NexoraAI 纽智
中日EN
Book a demo

On-premise AI · Automotive & advanced manufacturing

Turn documents, knowledge and testing into
one chain

A private-deployment AI engine for automotive and advanced-manufacturing R&D: documents stay on-premises, knowledge is queryable, and testing is traceable.

Book a 30-minute demo See how the products work together

// Three products plus one unified entry, all supporting on-premises deployment, with model endpoints fully local

≥95% Requirement-to-test traceability coverage NexTest Studio
≥80% Version change impact recall NexTest Studio
3.54s Knowledge Q&A end-to-end response Incl. generation 2.86s · With citations
0 Depends on an external RAG platform Self-built retrieval layer

The Problem

In R&D, the three most expensive problems all happen after the document

Finishing the specification is just the beginning. What really consumes engineering hours is version verification, hunting for knowledge and test traceability — and most of these steps still rely on people.

01

Versions everywhere, the latest one confirmed by someone shouting across the room

Specifications are scattered across drives and shared folders. Tests are written against old versions, and defects only surface at the vehicle validation stage.

02

The link from requirement to test case lives only in one person's memory

One transfer or resignation and the traceability chain breaks. When an audit comes, it cannot be pieced back together.

03

Mountains of documents, half a day to find one

Scanned files can't even be read. The same questions get asked and the same work redone, and onboarding relies entirely on word of mouth.

04

Public LLMs are useful, but R&D data cannot leave the premises

Either you can't use AI, or you carry the data risk yourself. Most R&D organizations won't accept that trade.

Product Line

Three products, each owning a stage — linked into one chain

They can be deployed separately or run together. Run together, documents, knowledge and tests all reference the same facts — and with a unified entry, cross-system questions are answered in one shot.

Source · govern documents

NexSpec Hub

Specification and document master data hub

The single authoritative source for specification master data, source files and version history. Non-overwritable source files, SHA-256 verification, a draft-to-release status flow and a trilingual Chinese / Japanese / English interface.

What it solves:Whether the copy in hand is the latest version no longer needs to be confirmed by hand.

View NexSpec Hub

Know · make knowledge usable

Nex Knowledge Base

On-premises Knowledge search and Q&A

Structure-aware chunking, dual-path retrieval, RRF fusion and reranker re-ranking all run on-premises, with no connection to any external RAG platform.

What it solves:Files can be judged without opening them, and Q&A answers open straight to their sources.

View Nex Knowledge Base

Test · turn it into an asset

NexTest Studio

AI test closed-loop platform for automotive software

Specification parsing, atomic requirement extraction, test specification generation, full-chain traceability and change impact analysis are carried out by ten collaborating Agents.

What it solves:From specification to test case, along one traceable, reusable automation chain.

View NexTest Studio

NexSpec Hub governs document master data → Nex Knowledge Base makes knowledge searchable and Q&A-ready → NexTest Studio turns knowledge into testing assets

Unified Entry

One entry that ties the three together

NexSpec Hub answers "what it should be", NexTest Studio answers "what has actually been verified", and the internal knowledge base answers "what happened before, and on what grounds". Nexora Assistant assembles the three into one traceable answer, without duplicating any product capability.

01 NEXSPEC HUB What it should be Version uniqueness, SHA-256 chained audit, fixed-version dereferencing. No one else can
02 NEXTEST STUDIO What has actually been verified Test cases, executions, defects, change impact and quality metric snapshots. No one else can
03 KNOWLEDGE BASE What happened before Cited unstructured search: minutes, reports, manuals, past issues. No one else can
04 NEXORA ASSISTANT Assembled into one traceable answer Expert routing, three-way parallel evidence gathering, and source annotation. Read-only: it never changes the state of any business database.
See how this entry works

Product Snapshots

See the interface first, then the story

Below are reconstructions of the core interfaces of all three products. The version numbers, statuses, citation sources and relevance scores you see are real fields in the system — not concept mockups.

01 / NexSpec Hub

Specification directory: see at a glance which version is authoritative

The top of the page shows four numbers first — total specifications, released versions, versions pending release and draft versions — and the list lets you read each document's current released version, document type, owner and last update side by side — no need to click in to confirm.

  • Files with the same structure and name are automatically appended as revisions, creating no duplicate documents
  • Source files are locked the moment they are ingested, verified with SHA-256
  • Released files cannot be overwritten; changes require uploading a new version

Go to NexSpec Hub

nexspec.internal/Specification library
Figure 1 · NexSpec Hub specification directory

02 / Nex Knowledge Base

Knowledge base Q&A: sources and scores attached beneath every answer

The sidebar groups items into 'Personal knowledge base / Shared knowledge base', and each answer states up front where the data goes. Every citation marks the referenced file and location, and shows the re-ranked relevance score — whether to trust it is your call.

  • Structure-aware chunking, where clause numbers and table rows inform boundary detection
  • Dual-path retrieval followed by RRF fusion and Reranker re-ranking
  • All three endpoints — generation, vectors and re-ranking — sit on-premises

Go to Nex Knowledge Base

kb.internal/Knowledge Q&A
Figure 2 · Nex Knowledge Base Q&A with citations

03 / NexTest Studio

Test execution: status and progress for every test case

The top bar shows the current module, and a command palette jumps straight to any Case or task. Each test case lists its system Case ID, revision number, run count and the software version it is based on, with step progress and pass / fail / blocked status on the right.

  • Document → Requirement → Atomic → TestCase → Defect
  • Version diffs automatically surface affected test cases and the recommended regression scope
  • Every AI output is traceable to the chapter, page and paragraph of the source

Go to NexTest Studio

nextest.internal/Test execution
Figure 3 · NexTest Studio test execution

Value Chain

Five stages on one chain

From a specification entering the repository to a change impact analysis producing results, every step in between leaves a record, is traceable and can be reviewed.

01 NEXSPEC HUB Specification Collects all of the company's specifications and source files, fixes versions, and leaves SHA-256 verification and audit records.
02 NEXTEST STUDIO Structured requirements Parses Word, Excel, PDF and PPT specifications and extracts testable atomic requirements.
03 NEXTEST STUDIO Test specification Combines the test viewpoint library, test factor library and defect knowledge base to generate specifications covering purpose, preconditions, steps, expected results and pass criteria.
04 TRACE GRAPH Traceability graph Builds full-chain traceability for Document → Requirement → AtomicRequirement → TestCase → Defect, with coverage ≥95%.
05 CHANGE IMPACT Change & Regression Version diffs surface affected test cases and the recommended regression scope, with change impact analysis recall ≥80%.
See how this chain runs

Deployment & Security

Bring the capability in, keep the data where it is

Rather than sending data out in exchange for capability, we deploy the models, search, indexing and auditing entirely inside your on-premises network.

Private deployment

Every product deploys inside your network. Neither the server side nor the model side needs a public egress.

Offline operation on-premises

Generation, vectorization and re-ranking are all handled by on-premises model endpoints, and it keeps running with the outside network cut off.

Data never leaves the domain

Documents, vectors, indexes and logs all stay in your data center; not a single piece of data flows to a third party.

Project-level permission isolation

Permissions are divided by project. The system decides who can see which project and which knowledge base.

Non-overwritable source files

Source files cannot be overwritten once ingested, with SHA-256 verification and duplicate detection giving them a verifiable identity from the moment they enter.

SHA-256 audit chain

Version baselines are fixed, and audit records are appended in a SHA-256 chain. Whether something changed — and who changed it — can be verified.

Corporate on-premises network · ON-PREMISE
NexSpec HubDocument master data
Nex Knowledge BaseHybrid search & Q&A
NexTest StudioClosed-loop testing
Nexora AssistantUnified entry
Local model endpointsGeneration · Embedding · Reranker all deployed on-premises
Data stays in placeSource files · vectors · indexes · audit logs
External RAG platforms / cloud LLMs DATA_EGRESS: BLOCKED

Figure 4 · Data boundary: all products and model endpoints sit inside your network, and no data needs to leave

Who It's For

Teams already doing this

Automotive R&D

From body and cockpit to ADAS, diagnostics and HIL / SIL validation, covering the full path from requirement to test.

Tier1 suppliers

One master data system handles the specification formats and acceptance requirements of multiple OEMs.

Manufacturing R&D centers

Turns equipment manuals, process documents and standards into searchable, queryable intranet knowledge.

ECU domain controller Smart cockpit ADAS Diagnostics HIL / SIL Vehicle validation Specification management Cross-border JP / CN / EN projects

FAQ

Six things you'll probably ask before go-live

Q1Can it run fully offline?

Yes. Every product supports on-premises deployment on your internal network, and the model endpoints also deploy internally, so no public egress is needed. Nex Knowledge Base connects directly to an internal llama.cpp model endpoint and runs even offline.

Q2Does it support Chinese, Japanese and English?

Yes. NexSpec Hub offers interfaces in Simplified Chinese, Japanese and English; search and Q&A are tuned for both Chinese and English documents — Chinese uses a bigram inverted index, English uses a word-level inverted index.

Q3How does it integrate with existing test management systems?

NexTest Studio outputs test specifications and traceability relationships through APIs, and can integrate with the field formats of your existing test management system. The exact API scope is confirmed during the environment assessment phase.

Q4How is data security ensured?

On-premises deployment, no data egress, project-level permission isolation, SHA-256 verification and append-only chained auditing, non-overwritable source files, and downloads via a five-minute, single-version read-only ticket that expires the moment it is revoked.

Q5How long does implementation take?

Implementation follows a standard process: environment assessment → on-premises deployment → knowledge base cold start → trial run → go-live. The timeline depends on data volume and integration scope; a concrete schedule is provided after assessment.

Q6Does it support local LLMs?

Yes. Generation, Embedding and Reranker can all connect to your own local model endpoints, or you can use the default configurations we support.

Request a Demo

In 30 minutes, watch it run on your own data

Bring a real specification and we will demo parsing, extraction, traceability and Q&A live.

Book a 30-minute product demo futurerr@sina.com