You don’t need to build any of this by hand. If you see something here that matches your use case, or you already know you want something this specific, just describe it to Prowork. It will build the flow for you. This page exists to spark ideas, not to be a how-to manual. Each section covers one higher-order capability. Within each section, the examples are scannable. Pick the ones that sound like your situation.Documentation Index
Fetch the complete documentation index at: https://parabola.io/docs/llms.txt
Use this file to discover all available pages before exploring further.
Generating notifications and alerts
Parabola can produce polished, formatted output, not just plain text. The recipient gets a message that’s tailored and easy to act on.- Branded HTML emails with KPI cards, section headers, subtotals, and styled tables. Used for executive reports, CFO summaries, weekly digests.
- Per-recipient email fan-out where each person gets only the data relevant to them. One manager gets their team’s open approvals, one vendor gets their open POs, one store manager gets their inventory issues.
- Slack messages with rich formatting: status emoji, formatted dollar amounts, clickable links, bulleted summaries grouped by category.
- Tiered notifications that escalate based on age. A friendly nudge at 7 days, a more direct ask at 30 days, a manager review at 45+.
- Conditional message routing with different subject lines, recipients, or templates based on the data (e.g. executives routed through their assistant, internal vs. external recipients).
- Per-recipient attachments. One email with one CSV attachment scoped to that recipient’s data.
- Mailto links inside reports that make it one-click to reply or forward.
- Inline product images in retail merchandising emails.
- Alert digests that only fire when there’s something worth surfacing, so the recipient isn’t trained to ignore them.
Monitoring changes and detecting state
Parabola can remember what it saw last time and compare against this time. Most teams use this to surface meaningful changes instead of dumping a full snapshot every run.- Inventory change detection. Only alert when a SKU’s status changes, not on every run.
- ETA change tracking. Surface only the orders where the ETA shifted since the last run.
- New-record alerts. Anti-join the current data against a “sent log” so each new record is only flagged once.
- Watermark-based filtering. Only consider records newer than the latest one already processed, even if the upstream source replays old data.
- Snapshot diffing. Keep dated snapshots of a source and compute what changed between any two of them.
- Rate-of-change monitoring. Flag when a metric spikes or drops past a threshold.
- Schema drift detection. Alert when a source file’s columns change unexpectedly.
- Audit trails. Append a row to a log table on every run so changes are reviewable later.
Reading and parsing data formats
Parabola handles most common business data formats, including ones that aren’t tabular.- CSV and TSV with quoting, embedded newlines, mixed encodings.
- Excel files with multi-row headers, merged cells, named ranges, and multiple sheets.
- JSON and nested JSON. Flatten, explode, or unpack nested objects and arrays.
- XML from APIs and legacy systems.
- PDF text extraction, both structured (forms, invoices) and free-text.
- PDF table extraction from scanned or native PDFs.
- EDI X12 documents (850 purchase orders, 856 ASNs, 810 invoices) with hierarchical Ship → Order → Pack → Item structures.
- Inbound email parsing. Extract structured fields from the subject, body, or attachments.
- HTML scraping from emailed reports or web pages.
- Recursive tree structures like QuickBooks GL reports, flattened into tabular form with breadcrumb columns.
- Multi-line cells. Parse cells that contain newline-delimited key=value pairs into separate columns.
- Image OCR. Extract text from scanned receipts, invoices, and packing labels.
Producing files and documents
Parabola can produce output in whatever shape the next system or person needs.- CSV and Excel files with formatting, multiple tabs, and formulas preserved.
- PDF reports generated from templates.
- JSON documents for downstream systems or APIs.
- HTML pages and dashboards that can be hosted or embedded.
- Per-record file generation. One statement per customer, one packing slip per order, one report per vendor.
- Compact encoded outputs like page-range strings (“1-3, 5, 7-9”) for downstream scripts.
- Multi-sheet Excel workbooks with one tab per category, region, or time period.
- Files written to Google Drive, S3, SharePoint, Dropbox with timestamped filenames.
Connecting systems that don’t talk to each other
Parabola is often the bridge between tools that lack a built-in integration.- CRM to accounting sync (e.g. HubSpot to NetSuite, Salesforce to QuickBooks).
- E-commerce to ERP sync for products, inventory, orders, and fulfillments.
- WMS to accounting for inventory adjustments and COGS.
- Spreadsheet to API. Read from a Google Sheet, push to any REST API.
- Email-only report ingestion. Turn a vendor’s emailed CSV into a live data feed.
- Webhook receivers that catch inbound events and route them to the right system.
- Bi-directional sync with diff detection so changes flow both ways without overwriting.
- System-of-record promotion. Pick one system as the source of truth and force the others to match.
Reshaping and pivoting data
Beyond simple joins, Parabola handles the reshaping work that usually requires a SQL query or a Python script.- Pivot a column into rows or rows into columns (long-to-wide and back).
- Multi-dimensional pivots with a column dimension and a row dimension, with values in the cells.
- Cross-tabs with marginal sums. A total column and a total row in the same output.
- Stacking files with different schemas into one normalized table.
- Explode array cells into one row per array element.
- Forward-fill and back-fill to handle merged cells and partial data.
- Window functions. Running totals, rolling averages, period-over-period change.
- Hierarchical aggregation with totals at multiple levels.
Calculating, scoring, and ranking
Parabola does numerical work too, not just data movement.- Allocation math. Split a pool across recipients proportionally with penny-adjustments so totals tie out.
- Aging buckets. Group records into 0-30, 31-60, 61-90, 90+ day cohorts.
- Scorecards with weighted inputs across multiple dimensions.
- Currency conversion with daily FX rates.
- Tax calculation by jurisdiction.
- Forecasting with backtest accuracy reporting.
- Variance analysis. Actual vs. budget, this period vs. last, etc.
- Ranking and top-N within groups.
- Penny-perfect reconciliation with tie-out checks.
Categorizing, classifying, and matching
Parabola can categorize records using rules, history, or AI.- Rules-based categorization with editable mapping tables maintained inside the flow.
- History-based learning. Apply the most common past categorization to new records, with confidence scores.
- AI categorization for records that don’t match any rule.
- Hybrid cascade. Use history first, AI second, “Uncategorized” last, with originals preserved.
- Fuzzy matching for merchant names, vendor strings, product titles.
- Lookup with fallback. Try exact match, then prefix match, then a default.
- Tariff and HS code classification using AI plus official reference data.
- Sentiment and topic tagging for customer feedback or support tickets.
Building human-in-the-loop workflows
Not all decisions can be automated. Parabola can route the unclear cases to people while handling the rest.- Review queues for low-confidence categorizations.
- Approval routing with escalation rules.
- Exception reports that surface what needs human attention.
- Editable in-flow lookup tables that ops leads can update without touching the flow.
- Test/preview routing that sends to a single review inbox during development before flipping to live.
- Audit columns that preserve the original value next to the override so spot-checking is trivial.
- Confidence flagging on AI-generated values.
Working with AI as a step
When teams use AI inside a flow, they treat it as one more transform step, not as a chatbot.- Extraction of structured fields from unstructured text (emails, PDFs, transcripts).
- Classification and tagging of records that don’t fit clean rules.
- Summarization of long-form content into a few bullets.
- Generating structured prompts for downstream creative or generation pipelines.
- AI-vs-rules validation. Run both in parallel and surface where they disagree.
- Batching for context windows. Split a large dataset into chunks sized to fit the model.
- Multi-step LLM chaining where one model’s output feeds the next.
- Vector embedding generation for semantic search applications.
Calling APIs and webhooks
Parabola is a competent API client. Each row in a table can drive one API call.- One row per outbound request. The cell holds the payload, the row identifies the target.
- GraphQL mutations for Shopify, Linear, and similar APIs.
- REST API calls with custom headers, auth, and request bodies.
- OAuth token rotation in-flow. Flag when a token is expiring, fetch a new one, write back to config.
- Batched requests in chunks of N (100 IDs per Spotify call, etc.).
- Paginated reads with cursor or offset handling.
- Rate-limit-aware retry with backoff.
- Webhook receivers for inbound events from external systems.
- Error accumulation across a multi-step API pipeline, surfaced in one user-facing message.
Scheduling and triggering
Flows can run on a schedule, on a trigger, or on demand.- Time-based schedules. Every hour, daily at 8am, every Monday morning.
- Webhook triggers that fire when an external system POSTs to a URL.
- Email triggers that fire when an email arrives at a specific address.
- Manual runs for one-off analysis.
- Timezone-aware scheduling with explicit IANA timezone support.
- Conditional skipping. Flow runs but exits early if there’s nothing to process.
- Sequencing barriers that force step A to complete before step B begins.
Building durable systems
For flows that need to behave responsibly over time, not just produce a single output.- Sent-logs to ensure each record is only acted on once.
- Snapshot tables for historical state.
- Config rows stored in a sheet, editable without touching the flow.
- Graceful degradation when an upstream source is down or empty.
- Fail-the-run-after-alerting patterns where the flow notifies first, then marks itself failed for monitoring.
- Per-row error accumulation so multi-step failures surface together, not one at a time.
- Backfill modes that combine one-time historical estimates with ongoing daily deltas.
- Version cutover handling when the data model changes mid-stream.