is in them, with results that open the exact moment or page. No keys, no addresses, no connector to set up; new and replaced files are added as they arrive, renamed ones are never processed twice, deleted ones leave search at once. It is charged on your search usage like any file you add; the switch is free (Make a bucket searchable).

  • Service: search on buckets and libraries (create and change), GET and DELETE /buckets/{id}/search (the DELETE needs an admin key), each file's and video's search state (videos and video.* events carry it), source on files, results and citations, ?bucketId= on /activity, GET /assets/{id}/playback. New code: asset_not_playable. A file whose search entry you deleted stays out of search (removed) until it is uploaded again. A skipped file is checked again by itself once its reason goes away. Searchable recordings now need video and Content Search in your plan, as every plan with video has.
  • Fixes in this release: reading a file's source or playback link as a reader now applies the reader's permissions (a file the reader may not see answers not found), and a resolver that returns no permissions no longer opens restricted files. A failed recording registered by its input is now withdrawn rather than left failed.
  • .NET: BucketSearch, GetBucketSearchAsync, UpdateBucketSearchAsync, RemoveBucketFromSearchAsync; GetPlaybackAsync, GetActivityForBucketAsync, Source on assets and results. The browser controller plays files of your own buckets and libraries.
  • Angular: results carry source; the preview plays a library video at the moment; [startSeconds] and [startKey] on <plycdn-player>.
  • React: results carry source; librarySource(result); startSeconds and startKey on <PlycdnPlayer>.
  • Dashboard: the switch in a bucket's and a library's Settings, each file's search state, "A plycdn bucket" on Storage connectors, the Library's source column and filter, and search results that play beside their transcript.

45Fixed in Content Search 1.15.1 and Storage & video 1.7.1

No breaking change and no code to edit: nothing you call changes. Install the new package versions together, as before.

  • Live players switch to new qualities smoothly. When a stream's other qualities arrive about a minute after it goes on air, the live player (@plycdn/angular, @plycdn/react, and the player in the dashboard) now moves over to them without a break. Before, it could stop on a black picture until the page was reloaded.
  • Restreaming always carries the sound. When you send a live stream on to other platforms, its sound now always reaches the destination. Before, some platforms received the picture with a silent, empty sound track and showed nothing. This is a fix in the service: it needs no package change, and it applies to every stream you restream.
  • Content Search 1.15.1 has no changes; it is released in step with the service.

46New in Content Search 1.15.0 and Storage & video 1.7.0

No breaking change and no code to edit: nothing you call today changes. Install the new package versions together, as before.

Narrow a search, and search selected documents. A search can now be made within earlier ones - up to five steps, each narrowing to its 2,000 most relevant documents - and over a selection of up to 1,000 documents. Pages continue with a cursor (it takes precedence over offset), and a selection can be saved as a collection in one call (Narrow a search and search selected documents).

  • Service: within, a selection of up to 1,000 assetIds and cursor on /search and /answer, which answer scope and nextCursor; GET /activity?order=filename finds documents by name with a cursor, as a reader with asReader and principals; POST /collections saves a selection. New codes: too_many_steps, selection_too_large, cursor_expired and cursor_mismatch (API reference). A search with steps or a selection is one search.
  • .NET: ContentSearchQuery.Within and .Cursor, ContentSearchResults.Scope and .NextCursor, FindDocumentsAsync and CreateCollectionFromDocumentsAsync. The browser controller's activity route now lists as the signed-in reader: a page that lists files through it shows only the files that person may see, as search always has.
  • Angular: <plycdn-scoped-search> and its parts (breadcrumbs, search within, the scope toggle, the selection pane with the document finder), PlycdnSearchSession, and [virtualize] and [selectable] on <plycdn-content-results>.
  • React: <PlycdnScopedSearch> and its parts, usePlycdnSearchSession, and findDocuments and createCollection on the client.

Ask your library. An assistant for every page of your app: people ask in their own words, follow up, and read the answer as it is written, each claim with a source that opens the exact moment or page. It answers from what the person may see, and from their selection when they search only that (Add the assistant to your app).

  • Service: /answer takes conversation (up to six earlier turns, never stored) and answers question; with Accept: text/event-stream it streams delta, citations and done events (API reference).
  • .NET: AnswerAsync(org, query, conversation) and StreamAnswerAsync; the controller relays streamed answers, and an answer's citations now point at your delivery host like search results do.
  • Angular and React: <plycdn-assistant> / <PlycdnAssistant>, with PlycdnAssistantSession / usePlycdnAssistant for your own interface; streamAnswer on the client.

Large libraries are faster. Searching, narrowing and listing stay quick as a library grows to many thousands of files, and one organization's large upload no longer delays another's.

Live streaming.

  • On air in seconds. A stream now goes live within seconds at the quality your encoder sends; its other qualities are added about a minute later and the players take them up by themselves. Until then a stream shows qualitiesComing: true and the players say More qualities coming (Live streaming).
  • Clearer reasons when going live fails. An encoder that connects but sends no picture within 15 seconds is ended with the reason input_no_video, which says what to fix (add a video source, set a keyframe every 1 to 2 seconds), listed with the other live reasons in the video guide.
  • No empty recordings. A stream that never went live leaves no video behind; only broadcasts that reached viewers become recordings.
  • Watch-page logo size. The brand's logoSize (small, medium or large; about 32, 48 or 72 pixels high) sets how large the logo is on watch pages; the dashboard has a picker with a live preview, and the watch components follow it (Brand your watch page).
  • Watch domains, step by step. Adding a hostname now shows exactly the records it needs for its type: a subdomain takes two CNAME records, a bare domain is refused with advice to use a subdomain, and Check now looks again straight away (Use your own domain for watch pages).
  • Share links for live streams, shipped in the previous release, are unchanged.

47Fixed in Content Search 1.14.1

No breaking change and no code to edit: nothing you call changes. Install the three Content Search packages together; Storage & video stays at 1.6.0.

  • Long PDFs and documents with many matches. They open and scroll smoothly, and use far less memory: pages are drawn a couple at a time, nearest the reader first, and pages far from the one being read are released and drawn again when you come back to them.

48New in Content Search 1.14.0 and Storage & video 1.6.0

No breaking change and no code to edit: nothing you call today changes.

Share a live stream. A share link is an address anyone can open to watch a live input, with no account and nothing for your backend to sign. It can have a password and an expiry, can show the recording when the stream ends, and can be turned off or deleted at any time. Any number of people can watch through one link, and an input can have up to 50. Viewers see Starting soon until the stream begins, then the stream, then This stream has ended and, if you want it shown, the recording (Share a live stream).

  • Service: POST and GET /live-inputs/{id}/share-links and PATCH and DELETE /live-inputs/{id}/share-links/{linkId} (admin key), and two public routes for viewers, GET /api/watch/{code} and POST /api/watch/{code}/session, which need no key and answer to any website. New codes: share_link_not_found, share_link_limit_reached, watch_link_not_found, watch_not_started (retryable), watch_ended, watch_password_required and watch_password_attempts (retryable) (API reference).
  • .NET (Plycdn.AspNetCore.Video 1.6.0): CreateShareLinkAsync, ListShareLinksAsync, UpdateShareLinkAsync and DeleteShareLinkAsync, with ShareLink, CreateShareLinkRequest and UpdateShareLinkRequest (its ClearLabel, ClearPassword and ClearExpiresAt remove a value), and the new codes in PlycdnLiveErrors.
  • Angular and React (1.6.0): the watch component, <plycdn-watch> and <PlycdnWatch>, for your own page: give it a link's address or code and it asks for the password, waits for the stream, plays it, renews its links and plays the recording afterwards. Its sentences and colours are one options object (Angular, React).
  • Dashboard: create, change, turn off and delete a live input's share links.

Brand your watch pages, and open them on your own domain. Set a logo, colours, a theme, one of three layouts (standard, side_panel, cinema), a title and description, your own sentences, a button to your site and up to three footer links once for the organization (Brand your watch page); a share link can change the title, description, sentences and layout for itself. Add a hostname of yours (up to five) and share links open on it over HTTPS, with a CNAME and a TXT record at your DNS host (Use your own domain for watch pages).

  • Service: GET and PUT /watch-brand, PUT and DELETE /watch-brand/logo, GET and POST /watch-domains and GET and DELETE /watch-domains/{id} (admin key); share links take and answer title, description, messages and layout, and answer urls; the watch info carries the resolved brand. New codes: logo_invalid, logo_too_large, watch_domain_limit_reached, watch_domain_not_found, domain_exists and hostname_not_allowed (API reference).
  • .NET (Plycdn.AspNetCore.Video 1.6.0): GetWatchBrandAsync, UpdateWatchBrandAsync (with Clear… flags), UploadWatchBrandLogoAsync, DeleteWatchBrandLogoAsync, ListWatchDomainsAsync, AddWatchDomainAsync, GetWatchDomainAsync and RemoveWatchDomainAsync; Title, Description, Messages and Layout on the share-link requests and Urls on ShareLink; the new codes in PlycdnLiveErrors.
  • Angular and React (1.6.0): <plycdn-watch> and <PlycdnWatch> draw the link's brand in the same three layouts; what you set in options still wins over the brand.
  • Dashboard: the Watch page area sets the brand with a live preview and manages your watch domains.

Also in these releases.

  • Live playback. Low-latency live playback keeps working when the player asks for the newest part of a stream: the players (@plycdn/angular/video, @plycdn/react/video, and the playback in @plycdn/content-search and @plycdn/content-search-react) now keep the link's own settings on every request they make.
  • New guide: your first live stream, step by step. The video guide walks from creating a live input to using the replay, in order, with a pointer to the full description of each step (Video).
  • New error code live_realtime_unavailable (503, retryable) is documented: a real-time link could not be made just now; ask again in a few seconds, or ask for mode: "broadcast". It is listed with the other live errors in the video guide.

49Added in Content Search 1.13.0 and Storage & video 1.5.0

No breaking change: nothing you call today changes, and you can move to these versions without editing code.

Live streaming. Broadcast from OBS or any encoder (RTMPS, RTMP or SRT) or straight from a web page, and viewers watch a few seconds behind (broadcast mode, for audiences of any size) or in under a second (real-time mode, for interactive audiences, with broadcast taking over automatically beyond a safety cap). Each stream is recorded as an ordinary video in the library, at the same address, with no second encode. A live input is the standing place to broadcast to; a browser goes live with a short-lived publish token that your server decides to hand out, so a stream key never reaches a page. Live streaming is included in every plan that includes video and is counted by the minute of streaming, by quality, and for real-time viewing by viewer-minute (Plans and prices).

  • Service: live inputs (POST /libraries/{libraryId}/live-inputs and the routes under /live-inputs/{id}), stream keys, publish tokens, streams (/live-inputs/{id}/streams, /streams/{id}, POST /live-inputs/{id}/end), POST /sign with a liveInputId and a mode, a new video status of live, four webhook events (live.connected, live.started, live.disconnected, live.ended), five usage metrics (live_source_minutes, live_sd_minutes, live_hd_minutes, live_fhd_minutes, live_realtime_viewer_minutes) and the live plan feature. New codes: live_input_not_found, live_not_started (retryable), live_input_disabled, live_input_busy, live_cap_reached, live_quota_exceeded, publish_token_invalid, stream_key_in_use, feature_not_available and feature_not_in_plan. Safety caps (25 simultaneous streams, 24 hours, 1080p, 12,000 kbps, 10,000 real-time viewers a stream, 1,000 inputs) can be raised on request. Creating a live input takes an Idempotency-Key like every other create, and GET /regions says which regions offer live streaming (live).
  • .NET (Plycdn.AspNetCore.Video 1.5.0): CreateLiveInputAsync, ListLiveInputsAsync, GetLiveInputAsync, UpdateLiveInputAsync, DeleteLiveInputAsync, RotateStreamKeyAsync, RevealStreamKeyAsync, CreatePublishTokenAsync, DisableLiveInputAsync, EnableLiveInputAsync, EndStreamAsync, ListStreamsAsync, GetStreamAsync and PlycdnVideoSigner.SignLiveAsync; the controller routes live-inputs/{id} and live-inputs/{id}/publish-token, decided by your new IPlycdnLiveAuthorizer (which refuses everyone until you register one); the live.* constants on PlycdnWebhookVerifier and PlycdnWebhookEvent.AsLive() (with each event's fields typed); StorageRegion.Live.
  • Angular and React (1.5.0): the broadcaster (<plycdn-live-broadcaster>, <LiveBroadcaster>) and the live player (<plycdn-live-player>, <LivePlayer>), with PlycdnBroadcastService and PlycdnLiveService (Angular) and client.publishToken, client.signLive and client.getLiveInput (React). Both show nothing while live streaming is not switched on for your service.
  • Dashboard: Video streaming, then a library's Live tab: live inputs, their settings, the stream key (shown again only on request, and recorded in the audit log), a test broadcast from the browser, the stream running now with its health, and past streams with their recordings.
  • Guides: Live streaming in the Video guide, Live inputs and streams in the API reference, and the live parts of the Integration guide and of the three platform guides.
  • More for live streaming (in the same release): backup ingest for a second encoder and redundant processing (backupIngest, redundantTranscoding; the live.failover webhook), a rewind window of up to two hours (dvrWindowSeconds, off by default), searchable recordings (indexRecording), per-stream viewer and health figures (GET /streams/{id}/viewers, /health), live captions (captions, GET /streams/{id}/captions, the live_caption_minutes usage), and, available on request, restreaming to other platforms (destinations and the live.destination_failed webhook), ad breaks and ad insertion (POST /live-inputs/{id}/breaks, ads, the live.ad_break webhook) and content protection (protection; playback links carry protection with each key system's addresses). Playback links also carry captions, dvrWindowSeconds and realtime (kind whep or rtc). New codes: live_setting_not_in_plan, live_setting_conflict (ads cannot be combined with protection or with a rewind window), protection_not_configured, destination_unreachable, destination_limit_reached, destination_url_invalid and destination_not_found; new safety caps live_dvr_seconds and live_destinations. The .NET package gains GetStreamCaptionsAsync, GetStreamViewersAsync, GetStreamHealthAsync, CreateDestinationAsync, ListDestinationsAsync, UpdateDestinationAsync, DeleteDestinationAsync, TestDestinationAsync and InsertAdBreakAsync, and PlycdnVideoSigner.SignVideoFromPlycdnAsync (always signed by plycdn: use it for the recording of a protected stream, since a link signed locally carries no protection); the Angular and React packages gain getStreamCaptions, and their players use captions, the rewind window, rtc real-time links, ad breaks ((adEvent), onAdEvent, adQuartiles) and protected streams. Creating a destination takes an Idempotency-Key like every other create (CreateDestinationAsync sends one, or yours). The video player does not play protected video yet: see content protection for a protected recording.

50Added in Content Search 1.12.0 and Storage & video 1.4.0

No breaking change: nothing you call today changes, and you can move to these versions without editing code.

plycdn links. New private buckets and video libraries normally sign links with a key plycdn holds for the bucket (scheme pl1, and pl1d for a whole video). A plycdn link keeps working whatever changes on our side, so a link you have already handed out does not break because of how we deliver. Buckets made before keep their links (ps2, ps3); we can switch a bucket to plycdn links for you, and its earlier links keep working for a week after. A bucket that allows credentialed cross-origin requests keeps ps2: a plycdn link is answered with a redirect, which a browser will not carry credentials across.

  • Service: the signing profile's scheme is pl1 or ps2, and directoryScheme is pl1d or ps3. On a pl1 bucket a key rotation takes effect within about a minute (wait a minute before handing out links signed with the new key), and turning cors.allowCredentials on is refused with 422 validation_error.
  • .NET (Plycdn.AspNetCore.Storage and .Video 1.4.0): PlycdnLinkSigner and PlycdnVideoSigner sign pl1 and pl1d in your process. Earlier versions ask plycdn to sign them and keep working.
  • Angular and React (1.4.0): no change for you; links still come from your backend.
  • Dashboard: a bucket's Signing tab shows which kind of link it signs, and the rotate confirmation says how soon old links stop.
  • Guides: Signing links yourself now gives the exact bytes to sign, with worked examples (including a key with non-ASCII characters), for signing without our packages.

51Added in Content Search 1.11.0 and Storage & video 1.3.0

No breaking change: nothing you call today changes, and you can move to these versions without editing code.

Delivery analytics: see what is behind your traffic. Switch it on for a bucket or video library and plycdn counts requests, bytes, cache results and status classes by the hour, and by up to three breakdowns you define: a part of the path, a query parameter, a regular expression, or a tag your server signs into a link (PlycdnSignOptions.Tags, private buckets). Read it over time or as a monthly breakdown (JSON, or CSV for a large month), by value, bucket or organization (one row per sub-organization for an account that has them), and try a rule on recent traffic before saving it. Each breakdown keeps a ceiling of distinct values per day that you can raise or return to your plan's. It is on every plan and adds two statement lines, Requests analysed and Tracked values, at the prices on your price list.

  • Service: GET /analytics/delivery, GET /analytics/delivery/breakdown, POST /buckets/{bucketId}/analytics/preview, the analytics section of buckets and libraries, and tags on POST /sign. A refused tag is a 422 validation_error whose detail says why; until link tags are switched on for your account, a request with tags is refused with "Link tags are not available yet". New codes: too_many_rows, too_many_organizations, export_busy and analytics_unavailable.
  • .NET (Plycdn.AspNetCore.Storage and .Video 1.3.0): PlycdnSignOptions.Tags, Bucket.Analytics, BucketPatch.Analytics, GetDeliveryAnalyticsSeriesAsync, GetDeliveryBreakdownAsync, GetDeliveryBreakdownCsvAsync, PreviewAnalyticsRuleAsync, and for browser-requested links PlycdnOptions.TagsFor and AcceptBrowserTags (off by default: tags a browser sends are ignored, so visitors cannot label their own links).
  • Angular and React (@plycdn/angular, @plycdn/react 1.3.0): an optional tags on a sign request; your backend decides whether to use it.
  • Dashboard: the Analytics tab of every bucket and library (charts over time, rules with a preview, and the ceiling), and Delivery → Delivery breakdown (by value, bucket or organization, with CSV).

Read Delivery analytics and the API reference.

52Content Search 1.10.0

For a team moving from 1.6.0, the last version you received. The packages are @plycdn/content-search (Angular) 1.10.0, @plycdn/content-search-react 1.10.0 and Plycdn.ContentSearch.AspNetCore 1.10.0. Install all three together: they move in step with the service.

The first section says what to do and the second what is new. The Content Search bundle's RELEASE-NOTES.md additionally lists every public difference between the 1.6.0 and 1.10.0 packages, read from the packages themselves.

What to do when upgrading from 1.6.0

Moving to plycdn.com (all packages)

Your server now talks to the hosted API at https://api.plycdn.com/. That is the default of the 1.10.0 C# package, so the server-side change is to remove any Plycdn:ApiBaseUrl setting (environment variable Plycdn__ApiBaseUrl) that names the old host; set it only if your plycdn contact gives you another HTTPS address. The browser packages never call the API directly, so they need no address. The full guides are at https://plycdn.com/docs, the one place they are kept up to date.

Angular

  1. Install the new package (1.18.0): npm install ./plycdn-content-search-1.48.0.tgz. Video and audio playback is built in: installing the package brings in everything it needs automatically (if you use a private npm registry or mirror, make sure it can serve the package's own dependency). The preview plays your file as it is.
  2. ContentClient.refreshSource and ContentUploadComponent.renewSource are gone (1.7.0). Source access is renewed by your server through the source-access callback; the upload row's Check again asks again. Remove any call to either.
  3. The preview plays recordings with its own controls (1.10.0). If you styled or relied on the browser's controls inside plycdn-content-playback, set [nativeControls]="true" on plycdn-content-playback, plycdn-content-results or plycdn-content-search. While focus is inside the preview, [ / ] and Alt+← / Alt+→ now move between matches (Alt+arrows are normally the browser's Back and Forward); [matchNavigation]="false" gives them back.
  4. ContentPlaybackComponent.refresh and ContentPlaybackComponent.seek are gone (1.10.0). The player renews an expired source link itself and resumes at the same second. Remove any call to either; to open another place, set result or startSeconds as before.
  5. ContentPlaybackComponent.constructor also takes an NgZone (1.10.0). Angular supplies it; only a unit test that constructs the component itself needs to pass one.
  6. ContentLabels has 26 new required keys (1.10.0), so code that types a complete translation as ContentLabels stops compiling (TS2739, "missing the following properties"). Spread the defaults under your own strings, const labels: ContentLabels = { ...DEFAULT_LABELS, ...mine }, or add the keys: back10, captions, captionsOff, captionsShort, exitFullscreen, forward10, fullscreen, matchAnnounce, matchNext, matchPosition, matchPrevious, mute, pause, play, pointerLine, pointerPage, pointerRow, pointerSheetRow, pointerSlide, pointerTime, quality, qualityAuto, seek, speed, unmute, volume. A Partial<ContentLabels> (what the components' labels input takes) is unaffected.

C#

  1. ContentController.RefreshSource(id,ct) and ContentController.Settings(ct) are gone (1.7.0), and with them the routes assets/{id}/refresh-source and settings. Renew source access through the source-access callback, and read settings on your server with GetSettingsAsync.
  2. API keys travel in the header only. A key in the query string was deprecated in 1.7.0 and, since 1.9.0, the service refuses it, in the query string or in a request body, with 401 api_key_required. The package has always sent the Authorization: ApiKey header; this matters only to code that calls the API directly.

React

  1. ContentLabels has 26 new required keys (1.10.0), so code that types a complete translation as ContentLabels stops compiling (TS2739). Spread the defaults under your own strings, const labels: ContentLabels = { ...DEFAULT_LABELS, ...mine }, or add the keys: back10, captions, captionsOff, captionsShort, exitFullscreen, forward10, fullscreen, matchAnnounce, matchNext, matchPosition, matchPrevious, mute, pause, play, pointerLine, pointerPage, pointerRow, pointerSheetRow, pointerSlide, pointerTime, quality, qualityAuto, seek, speed, unmute, volume. A Partial<ContentLabels> is unaffected.

Nothing else to change: the rest of the package has only gained.

What is new since 1.6.0

Angular

  • Moving between a file's matches in the preview (1.10.0). Previous and Next under the video, audio or document, with "Match 2 of 9" between them; the keys ] / Alt+→ and [ / Alt+←; a polite announcement of each move; (matchChanged) with the result as at the new match, which the results list follows. [matchNavigation]="false" turns the row and the keys off.
  • The preview's own player (1.10.0): labelled controls, the plycdn player's keyboard map, and a tick on the seek bar for every match (ticks closer than 6 pixels merge and cycle). The same player engine as the plycdn video player, so expired links are renewed without interrupting the reader. [nativeControls]="true" keeps the browser's controls, with the ticks in a strip.
  • Match helpers (1.10.0): matchesInOrder, matchIndex, adjacentMatch, markerPositions, matchPointer and resultAtMoment, for an interface of your own.
  • New labels (1.10.0): matchPrevious, matchNext, matchPosition, matchAnnounce, the pointer* templates and the player controls' names. All have English defaults.
  • Theming (1.10.0): --pcs-marker, --pcs-marker-active, --pcs-player-bar, --pcs-player-text.
  • Two more label keys (1.10.0): qualityAuto ("Auto") and captionsShort ("CC"), the short names the player's quality and captions buttons show. Both packages have them, with English defaults.
  • PDF viewer (1.9.1): the PDF viewer inside the package now includes a security update. Nothing to change.

React

  • Match helpers and labels (1.10.0): matchesInOrder, matchIndex, adjacentMatch, markerPositions, matchPointer, and the same new labels in DEFAULT_LABELS. The API reference ("Match navigation in your own interface") shows Previous / Next and seek-bar markers in about twenty lines.

C#

  • Sub-organizations and the account audit from code (1.7.0): PlycdnAccountApi, with an admin key.
  • Settings in part (1.7.0): UpdateSettingsAsync with a ContentSettingsChange changes only what you name; ClearSettingsAsync returns an organization to its parent's settings.
  • Transcripts from code (1.7.0): GetTranscriptAsync and CorrectTranscriptAsync.
  • Usage, quotas and statements: GetUsageAsync by month, GetUsageDailyAsync, GetUsageQuotasAsync, ListStatementsAsync, GetStatementAsync and ExportStatementCsvAsync. First and last months are prorated when you go live mid-month.
  • Errors you can act on: PlycdnApiException carries Code and RetryAfter.
  • Everything the API offers, from C# (since 1.6.0): AnswerAsync, SuggestAsync, GetRelatedAsync, GetLibraryHealthAsync, GetSearchAnalyticsAsync, GetMetadataAsync, GetPermissionsAsync, collections (ListCollectionsAsync, UpsertCollectionAsync, DeleteCollectionAsync, SetCollectionMembersAsync), GetSynonymsAsync / SetSynonymsAsync, boosts and pins. ContentSearchQuery takes a Filter (ContentSearchFilter), Principals and Group.
  • Duplicate-charge waivers and usage detail: SourceReference.ContentSha256 declares the checksum of a large recording we read in ranges, so a copy of it is not charged twice (see "How usage is counted" in the Integration guide); ContentUsage also carries Proration, PayAsYouGo and an Organizations breakdown.
  • ContentSearchOptions.MaxUploadBytes (default 20 GB) caps what the upload route accepts.

Fixed

  • On-screen text setting for connectors, from C# (1.10.0). ContentConnectorRequest.ExtractContentFromVideoFrames is now sent when you create a connector and when you update one, and an update that leaves it unset keeps the value already stored (including one set in the dashboard). If you set it from C# earlier and saw no effect, set it again.
  • C#: registering a file and importing no longer fail when you leave the language unset (1.10.0). RegisterAssetAsync and ImportAsync now leave language and audioTrack out of the request unless you give a value; before, an unset value was sent as null and refused with 422.

Documented error codes

The API reference's error table documents more codes than it did in 1.6.0 (including the storage, video and webhook codes). None was removed, so no code you match on today has changed meaning. The bundle's RELEASE-NOTES.md lists every one added.

53Content Search 1.5.0 to 1.9.0

What each of these releases added, newest first. If you are moving from 1.6.0 or earlier, the 1.10.0 section above says what to change; this part says what you gained on the way. Everything here is additive unless it says Changed.

Version 1.7

  • Two keys at once, and admin keys. Up to two API keys can be active, so a key is replaced without downtime: rotate, deploy, then deactivate the old one. Keys are managed in the dashboard only - they are credentials, managed by people - and the key routes refuse every API key with dashboard_session_required. An admin key, created only in the dashboard, manages sub-organizations and reads the account audit from code (PlycdnAccountApi).
  • Sub-organizations from code, with your own Metadata on each.
  • Settings in part: UpdateSettingsAsync changes only what you name; ClearSettingsAsync returns an organization to its parent's settings.
  • Transcripts from code: GetTranscriptAsync and CorrectTranscriptAsync.
  • Changed: renewing source access and reading settings are server actions only. Angular's refreshSource() and the controller routes assets/{id}/refresh-source and settings are gone. The upload component's waiting row offers Check again, and access is renewed by your server through the source-access callback. Read settings on your server with GetSettingsAsync.
  • Changed: sub-organization and account-audit routes refuse a standard key (admin_key_required): use an admin key, or the dashboard. People, invitations and the organization profile routes (/api/users, /api/organizations) refuse every API key (dashboard_session_required): manage them in the dashboard.
  • Recorded: the account audit now also lists every dashboard "search as" (dashboard.search_as): who, which organization and how many principals, never their names.

Version 1.6

  • One result per file. A file that answers a query in several places is one result, listing every place it matched as a jump-to link - 02:14 · 07:51 · 12:03 on a recording, Page 4 · Page 11 in a document, Slide 3 in a course. The Angular <plycdn-content-search> does this by default ([groupByFile]="false" restores one result per passage). React: search(q, { group: 'file' }); .NET: new ContentSearchQuery { Group = "file" }. Paging then counts files. See "One result per file" in the API reference.
  • The words that matched are marked - in each result's snippet and, when a document opens in place, in the matched line, paragraph or row. PDFs now render in the component at the matched page with the passage highlighted and the words marked, with zoom and page controls; this needs the same storage CORS rule as Word and Excel, and falls back to the browser's viewer without it. A result found by meaning alone, sharing no words with the question, says so ("Matched by meaning"). Results carry highlights, pages carry terms; React exports markSpans and markTerms for your own interface.
  • SCORM packages of every common shape are read: launch pages that navigate between lesson pages, slide players whose course is in a data file, Articulate Rise exports, PDF and Word files inside a package, and packages zipped inside a folder. See "What can be indexed".
  • The result button says where it goes: "Play from 2:14" on a recording, "Go to page 4" in a PDF, "Go to slide 3" in a course, "Go to row 12" in a spreadsheet, "Go to line 40" in text, "Go to the passage" otherwise - instead of a generic "Preview video". Translate it with the playFrom, goToPage, goToSlide, goToRow, goToSheetRow, goToLine and goToPassage labels; {time}, {page}, {slide}, {row}, {sheet} and {line} are filled in wherever your translation puts them (goToPage: 'पृष्ठ {page} पर जाएँ').
  • Result locations gain slide, title, resource, element and source in every package's types, so a course result can read "Slide 4 · Definition and Scope".

Everything is additive: an integration built on 1.5 keeps working unchanged.

Version 1.5

  • New package names. The packages are now Plycdn.ContentSearch.AspNetCore, @plycdn/content-search and @plycdn/content-search-react. Only names changed: the registration call is AddPlycdnContentSearch, the Angular module PlycdnContentSearchModule, components <plycdn-content-upload>, <plycdn-content-search>, <plycdn-content-results> and <plycdn-content-playback>, the configuration section Plycdn, the default customer route /api/plycdn-content/v1, the organization claim plycdn_organization_id, CSS variables --plycdn-*, and the default service address https://api.plycdn.com/. If you are upgrading, rename these and update the callback URL registered in the dashboard to the new route.
  • A React package for search, answers, suggestions, related files, the library and progress. See "Using React".
  • Clearer validation errors. A validation_error now says which field was wrong in detail, for example that a storage host was given with https:// or a path.
  • Spoken language is honoured. The language you register a recording with (for example hi or ta) is passed to transcription, which improves accuracy for Indian languages.

54Storage & video 1.2.0

The storage and video packages are split by purpose. Before 1.2.0, Plycdn.AspNetCore held storage and video together, and @plycdn/angular and @plycdn/react each had one entry point for everything. From 1.2.0 you install and import only what you use: storage, video, or both. The names changed; what the packages do, the routes your frontend calls and everything plycdn answers did not. Nothing about the service, your data, your API key or your plan changes.

Most teams simply install the 1.2.0 names. If you installed a 1.1.x package, the table and the steps below take you across.

New names

Before (1.1.x) From 1.2.0
NuGet Plycdn.AspNetCore (storage and video) Plycdn.AspNetCore.Storage (buckets, uploads, delivery, domains, cache, signing, the browser controller) and Plycdn.AspNetCore.Video (libraries, playback, webhooks, the video route). Video depends on Storage
NuGet Plycdn.AspNetCore.S3 Plycdn.AspNetCore.Storage.S3
NuGet Plycdn.AspNetCore.AzureBlob Plycdn.AspNetCore.Storage.AzureBlob
Namespace Plycdn Plycdn.AspNetCore.Storage, Plycdn.AspNetCore.Video, Plycdn.AspNetCore.Storage.S3, Plycdn.AspNetCore.Storage.AzureBlob (the namespace matches the package)
services.AddPlycdn(...) services.AddPlycdnStorage(...), or services.AddPlycdnVideo(...), which registers storage as well
Library, video and webhook calls on PlycdnStorageApi (CreateLibraryAsync, CreateVideoAsync, CreateWebhookAsync and the rest) The same methods on PlycdnVideoApi, which AddPlycdnVideo registers and which is constructed from a PlycdnStorageApi
PlycdnLinkSigner.SignVideoAsync PlycdnVideoSigner.SignVideoAsync, with the same arguments
PlycdnWebhookVerifier, PlycdnWebhookEvent, PlycdnWebhookSignatureException, PlaybackLink, Library, Video The same names, in Plycdn.AspNetCore.Video
GET api/plycdn/v1/videos/{id} served by PlycdnStorageController The same route, now served by PlycdnVideoController (the sign route stays on PlycdnStorageController)
npm @plycdn/angular: everything from one entry point @plycdn/angular/storage (PlycdnStorageModule, PlycdnStorageClient, PlycdnUploadComponent, PlycdnFileBrowserComponent, PlycdnUploader) and @plycdn/angular/video (PlycdnVideoModule, PlycdnPlayerComponent, PlycdnPlaybackService, PlycdnPlayerCore). The root @plycdn/angular holds the shared types, failures, messages and theme, and the framework-free upload and player engines that the two entry points re-export
PlycdnModule.forRoot(...) PlycdnStorageModule.forRoot(...) for uploads and the file browser, PlycdnVideoModule.forRoot(...) for the player. PlycdnModule is removed
npm @plycdn/react: everything from the root @plycdn/react/storage (provider, upload, file browser, hooks, uploader, client) and @plycdn/react/video (provider, player, usePlayback, useVideo, PlycdnPlayerCore). The root holds types only

The packages are one version line again: 1.2.0 for all of them.

Regional hostnames use a short label

Storage hostnames carry the region's short host label, not its region code: s3.in-mum.plycdn.net, <bucket>.s3.in-mum.plycdn.net, blob.in-mum.plycdn.net, <account>.blob.in-mum.plycdn.net and, for upload parts, in-mum.upload.plycdn.com (Mumbai; later regions follow the same pattern, for example in-hyd and sg-sin). The region code does not change. mumbai is still what you pass to the API and the SDKs, what you set as the region in S3 tools, and what requests are signed with. The packages keep no list of regions or labels: region codes are data from GET /regions. Create clients with await PlycdnS3.CreateClientAsync(api, organizationId, "mumbai", ...) or await PlycdnBlob.CreateServiceClientAsync(api, organizationId, ...), which read the label from the region's endpoints. To skip the lookup, set HostLabel (PlycdnCompatOptions.HostLabel, or the hostLabel argument) from that same listing. The region code is never used as a host label: a client that cannot find the label fails with a message that says so, instead of connecting to a host that does not exist. If you register a client from configuration (AddPlycdnS3, AddPlycdnBlob) or call CreateClient / CreateServiceClient without the async lookup, set Plycdn:S3:HostLabel or Plycdn:Blob:HostLabel (the label in the region's S3 endpoint from GET /regions).

What to change

ASP.NET Core.

  1. Remove Plycdn.AspNetCore (and .S3 or .AzureBlob if you use them) and install the new packages: Plycdn.AspNetCore.Storage, and Plycdn.AspNetCore.Video if you use video.
  2. Replace using Plycdn; with using Plycdn.AspNetCore.Storage; (and using Plycdn.AspNetCore.Video;, using Plycdn.AspNetCore.Storage.S3; or using Plycdn.AspNetCore.Storage.AzureBlob; where you use them).
  3. Replace AddPlycdn(...) with AddPlycdnStorage(...), or with AddPlycdnVideo(...) if you use video. The Plycdn configuration section and its settings are unchanged, and so is your IPlycdnStorageAuthorizer.
  4. Where you call library, video or webhook methods, inject PlycdnVideoApi instead of PlycdnStorageApi; where you sign video links, inject PlycdnVideoSigner instead of PlycdnLinkSigner.
// Before (1.1.x)
builder.Services.AddPlycdn(builder.Configuration);
var link = await signer.SignVideoAsync(orgId, libraryId, videoId);        // PlycdnLinkSigner

// From 1.2.0
builder.Services.AddPlycdnVideo(builder.Configuration);
var link = await videoSigner.SignVideoAsync(orgId, libraryId, videoId);   // PlycdnVideoSigner

Angular.

  1. Install plycdn-angular-1.39.0.tgz (the one tarball carries both entry points).
  2. Replace PlycdnModule.forRoot(...) with PlycdnStorageModule.forRoot(...) imported from @plycdn/angular/storage and, if you play video, PlycdnVideoModule.forRoot(...) from @plycdn/angular/video.
  3. Change imports of services, components and the uploader to the entry point that now holds them (table above). Templates (<plycdn-upload>, <plycdn-file-browser>, <plycdn-player>) are unchanged.

React.

  1. Install plycdn-react-1.39.0.tgz.
  2. Change imports from @plycdn/react to @plycdn/react/storage or @plycdn/react/video. Both export the same PlycdnProvider, so one provider serves a page that uses both. Props and hooks are unchanged.

Player type names

The exported player types and values no longer name the playback technology. If your code imports the player's engine type or its loader types from @plycdn/angular/video or @plycdn/react/video, use the new names. Behaviour is unchanged.

Before Now
HlsLike AdaptiveStreamLike
HlsLevel AdaptiveStreamLevel
HlsConstructor AdaptiveStreamConstructor

The loader is now loadAdaptiveStream, and the player reports its engine as 'adaptive'.

What did not change

  • The wire contract: every route, field and error code, and what your frontend sends to your backend.
  • Behaviour of uploads, signed links, playback, webhooks and the S3 and Azure clients.
  • Your API key, buckets, libraries, domains, webhooks and plan.

The full guides for the new packages are Storage & video with ASP.NET Core, with Angular and with React.