Client Pilot Information Request

Section A · Questions 01–05

Use case

Define the data product and the evidence that will make this pilot successful.

QUESTION 01

What data product / report should the pilot produce?

Example answer: A weekly customer-revenue dashboard for Finance and Sales, showing recurring revenue, overdue invoices, and account-level trends.

QUESTION 02

Which 1–3 source tables should we use?

Use fully qualified names where possible, one table per line.

Example answer: crm.accounts, billing.invoices, and product.usage_events (one table per line).

QUESTION 03

What does success look like for you?

Example answer: Refresh completes within 20 minutes, totals reconcile to Finance within 0.5%, and the Finance Director approves the result.

QUESTION 04

Approx. data volumes (rows / size)?

Approximate rows Approximate size

Example answer: Approximately 2,500,000 rows totaling 18 GB.

QUESTION 05

Pilot duration and who signs off?

Duration in weeks Target end date Sign-off owner

Example answer: Four-week pilot ending 15 September 2026; sign-off by Sarah Chen, VP of Data.

Section B · Questions 06–11

Source database

PostgreSQL / Snowflake / Oracle / MySQL / SQL Server / etc.

QUESTION 06

Which database is it? Host / port / database (or service) name?

Database type * Host Port Database / service name

Example answer: PostgreSQL; host db.pilot.example.internal; port 5432; database analytics.

QUESTION 07Can you provide an account with SELECT on the source schema (profiling)?

Yes Pending approval Not currently

Account / access notes

Example answer: Yes—profiling_reader can receive SELECT access to the CRM and billing schemas; credentials will be shared through an approved secure channel.

QUESTION 08Can you provide a sandbox schema with write access for deployment?

A single “gateway” user with both source SELECT and sandbox ownership is fine.

Yes Pending approval Not currently

Sandbox schema / notes

Example answer: Pending security approval; the celvari_pilot sandbox schema will be owned by the gateway user.

QUESTION 09

Database version?

Example answer: Oracle Database 19c Enterprise Edition.

QUESTION 10

How is the source schema populated / refreshed (Airbyte / CDC / ETL / static snapshot)? Celvari reads it; it does not load it.

Refresh method Refresh frequency

Additional details

Example answer: Airbyte sync every 15 minutes, followed by a reconciliation job nightly at 02:00 UTC.

QUESTION 11

Any known data-quality / business rules for these tables?

Example answer: account_id must be unique, invoice totals cannot be negative, and every invoice must reference an existing account.

Section C · Questions 12–13

Metadata catalog

QUESTION 12 Do you already run OpenMetadata? If yes, URL + ingestion-bot token? (If no, we will deploy it.)

We already run itProvide the base URL and token below. Celvari should deploy itInclude it in the pilot stack.

OpenMetadata base URL Ingestion-bot token

Example answer: We run OpenMetadata at https://metadata.example.com; the ingestion token is available and will be shared through an approved secure channel.

QUESTION 13 · OPTIONAL

An OM ingestion-pipeline ID to trigger after deploys?

Example answer: Pipeline ID 3f77c70b-1f73-4d91-a2d1-7fb72ebc82ae.

Section D · Questions 14–16

AI / LLM

QUESTION 14

Agent LLM key — OpenAI / Anthropic / Google (or an OpenAI-compatible endpoint such as OpenRouter)?

Agent provider * Preferred model Agent API key Compatible base URL

Example answer: Azure OpenAI with GPT-4.1 at https://pilot-ai.example.com/v1; the API key is available through an approved secure channel.

QUESTION 15

Embeddings key — an OpenAI key for RAG/glossary/knowledge-base features.

Required even if your agent provider is Anthropic/Google because embeddings are OpenAI-only. Without it, AI profiling runs with no knowledge-base context.

Example answer: An OpenAI embeddings key is available and will be shared through an approved secure channel.

QUESTION 16 Any rule that data must not leave your tenant / region? (Affects both chat and embeddings.)

Yes No Needs confirmation

Constraint details

Example answer: Yes—data storage and model inference must remain within our Azure West Europe tenant.

Section E · Question 17

Documents

QUESTION 17Can you share a business glossary / data dictionary for these tables? (Digital PDF or Word — no scanned images.)

Business glossary Data dictionary Domain documentation None available

Delivery link or notes

Example answer: A business glossary and table-level data dictionary are available as digital Word and PDF files through the client SharePoint.

Section F · Questions 18–20

Hosting & network

QUESTION 18 Should we host, or will you provide a VM (Linux, Docker)?

Celvari hostsPilot data flows to the Celvari environment. We provide a VMLinux x86_64 with Docker and Docker Compose.

PROFILE A

Own OpenMetadata: ~8 vCPU / 16 GB RAM / 120 GB disk.

PROFILE B

Celvari deploys OpenMetadata: ~16 vCPU / 32 GB RAM / 200 GB disk.

Expected profile Raw pilot data (GB)

Example answer: Client-hosted Ubuntu 24.04 VM with Docker, 8 vCPU, 32 GB RAM, 250 GB disk, and approximately 18 GB of pilot data.

QUESTION 19

How will the platform reach your source DB (VPN / tunnel / allowlist)?

Connectivity method * Network owner / team

Connection details

Example answer: Site-to-site VPN; the DBA team owns allowlisting; permit TCP 5432 from the Celvari VM to the source database.

QUESTION 20Do you need the UI over HTTPS with your own domain? Can you provide the DNS records (and origin cert, if used)?

Yes No Undecided

Preferred hostname DNS / certificate notes

Example answer: Yes; use pilot.example.com, with DNS managed in Cloudflare and an origin certificate supplied by the client.

Section G · Questions 21–27

Access & security

QUESTION 21

Who is the initial admin (name + email)?

Admin name Admin email

Example answer: Jordan Lee, jordan.lee@example.com.

QUESTION 22

Login: built-in accounts, or SSO (SAML / OIDC)? If SSO, IdP details?

Login method Public callback base URL

IdP / OIDC details

Example answer: Microsoft Entra ID via OIDC, callback base URL https://pilot.example.com; the IAM team will provide client details.

QUESTION 23

Who are the pilot users (emails) and who should be admin?

Example answer: Jordan Lee — jordan.lee@example.com — admin; Priya Shah — priya.shah@example.com — user.

QUESTION 24

Data residency requirements?

Residency requirement * Country / region / tenant detail

Example answer: EU-only residency within the client's Azure West Europe tenant.

QUESTION 25 Will pilot data contain PII? Can it be masked / anonymized?

Contains PII *

Yes No Unknown

Masking available

Yes No N/A

PII / masking details

Example answer: Yes—customer names and emails are present; use hashed customer IDs and masked email addresses in the sandbox.

QUESTION 26

NDA / DPA needed before we receive access?

NDA required DPA required

Current legal status Legal owner / notes

Example answer: NDA signed; DPA in review; legal owner Maya Patel with completion targeted for 30 August 2026.

QUESTION 27

Security contact for firewall / access approvals?

Security contact name Security contact email

Example answer: Omar Hassan, security@example.com.

Section H · Question 28

Contacts

QUESTION 28

Business owner, DBA, IT/security, and day-to-day champion (names + emails)?

Business owner

DBA / data engineer

IT / security

Day-to-day pilot champion

Example answer: Business: Jordan Lee; DBA: Priya Shah; Security: Omar Hassan; champion: Maya Patel—each with their company email and phone.

Section I · Question 29

Logistics

QUESTION 29

Preferred kickoff date and communication channel?

Preferred kickoff date Communication channel Preferred cadence

Example answer: Kickoff 1 September 2026; Microsoft Teams; 30-minute stand-up every Tuesday and Thursday.

Let’s shape a pilot that proves value fast.