A knowledge layer for Vietnamese enterprises
Knowledge helps leaders and IT, Operations, Product, HR and Support teams reach information in context while respecting source permissions. Every answer stays connected to evidence that people can inspect instead of becoming an opaque conclusion.
One workbench, three layers of context
The workbench below uses fictional data to illustrate how Search, Grounded Assistant and Context Graph can work together. It is not a screenshot of a live product environment.
Results stay within your access scope
Updated 2 days ago · Revenue Ops
Source: ConfluenceApproved version · Finance
Source: Internal fileWho needs to approve, and what evidence is required?
The pricing owner confirms the proposal, Finance checks the margin, and Operations records the effective date. The final decision should cite the policy and review minutes.
Both sources are inside the access scope, refer to the same “Q4 pricing” entity and are ranked using the trust of controlled documents.
The problem is not a lack of documents
When information, access and context are separated, teams spend time verifying what should have been clear from the start.
Files, wikis, project discussions and operating guidance live across systems. Different names make one issue look like unrelated fragments.
A result is useful only when the viewer has access at the source. Ignoring ACLs may create a fast answer, but it does not create a safe one.
A summary without citations is hard to inspect, difficult to explain to colleagues and risky to use as the basis for the next action.
Value model
Connect → Understand → Find → Act → Govern is an operating model for how Knowledge creates value. It does not claim every step is complete; the public status ledger below states what is actually available.
Bring content and access metadata from supported sources into a controlled flow.
→Recognize relationships between documents, entities, owners and decisions.
→Retrieve by semantics, keywords, trust and ACL rather than a single signal.
→Bring cited answers into work within the limits of current capabilities.
→Observe explanations, application access and technical operations.
Core capabilities
Every capability references narrow items in the status ledger. The labels below distinguish what is available, in progress and planned.
ACL-aware retrieval is available; the cross-source Unified Search experience remains in progress.
Answers combine citations, trust and reranking so people can inspect the source and the reason it was prioritized.
File upload and Confluence are available. Production Jira, GitLab, Drive and SharePoint connectors are in progress.
The graph connecting sources, entities and decisions is available; people and expertise remain in progress.
Hybrid retrieval and explainability show what evidence was found and why a result received priority.
ACL-aware retrieval and API-key Apps form a control foundation; curation is in progress and DLP/Protect is planned.
The operations workbench offers technical visibility for operating the system. This is not a claim about customer or business analytics.
Agents/actions, SDK/MCP and browser, Slack and Teams surfaces are planned and are not represented as available.
How it works
Knowledge does not separate an answer from the source and permissions that produced it.
Ingest files or Confluence with source metadata and ACLs. Other production connectors are represented accurately as in progress.
Combine Qdrant/OpenSearch, graph context, trust and reranking to find evidence using more than one signal.
Grounded chat presents citations and explanations beside the answer so teams can evaluate it before taking action.
Permission architecture
The architecture shown here is limited to evidenced ACL metadata, retrieval filtering and explainability. Knowledge does not claim unverified SSO, SCIM, data residency, compliance certification or production hardening.
Files and Confluence in the current scope
→Preserve source access metadata
→Qdrant + OpenSearch
→Filter before results return
→Rerank by context and trust
→Citations + explanation
→Technical operations visibility
DLP/Protect is planned and is not represented as an available protection layer.
Transparent status
This ledger is the public claim boundary. It contains no customer metrics, connector count, compliance badge or feature-parity promise without evidence.
Supported by evidence in the current code.
Being completed for production use.
A direction, not a commitment of availability.
Frequently asked questions
Clear answers about scope, permissions, sources and how to start a useful conversation.
No. This is a B2B landing page that explains the product direction, architecture and status. It has no account, product backend, live search or customer data.
The evidenced scope preserves source ACL metadata and filters retrieval results by access before returning them. This page does not extend that claim to SSO, SCIM, data residency or compliance certification.
File upload and Confluence ingestion are available. Production Jira, GitLab, Drive and SharePoint connectors are in progress, and the page deliberately publishes no aggregate connector count.
Grounded chat, citations, trust/reranking and explainability are in the available scope. People can inspect referenced sources and why evidence was prioritized within their access boundary.
Not yet. Agents/actions, SDK/MCP and browser, Slack and Teams surfaces are explicitly planned. The landing page does not simulate real actions or request integration access.
Choose one concrete decision flow, list the sources involved, describe access groups and define what makes an answer trustworthy. Cloud Solution can then discuss that context against the current product scope.
Start with a real decision
Choose a workflow that currently consumes time in searching or verification. We can map its sources, permissions, evidence and path to a suitable rollout without using a fake lead-capture form.