The Customer 360 Banking Demo brings customer, account, transaction, and CRM data together from separate banking systems. TapData CDC captures changes, transformations build a unified Customer 360 model in MongoDB, and published APIs supply the dashboard.
BankStream AI Dashboard displays this integrated data and provides AI Analysis, AI Chat, and suggested next actions. It is a reference application for internal training, partner demonstrations, and development.
AS400 + Oracle → TapData CDC → FDM (MongoDB)
→ MDM (MongoDB Customer 360) → TapData APIs → AI Dashboard
AS400 holds core banking records, and Oracle holds CRM data. TapData replicates source data into FDM, then joins and transforms it into MongoDB MDM models. The dashboard reads published APIs through its Node.js backend. TapData maintains the models; API and AI credentials stay on the backend.
- Git, npm, and Node.js 22.9+ or 20.18+ for the dashboard.
- IBM i / AS400 (DB2 for i) and Oracle databases with demo schemas and sample data. Partners using Tapdata-hosted databases do not need to prepare these.
- Access to the TapData APIs from the environment where the AI dashboard is deployed.
- Published TapData API URLs and authentication credentials accessible from the dashboard host.
- Optional: A DeepSeek or OpenAI API key and provider connectivity for live AI features.
The complete demo uses two repositories: BankStream Feeder simulates banking activity in the AS400 and Oracle source databases, including KYC updates, account balance changes, and transaction inserts. BankStream_AI_Dashboard displays these updates in real time. The Feeder repository also includes source database preparation, TapData tasks, configuration files, and setup instructions. Each repository documents its own setup requirements.
For source database and TapData pipeline setup, follow the BankStream Feeder README. Once the published APIs are available, install and configure the dashboard below.
-
Clone the dashboard repository.
git clone https://github.com/tapdata/BankStream_AI_Dashboard.git cd BankStream_AI_Dashboard -
Install dependencies.
npm ci
-
Copy the example configuration. Choose the command for your terminal.
Windows PowerShell:
Copy-Item .env.example .envLinux/macOS:
cp .env.example .env
-
Choose sample data or live data. Set one of these values in
.env:Value Dashboard data What npm testdoesDATA_SOURCE=demoSample data; no TapData connection needed. Runs application tests and skips real TapData API calls. DATA_SOURCE=tapdataReal data from TapData. Also calls the 13 TapData APIs that feed the dashboard and checks that each returns records. -
Optional: enable AI Analysis and AI Chat. In
.env, setAI_PROVIDER,AI_MODEL,AI_BASE_URL, and your provider's key:DEEPSEEK_API_KEYorOPENAI_API_KEY. See .env.example for examples. You can skip this step to view dashboard data and runnpm test. -
Run tests and start the dashboard. If you chose
demo, no TapData credentials are needed.TapData API Configuration
If you chose
tapdata, use the URLs for your deployment in.env.TapData-Hosted
Use these URLs only if TapData hosts your TapData environment.
TAPDATA_API_BASE_URL=http://113.98.206.139:43080 TAPDATA_TOKEN_URL=http://113.98.206.139:43030/oauth/token
Self-Hosted
Configure all relevant
.envvariables for your own TapData deployment. These URLs are examples for a local deployment. Use your own hosts and ports as needed.TAPDATA_API_BASE_URL=http://127.0.0.1:3080 TAPDATA_TOKEN_URL=http://127.0.0.1:3030/oauth/token
Client credentials
For either deployment, log in to your TapData environment and go to Data Service → API Client. Find the API Client ID and Client Secret and use them in
.env.TAPDATA_CLIENT_ID=YOUR_TAPDATA_CLIENT_ID TAPDATA_CLIENT_SECRET=YOUR_TAPDATA_CLIENT_SECRET
For the reference setup, keep the other TapData settings from
.env.example.From the dashboard folder, run:
npm testIf
npm testfails only on the live TapData checks, you can still run the start commands below to explore the dashboard. However, live data may be missing or incomplete, so its accuracy cannot be guaranteed until the TapData issue is fixed. Check your connection and credentials, or useDATA_SOURCE=demofor sample data. Investigate any other test failures before continuing.Start the dashboard:
npm run service:start npm run service:status
-
Open http://127.0.0.1:3001. Confirm customer and transaction data appears, then try AI Analysis and AI Chat. Restart with
npm run service:restartafter changing.env.For foreground operation, use
npm startand stop with Ctrl+C. Background mode does not automatically restart after a computer reboot.
For a UI preview without live integration, set DATA_SOURCE=demo. Sample data does not demonstrate CDC.
Optional LAN access: Set HOST=0.0.0.0 and the desired PORT in .env, restart, and use the LAN URL shown by service status from another computer. The HOST=127.0.0.1 allows access only on the host. 0.0.0.0 is a listening address, not a browser URL. Your host firewall and network must permit the connection; the application does not configure them. Production requires the existing trusted identity proxy and network controls.
- Open the dashboard and go to Customer 360. Search for customer ID 100002 and open their details. In Accounts, find the Checking account with account ID 1000021 and note its balance. Review the transactions for this account and note a transaction amount, then open KYC and note the current KYC status. Compare these values after making the source changes below.
- In TapData, confirm that the source-to-FDM replication tasks and FDM-to-MDM transformation tasks are running, and that the published APIs are available. Keep these tasks and APIs running during the demo.
- Use BankStream Feeder to generate a transaction or KYC update for that customer. Follow the Feeder README for all simulation and verification commands.
- Allow time for the source changes to reach MDM and appear in the API results. Refresh or reopen the dashboard view and confirm the same customer's updated information.
| Area | What it shows |
|---|---|
| Overview | Customer and transaction KPIs, segments, and top recommendations. |
| Customer 360 | Unified customer details and related banking information. |
| Transaction & Fraud Intelligence | Transaction activity, filters, review signals, and investigation context. |
| AI Analysis | AI-assisted insights from available customer data. |
| AI Chat | Natural-language questions through the configured AI provider. |
| Next Best Actions | Suggested service or engagement follow-ups. |
Fraud indicators are deterministic review signals, not confirmed fraud. AI suggestions require human review before any banking action.
Run these from the dashboard repository:
| Action | Command |
|---|---|
| Start in the background | npm run service:start |
| Stop with confirmation | npm stop |
| Restart | npm run service:restart |
| Show status and addresses | npm run service:status |
| View recent logs | npm run logs |
At the stop prompt, enter y to stop; other answers cancel. For non-interactive shutdown, use npm stop -- --yes. Restart does not ask for shutdown confirmation.
| Problem | What to do |
|---|---|
| Dashboard cannot start | Check Node.js version, run npm ci, confirm .env is in the project folder, and check that the configured port is free. View npm run logs. |
| TapData API connection fails | Check API URL/path, published service names, OAuth credentials, and network access. See API integration details. |
| Customer or transaction data is missing | Confirm DATA_SOURCE=tapdata and restart. Check API results, service mappings, and scan limits in the integration guide. |
| AI features do not work | Check provider, model, API root, key, and connectivity. Restart after configuration changes and inspect backend logs. |
For source database or simulation issues, use the Feeder documentation. For replication, transformation, or API publishing issues, use the BankStream Feeder repository.
| Repository | Purpose |
|---|---|
| BankStream Feeder | Source schema/sample-data scripts and commands to generate banking activity. |
| BankStream AI Dashboard | Customer 360 dashboard, transaction intelligence, and AI-assisted insights. |

