nimitFlowConnect data, machinery
and AI in a single system

When data is in different files, databases, and apps, reporting requires re-collecting and re-verifying accuracy. nimitFlow uses Flows as intermediary pathways to receive, transform, process, and transfer data, allowing teams to see steps and collaborate on work

Design the Flow by dragging nodes from the Toolbox and connecting them, define the source, processing steps, and destination. The presentation document calls this approach to creating jobs nimitGraph, while specific jobs use Python within nodes to define additional conditions

1. 1

   Receive data

   Select PostgreSQL, MySQL, HTTP/REST, Webhook, or files used in the work, and configure access settings

2. 2

   Set up operation

   Connect data nodes, transform values, calculate, loop, or call Python and AI in sequence

3. 3

   Test and set Trigger

   Run test, check results, then select to run yourself, set timing, or receive events from Webhook

4. 4

   Send results to the user team

   Create graphs, Excel/CSV reports, or send data and alerts through connected channels

• Python for specific conditions and AI Code Generator helps draft code for developers to review before use
• nimitFlow Drive stores files and divides owner, editor, and viewer-only permissions
• Export and import Flow as a .nim file to bring back work or continue development
• Systems or databases without ready-made nodes, connect further through Python or API according to supported endpoints

Use nimitFlow to route data from the existing SQL/NoSQL systems, APIs, files, and on-site devices into the central data lake. Separate the storage step from the processing step to select resources suitable for the volume of data and the frequency required.

• Select data intake jobs either in batch mode or streaming mode based on the readiness of the source
• Design S3-compatible file storage area and separate processing or search layers by project
• Define data retention periods for each layer, archiving, and recovery together with the system administrator

Implement governance alongside data flow by assigning Data Owners responsible for data and Data Stewards who manage meanings and quality. Apply organizational rules to Flows and integrated Catalog systems.

• Data Catalog / Metadata: Record descriptions, sources, stewards, and categories of datasets for discoverability
• Data Lineage: Show which steps data has passed through and where it was used in reports to analyze impact when changing the source
• Critical Data Elements: Identify key data and validation rules, such as blank values, data format, duplicates, and completeness
• Data Quality: Set validation rules and report anomalies to responsible parties according to the workflow
• Security: design access permissions, data masking, and key or secret management together with the systems selected for deployment

The data lifecycle covers creation, storage, use, publication, archiving, and destruction, with responsibilities and conditions defined for each stage in accordance with organizational policies

1. 1

   Handle data requests

   Record access requests, modifications, or management of data entering the organization's process

2. 2

   Search and verify

   Use Catalog data and connection channels to search for relevant items, along with verifying permissions and scope

3. 3

   Approve

   Send responsible parties to review and approve requests before proceeding with actions

4. 4

   Execute and record history

   Manage data according to approval results, record outcomes and evidence for retrospective review

Each device and app has specific duties. nimitFlow helps collect data for further use in reports or joint workflows by selecting signals, interfaces, and gateways appropriate to machinery

For vibration analysis work, it can be designed to receive data → process signal and FFT graph → analyze according to criteria or model → send results to the repair and maintenance team. Analysis using the model is additional work that must be prepared and tested separately from InCBM's threshold monitoring.

Provide prepared and inspected data as the starting point for AI work. Use Flow to sequence data preparation, call models, and send analysis results to users or systems.

• Time Series and Anomaly Detection: Create variables from time-series data, compare baseline values, and test anomaly detection.
• AI Vision: Receive inspection results from cameras processing real-time data, link them to production quantities, quality, and station history.
• LLM: Connect models through AI nodes or use Ollama for models running in-house.
• Dashboard and reports: Display trends, model results, and job statuses alongside user-accessible data.

Choose between Cloud Subscription or on-premises server installation (On-Premise) by evaluating the number of users, data sources, flow frequency, storage space, and responsible personnel

Enterprise architecture in a separate document defines areas for Edge, App, Data, and Execution to design permissions and resources for each section. The choice of Container approach or Cluster system must match the environment and version being installed

1. 1

   Review existing data and work

   Collect upstream data samples, report formats, users, and network limitations

2. 2

   Select pilot project

   Define one flow task with results and acceptance criteria agreed by the inspection team

3. 3

   Connect and test

   Test with real data, check errors, permissions, and operation when the upstream system stops

4. 4

   Handover and scale up

   Prepare manuals, responsible parties, backup and follow-up work before adding new data sources or models

Evaluate cost-effectiveness from data collection time, reduced duplication work, detected data quality, and organizational system maintenance costs. Do not use savings figures or accuracy from demonstration images as proof for all projects.