What this page is
We don't sell software seats. We deliver a finished asset: your scattered documents turned into a knowledge base that answers with citations and passes acceptance β then wired into your customer service and public content.
So this page covers two things only: what each service delivers and what counts as delivered. Pricing is quoted per project β contact us on WeCom (QR at the bottom).
Service 1 β Enterprise knowledge base
| Deliverable | What it contains |
|---|---|
| Inventory & cleaning report | What you have, duplicates and expired items flagged, usability tiering |
| Structure design | Bucket scheme (FAQ and policy clauses in separate indexes), metadata fields |
| Ingestion & retrieval setup | Chunking rules, index isolation, citation back-fill |
| Evaluation set | Real questions with expected answers and sources |
| Operations manual | Routine maintenance, incremental ingestion, self-troubleshooting |
Acceptance: re-run the evaluation set β accuracy, citation correctness and empty-answer count reported. We fix and re-submit until it passes.
Service 2 β AI customer service & national-standard compliance
Handover rules and audit trails designed against GB/T 47746β2026, not just a chat widget.
| Deliverable | What it contains |
|---|---|
| Self-check sheet | All 61 check items (48 shall / 4 should / 9 may) with verdict and clause reference; the 5 one-vote items flagged separately |
| Gap list & remediation advice | Ranked by cost Γ risk, each item mapped to a clause |
| Handover rule file | Which scenarios must escalate, emotion triggers, what the customer is told β configurable and traceable |
| Regression acceptance report | Resolution rate, wrongly-escalated count, missed-escalation count, empty answers |
Acceptance: spot-check 10 items against the standard text; re-run the regression set on site and compare.
Service 3 β Answer-layer optimisation (AEO, also called GEO)
Making AI search able to find you, answer about you and cite you β assets, not ranking games.
| Deliverable | What it contains |
|---|---|
| Fact base & 9-column provenance ledger | Every public claim carries source, collection date and verification method |
| Source-tier escalation plan | Own site first (structured data + fact pages), then authoritative platforms and public registries |
| Process verification record | Regression re-runs, rule-change log, monthly observations (same question β₯3 samples per week) |
Acceptance: facts can be spot-checked back to source; process records can be re-run by the client. We do not promise rankings or citation rates β the applicable standard, T/CAPT 026β2026, explicitly prohibits service providers from promising such outcomes.
Self-serve tools
Portable, zero-install kits are sold through our Taobao shop, at the price shown there. They are separate from enterprise knowledge-base and compliance delivery.
What we don't take on
- Projects requiring enterprise qualifications we don't hold (we quote within our real capability)
- Hardware, motion control, model training
- Any promise of rankings, citation rates or blanket coverage
- Restricted data (personal data, contract floors) is never sent to overseas APIs
How quoting works: document inventory first β problem list and plan β then a quote. Remote-first; on-site in Chongqing available.
For pricing, contact us on WeCom (note “Knowledge Base”).
WeCom: scan the QR code above (note “Knowledge Base”)
FAQ
Why don't you publish prices?
Because a knowledge-base project's price depends on volume, structure and integrations. A public number would not be meaningful for your case. We inventory, then quote.
What should I prepare before asking for a quote?
Company size, the scenario (internal Q&A or customer-facing), and the rough volume and formats of your documents.
How do I reach you?
Add us on WeCom (note “Knowledge Base” or “AI Service”), or use the contact section on our About page.