AI & automation.
Built into the work.

Inside the project

• AI integration: optional Ollama narratives explain validated report aggregates, with deterministic wording when the model is unavailable.
• Resolve the saved layout version and its configured template datasets as the question scope.
• Interpret measures, dimensions, periods, aggregations, and requested output forms.
• Support cards, tables, matrices, bars, lines, areas, composition charts, heatmaps, and scatter requests.
• Offer compact choices when the question explicitly asks for several visual forms.
• Explain when a requested subject belongs to a different dashboard version.
• Accept browser speech-recognition input through the same version-bound question path.
• Retain a deterministic narrative when optional model wording is unavailable.

The toolkit

Inside the project

• AI integration: the version-aware assistant supports executive reports 213–217 and can use Ollama to explain their computed facts.
• Provide separate versions for E-Bill, A/R, Disconnects/Reconnects/Bankruptcies, Final Bill Recovery, and Customer Payments.
• Reuse each report's existing configured SQL or reporting source.
• Export KPI summaries and comparisons alongside the underlying chart-data tables.
• Include rendered graph images and separate data-table worksheets in Excel exports.
• Export the full report version to PNG, including its KPI cards, graphs, table previews, and notes.
• Support per-report email/export endpoints and a SQL Agent PowerShell delivery script.
• Retry scheduled requests and validate exported XLSX/PNG signatures before accepting the output.

The toolkit

Inside the project

• Gemma integration: Gemma through Ollama explains code-calculated load adjustments; the source defaults to gemma3:4b when no model is configured.
• Read the operating workbook and current REG forecast as the report inputs.
• Apply the configured load-ramp adjustments to calculated daily peaks.
• Supply the calculated forecast facts to the narrative-generation step.
• Restrict report generation to the configured trigger user.
• Track generation status and prevent duplicate concurrent generation for the same date.
• Reuse an already generated daily report from the cache.
• Skip Gemma generation when no current forecast dates have available operating-log ramp values.

The toolkit

Inside the project

• AI integration: Ollama supports question comprehension and structured query planning; catalog checks and server-side compilation validate execution.
• Search the surfaced database catalog and inspect object fields.
• Explore the relationship neighborhood around an object.
• Bind live querying to the currently selected, query-ready semantic source.
• Use local question comprehension with a literal fallback when that pass fails.
• Check proposed objects, columns, joins, measures, and filters against the catalog.
• Compile the validated plan into parameterized SQL on the server.

The toolkit

Inside the project

• AI integration: the configured Ollama model supports task intelligence, with optional prior-conversation analysis for newly received replies.
• Enroll each employee separately through one-click Outlook intelligence activation.
• Start new-mail ingestion at that employee's activation timestamp, excluding older messages and old unread mail.
• Use a workstation Classic desktop mail connector in an interactive signed-in user session.
• Offer bounded prior-conversation context for newly received replies without importing it as new tasks.
• Show future calendar events using each employee's configured preferences.
• Expose connector/provider health with pause, resume, and disconnect controls.
• Remember System, Light, or Dark theme preferences in the browser.

The toolkit

Inside the project

• Show ticket-volume measures as compact summary cards.
• Present service-level performance alongside demand measures.
• Break down tickets by status.
• Compare ticket priority composition.
• Include security-training and phishing components.
• Place service-uptime components in the same payload-driven report interface.

The toolkit

Inside the project

• AI integration: every dashboard layout has a bot bound to its datasets; optional Ollama wording explains computed results within that layout context.
• Show CX measures including call volume, E-Bill adoption, KA meetings, KA engagements, KA projects, and dialing volume; layout 186 provides one six-KPI composition.
• Present cumulative call-handling counts at 30, 60, and 120 seconds.
• Include call volume in the handling table alongside the corrected abandoned count.
• Use Contact Center nOutboundAttempted for the configured outbound campaign population.
• Bind custom HTML visuals to the existing dashboard configuration and SQL-backed sources.
• Provide a bot on every layout, bound to that layout's datasets for contextual questions and explanations.

The toolkit

Inside the project

• Filter transactions by serial number, status, location, selected date field, and date range.
• Search work-order identifiers, warehouse fields, and transaction notes.
• Inspect transaction details with related inventory and transformer-part lookups.
• Create transactions from a prior transaction using the copy workflow.
• Restrict transaction creation and editing through the transaction-administration policy.
• Require an existing real inventory serial number before creating lifecycle transactions.
• Reject new lifecycle transactions after Scrap and avoid saving an unchanged transaction state.
• Apply transaction validation and synchronize the equipment's inventory state after a successful save.

The toolkit

Inside the project

• Create and edit service-layout records, including uploaded PDF content.
• Keep location, customer, service-area, and AutoCAD-link context with the layout.
• Normalize address components when finding Customer Information System service-order candidates.
• Require the saved layout and valid issued-date context for candidate lookup.
• Allow the reader-specific save path to update the service-order number without applying other posted fields.
• Resolve the service-order completion date during the relevant save path and check whether completion email is needed.
• Provide a print view and PDF retrieval for the saved layout.
• Email the layout with a composed PDF attachment when available, including recipient validation.

The toolkit

Inside the project

• Present technician performance across a 13-month horizon.
• Use interactive SVG chart elements for exploring the report.
• Open supporting request-level drilldown tables.
• Inspect individual requests in detail dialogs.
• Switch between light and dark visual palettes.
• Adapt chart and detail arrangements to smaller screens.

The toolkit

Inside the project

• Read Carrier Mobility invoices delivered to the file-share intake.
• Try native PDF text extraction before OCR.
• Use Tesseract OCR for the required invoice regions when native extraction is insufficient.
• Reconcile financial columns and department totals to the cent.
• Hold ambiguous invoices for review before production processing.
• Separate dry-run validation from production execution.
• Import validated results to SQL and produce SMTP notifications through the pipeline.

The toolkit

Inside the project

• Read call analytics from the Contact Center Cloud Analytics API.
• Populate the existing SQL call-volume reporting shapes.
• Apply the specified answered-call and abandoned-call business rules.
• Write call-handling and response-time measures for dashboard use.
• Include outbound dialing measures in the CX integration.
• Retain the rolling 13-month response-time series.

The toolkit

Inside the project

• Resolve a saved workbook layout by its title.
• Resolve a layout by its layout-version identifier.
• Render dashboard PNG snapshots using Playwright.
• Configure image width and scale for snapshot generation.
• Configure the render-wait time before capturing the layout.
• Build Excel workbooks from the SQL-backed tile sources configured in that layout.

The toolkit

Inside the project

• List the available historical snapshot dates.
• Return filter options for the usage exploration interface.
• Scope the result by billing cycle and year.
• Filter the population by utility type and sector.
• Narrow the result by postal prefix.
• Inspect account, customer, service-address, and billed-consumption fields within the selected population.

The toolkit

Inside the project

• Use the aging-cube source for authoritative reporting balances.
• Use that source for account inclusion and aging qualification.
• Keep active versus final status consistent with the reporting source.
• Enrich missing customer information from Customer Information System.
• Add due-date, payment, and deposit context where available through enrichment.
• Apply the shared source rule to both dashboard and Excel-export loaders.

The toolkit

Inside the project

• Read service-layout records and the relevant Customer Information System account/order candidates.
• Normalize addresses before comparing candidate records.
• Evaluate completed and created dates through the defined date-distance rules.
• Produce the default Excel review without writing source changes.
• Support configurable updates restricted to blank destination values.
• Expose overwrite behavior as a separate configured choice.

The toolkit

Inside the project

• Accept the billing files delivered by the existing cloud workflow platform flow.
• Use a file-share folder as the processing intake.
• Read the delivered Excel billing workbooks.
• Run the existing SQL billing-processing workflow.
• Produce SMTP reporting as part of the pipeline.
• Operate from the delivered files without requiring Outlook or browser mailbox access.

The toolkit

Inside the project

• Select the active Operations user population.
• Restrict enrollment selection to the current calendar month.
• Use completion dates when identifying unfinished enrollments.
• Use enrollment status alongside completion information.
• Keep each unfinished enrollment as a separate report row.
• Generate and email the report through the script.
• Run a no-email validation mode before sending a report.

The toolkit

Inside the project

• Apply dashboard titles from the supplied payload.
• Build the dashboard's tab structure from its payload contract.
• Populate measures and detail rows from incoming data.
• Render the retained trend, donut, bar, and table visual families.
• Refresh a selected visual family through events.
• Refresh the full dashboard through its update events.
• Omit unsupported visual families when the payload does not provide the required information.

The toolkit

Inside the project

• Inspect standard phishing campaign paths.
• Inspect callback campaign and test paths.
• Explore both global and campaign-specific security tests.
• Check regional API endpoints while locating the account data.
• Fetch recipient records for tests in the configured target month.
• Write read-only diagnostic output to a timestamped folder.

The toolkit

Inside the project

• Download the workbook template for the enabled CX visuals.
• Accept XLSX files through the upload workflow.
• Validate the incoming file type.
• Check expected sheets and workbook headers.
• Parse populated visual tabs into the configured measure shapes.
• Ignore empty visual tabs instead of treating them as populated data.
• Return actionable validation feedback for workbook corrections.

The toolkit

Inside the project

• Treat existing SQL table definitions as the destination contract.
• Match Oracle objects by name with case-insensitive comparison.
• Match source and destination columns with optional explicit overrides.
• Stage the incoming rows using destination-compatible SQL types.
• Use a usable key when comparing staged and destination rows.
• Insert new rows and update rows whose values changed.
• Preserve target-only rows and avoid replacing or clearing destination tables.

The toolkit

Inside the project

• ML integration: baseline and machine-learning candidates are compared through temporal backtests and conservative champion/challenger decisions.
• Read the actual-load workbook and convert observations to a compatible hourly target.
• Run expanding-window, rolling-origin 14-day backtests.
• Keep future actual observations out of the feature set.
• Compare baseline and machine-learning candidates under the same test design.
• Archive vendor REG/DNN forecast workbooks and score them as actual observations become available.
• Maintain a statistically conservative champion/challenger decision.
• Keep the retention-harmonic inspired challenger distinct from the user's optional exact S2 adapter.

The toolkit

Inside the project

• ML integration: a TF-IDF and logistic-regression classifier combines with structured rules to suggest roles and retain uncertain cases for review.
• Combine structured ticket fields with operation-pattern evidence.
• Train the text classifier from sufficiently supported seed examples.
• Require adequate classifier probability and margin before accepting an inferred operation.
• Weight recent support more heavily than legacy activity.
• Separate classification evidence from ownership confidence.
• Produce role catalogs and proposed routing-rule worksheets.
• Retain low-confidence or conflicting cases in exception review with methodology documentation.

The toolkit