QVCCS innovation · Audit and governance
What is really in your knowledge bases? A point-in-time Genesys Cloud knowledge inventory
Knowledge now feeds agents, bots and customers alike, so the question ‘what exactly is in our knowledge bases, and is it any good?’ matters more than ever. Knowledge Manager answers it with one click: a complete, read-only, point-in-time snapshot of the Knowledge workbench – every base, category, label and article, with full bodies in both draft and published form.
Did you know that a Genesys Cloud knowledge article can carry touchpoint variations, so the same answer is tailored for Agent Assist, bot flows, the knowledge portal and the Knowledge app for Messenger? One article can therefore hold several versions of its content.
Did you know that when you import or export knowledge articles, the externalID attributes and the category and label structure are only available in JSON format? CSV and XLSX suit basic FAQs and spreadsheets.
Did you know that every knowledge article has a draft side and a published side, each with its own set of content variations? Reviewing only what is live misses what is about to go live.
01
The knowledge review nobody has time to do
A new bot is launching on Monday and it will answer from your knowledge bases. The contact centre manager wants to know which articles are still in draft, the compliance officer wants every article that mentions refunds, and the knowledge author wants to see which categories have grown unwieldy. All of that content lives in Genesys Cloud CX. Reviewing it article by article, knowledge base by knowledge base, is the kind of job that gets postponed until something goes wrong in front of a customer.
Knowledge Manager makes the review a one-click job. Press Harvest and the app walks the Knowledge APIs and captures everything: every knowledge base, the full category tree, all labels, and every article with its metadata – state, visibility, authorship, dates, published version and alternative phrases – plus its complete body content. The result is saved as a point-in-time snapshot, and the review portal then works entirely from it. Browsing never re-queries Genesys, so review sessions are instant and put no load on the platform.
02
The knowledge model, endpoint by endpoint
The app mirrors the Genesys knowledge model faithfully. A knowledge base is a container of articles with one core language, and an org may have several – per brand, department or bot. Categories are hierarchical folders inside one knowledge base, and an article references at most one; articles without one appear under Uncategorised. Labels are flat, colour-coded tags that cut across categories. Each article has a state of Published, Draft or Archived, a visible flag that controls whether knowledge search surfaces offer it, and alternative phrases that should match it.
The harvest follows that structure through the Genesys Public API. GET /api/v2/knowledge/knowledgebases enumerates every knowledge base; for each one, the categories, labels and documents endpoints return the category list, which the app rebuilds into a hierarchy from parent references, the labels with their colours, and every article's metadata. Finally, the variations endpoint is called separately with documentState=Draft and documentState=Published, so both sides of every article are captured. A missing side is normal and skipped silently.
Content itself is a tree of blocks: paragraphs, ordered and unordered lists, images, videos and tables, with bold, italic, underline and strikethrough marks and hyperlinks. The snapshot keeps that tree verbatim, so nothing is flattened or lost before review. To keep the portal fast even for large orgs, it boots from a light form of the snapshot – every document reduced to has-draft and has-published flags and variation counts – and streams an article's full bodies only when the reader opens it.
Read this diagram as text
Three columns show the Genesys Knowledge API hierarchy on the left converging on a point-in-time snapshot in the middle, which feeds the QVCCS Knowledge Manager review portal on the right.
- Left panel, "Knowledge API · GET only: walk the whole workbench", starts with Knowledge bases (/api/v2/knowledge/knowledgebases).
- Under each knowledge base, three calls fetch categories (hierarchy), labels (name, colour) and documents (metadata).
- Under documents, the document variations call is made with documentState Draft or Published, giving a Published side and a Draft side.
- Knowledge bases, the per-base calls and the variations all arrow into the middle column, "Point in time: snapshot", a gzipped JSON file that is owner-scoped, with atomic save and deltas versus the previous snapshot.
- The snapshot arrows into the right panel, "QVCCS Knowledge Manager: review without re-querying".
- The portal has four parts: Overview with five stat cards (bases, categories, articles, variations, labels); Knowledge base browser (category tree, search, state, label chips); Article reader (Published and Draft tabs, safe rendering); and Exports (inventory CSV, snapshot JSON.gz, article JSON).
- A banner says: harvest once, then browse instantly, so reviews put no further load on Genesys Cloud.
03
What the native Knowledge workbench already offers
The Knowledge workbench V2 is a capable authoring environment, and authors should stay in it. It lets you create and organise articles with categories and labels, create touchpoint variations, test queries, publish, edit multiple articles at once, view and restore previous versions, review article feedback, follow the Knowledge Performance dashboard and use the knowledge optimizer to improve the health of a knowledge base. For data movement, a knowledge base can be exported to CSV, JSON or XLSX – all articles or a specific time period – and the export includes the knowledge base ID and the linked category and label structures.
Knowledge Manager is not an authoring tool and never writes anything. It adds a review lens across the whole workbench: every knowledge base in one overview, five clickable totals that open the exact rows behind each number, a category tree with counts, side-by-side published and draft content in a reader, deltas against the previous harvest and a single archivable snapshot of the org's knowledge at a moment in time. It is built for reviewers, auditors, bot designers and project teams, alongside the authors who keep using the workbench.
04
Engineering a safe, resumable harvest
A harvest is a server-side job keyed to the region and OAuth client that started it, not to a browser session. Navigate away or sign out and it keeps running; sign in again with the same client and the UI reattaches to it. Only one harvest per client runs at a time, so if a colleague on the same client has already started one, your Harvest button simply adopts its progress. A banner shows the phase, knowledge base and article counters, the running API-call count and any warnings.
Rate limits shape the design. Genesys calls are paced at least 210 ms apart per session, about 4.7 requests per second, and article-content fetches run through a four-worker pool that still shares the pacer. Any 429 honours the Retry-After header and widens the spacing, so harvests are safe during business hours. The guide's rough cost model is about two API calls per article plus listings: a 500-article org is around 1,000 to 1,100 calls and about four minutes.
Failure handling is explicit. A failed item – one knowledge base's labels, one article's content – becomes a warning while the rest of the harvest carries on, and up to 50 warnings are stored in the snapshot for review afterwards. Cancel stops cleanly at the next step boundary and writes no partial snapshot, and saves are atomic, so a crash mid-save can never leave a corrupt snapshot behind. The Genesys client is GET-only by construction: there is no code path that can create, modify or delete anything in your org.
05
Browse, read and review from the snapshot
The overview opens with five stat cards – knowledge bases, categories, articles with a published, draft and archived split, content variations and labels – each with change-since-last-harvest markers. Click any card and a drawer lists the rows behind the number, with a filter box, click-to-sort columns and a button that exports exactly the filtered, sorted view to CSV. Click a knowledge base to browse it: a category tree with per-node counts, search across titles and alternative phrases, a state segment, label chips and a sortable article table.
The article reader shows the full breadcrumb, state, a hidden-from-search pill where relevant, labels, authors, dates, published version, word count and every alternative phrase. Published and Draft tabs appear only for sides that hold content, with a picker when a side has several variations, so an in-flight edit is obvious at a glance. Rendering is faithful – nested lists, tables, images, video links and marks – and safe: bodies are built strictly as DOM text, never injected as HTML, so a hostile article cannot script the page. Unknown block types appear as an inspectable panel rather than being dropped.
Every article has a deep link, so a reviewer can send a colleague straight to the article that needs attention; after they sign in, the link opens onto it. The reader footer also offers Copy JSON and Download JSON for the complete raw article record, which is often the quickest way to show a bot designer exactly which blocks, marks and variations an answer contains.
Read this diagram as text
Three stacked bands show the sequence: harvest, then review entirely from the snapshot, then export.
- Band 1, "Harvest · paced, resumable: one click walks every knowledge base", runs left to right through Knowledge bases, Categories and labels, Listing articles, Article content and Saving snapshot.
- Band 2, "Review · no further Genesys calls: from a number to the article behind it", runs left to right through four steps.
- The review steps are Overview (five stat cards, deltas), Drawer (rows behind the number), KB browser (category tree, labels) and Article reader (Published and Draft tabs).
- Band 3, "Export", lists four outputs: Inventory CSV of every article, Raw snapshot as JSON.gz, Drawer view as CSV, and One article as JSON.
06
Exports that make knowledge auditable
Four exports cover the common needs. The inventory CSV holds one row per article across every knowledge base: knowledge base, category path, title, state, visibility, labels, alternative-phrase count, word count, has-published and has-draft flags, dates, published version, authors and document ID – in UTF-8 with a byte-order mark so Excel opens it cleanly. The raw snapshot downloads as gzipped JSON with all article bodies included: an archivable, machine-readable record of the org's knowledge at a point in time. Each stat drawer exports its current view, and each article exports its full record as JSON.
The app keeps one snapshot per owner rather than a history, so the guidance is simple: download the raw snapshot before re-harvesting if you want to compare two points in time. Snapshots are owner-scoped to the region and client ID that harvested them; a different client, even in the same org, gets its own, and the purge button deletes only your own snapshot. Drafts may contain unreleased material, so we recommend treating snapshots with the same care as the knowledge base itself.
07
Built for the knowledge lifecycle, backed by the practice
Knowledge Manager follows the same lifecycle as our client work. Requirements came from knowledge reviews before bot and Agent Assist launches; the design separates a light snapshot for fast boot from full bodies streamed per article; and the build was peer reviewed against the Senior Platform Practice Lead's standards, with negative paths – 403s, permission propagation lag, legacy V1 knowledge bases, missing sides – each producing a clear, documented outcome. The client role needs only the Knowledge view permissions for knowledge bases, categories, documents and labels, within the divisions it covers.
The app is part of the QVCCS App Suite included with every Managed Professional Services tier. When the review turns into a knowledge programme – restructuring categories, designing touchpoint variations, preparing content for bots and Agent Assist – we muster a squad from our own bench: Senior Business Consultants and Business Analysts for content governance, a Solution Architect for the knowledge and bot design, Trainers for the authors, and the support engineers who keep it healthy in run and evolve.
How it compares
The Knowledge workbench and Knowledge Manager
The workbench is where knowledge is authored and published; Knowledge Manager is a read-only review and inventory lens over it.
| Aspect | Native Genesys Cloud CX | QVCCS Knowledge Manager |
|---|---|---|
| Purpose | Create, organise, test, publish and optimise articles, with version history and a Knowledge Performance dashboard. | Read-only, point-in-time inventory and review of everything in the workbench; never edits content. |
| Scope of export | Per knowledge base: all articles or a specific time period, in CSV, JSON or XLSX. | Every knowledge base the client can see in one harvest, plus an inventory CSV across all of them. |
| Structure | Exports include the knowledge base ID and linked category and label structures; externalID and structure only in JSON. | Full category hierarchy, labels with colours and per-node counts, rebuilt from parent references. |
| Draft and published | Articles are authored as drafts and published in the workbench. | Bodies captured for both sides of every article, shown in Published and Draft tabs side by side. |
| Change over time | View and restore previous versions of an article. | Overview deltas against the previous harvest; raw snapshot downloadable for external comparison. |
| Review workflow | Filter articles, review feedback and use the knowledge optimizer inside the workbench. | Clickable totals with row drawers, search across titles and alternative phrases, deep links to articles. |
| Platform load | Interactive use of the workbench. | Paced harvest, then browsing works entirely from the snapshot with no further Genesys calls. |
Knowledge Manager harvests Knowledge Workbench V2; legacy V1 knowledge bases are listed with their metadata, but their content is not captured.
The takeaways
- One click captures every knowledge base, category, label and article, with full bodies.
- Draft and published content side by side, so reviews catch what is about to go live.
- Clickable totals, a category browser and a safe reader make reviews fast and focused.
- Inventory CSV and a gzipped JSON snapshot give auditors an archivable record.
- GET-only, paced and owner-scoped, so harvests are safe during business hours.
Knowledge Manager is part of the QVCCS App Suite, included with every Managed Professional Services tier and built by the same certified team that designs, builds and supports Genesys Cloud CX solutions.
Sources
- About the knowledge workbench V2 – Genesys Cloud Resource Centerhelp.genesys.cloud
- Import or export articles in a knowledge base V2 – Genesys Cloud Resource Centerhelp.genesys.cloud
- Limits – Genesys Cloud Developer Centerdeveloper.genesys.cloud