Schogini

Use the Schogini utility API for image editing and inspection, QR codes, text analysis, encoding, and link shortening. Agents close small gaps in a workflow behind one credential.

34 actions API key available

Schogini website ↗ Platform Documentation ↗ Capability data verified 2026-08-01

A customer submits a product review with a photograph

Agent normalises the photograph with Create Image Thumbnail and Convert Image Format

Agent reads Get Image Metadata to see what the file carries beyond the picture itself

Agent runs Detect Spam or Phishing and Moderate Content across the review text

Agent runs Detect Language and Analyse Sentiment so the item routes to a team that can read it

Agent posts every flagged submission to the community channel with the customer's original wording intact

A moderator publishes, edits, or suppresses each flagged review, and the agent applies the decision

What This Integration Enables

Schogini is the connector for user-generated content, and it earns that description because it covers both halves of a submission. On the image side there are twelve editing operations and four inspection ones: resize, crop, rotate, blur, grayscale, thumbnail, border, watermark, text overlay, compositing, format conversion, SVG rasterisation, plus metadata reading and conversion to and from inline data. On the text side there are eight: sentiment, moderation, spam and phishing detection, language detection, translation, contact extraction, configurable entity extraction, and YouTube listing metadata. QR generation and reading, hashing, Base64, page fetching, and a link shortener fill in around them. Image operations take a public image URL rather than a file upload.

The screening actions are the reason to reach for it, and they are also the reason to be careful with it. Moderate Content and Detect Spam or Phishing return a judgment about something a real person wrote. Schogini's own OpenAPI document declares no response schemas, which means field names should be confirmed against a live call, and no enums, which means format, colour, and language parameters take free text rather than a validated list. Those are the vendor's stated limits, and they are worth knowing before a flow is built on top of them. FlowRunner's answer is not to treat the verdict as a decision. It is to treat the verdict as evidence, hand it to a moderator, and let the moderator decide. An agent that suppresses a customer's words on a thin signal has not saved anyone work; it has moved a mistake somewhere nobody is looking.

Without FlowRunner

Two pipelines for one submission The image goes through one tool, the text through another, and neither knows the other ran

Assets published as uploaded Whatever resolution and format the customer's phone produced is what appears on the page

Moderation happens after publication Objectionable content is removed once somebody reports it

With FlowRunner

One pass over the whole submission Image normalisation and text screening run as steps in the same flow, on the same record

Assets normalised on arrival Every published image is a consistent size and format regardless of what was uploaded

Screening before publication Flagged submissions are held as queued decisions rather than appearing and being pulled down

Use Case Scenarios

Review intake with the image and the text handled together

A customer posts a review with a photograph. The agent calls Get Image Info to confirm the file meets minimum dimensions and Get Image Metadata to read the embedded EXIF fields, which is where camera settings, capture dates, and location tags live. It calls Create Image Thumbnail to bound both dimensions inside the listing box, and Convert Image Format to bring everything to one delivery format. On the text it runs Detect Language so a review written in French does not sit in an English queue, Analyse Sentiment to prioritise the angry ones, and Moderate Content and Detect Spam or Phishing to catch what should not go live. Clean reviews publish. Flagged ones queue.

Catalogue photography prepared and branded in one pass

Supplier photography arrives at wildly different sizes with no consistent treatment. For each asset the agent calls Resize Image to fix a width with the aspect ratio preserved, Add Border to Image to apply the brand frame, and Add Watermark to Image with position and opacity given as percentages so the same settings land identically on images of different sizes. Convert SVG to Image rasterises any vector logos that a downstream email client or social platform cannot render. Convert Image URL to Data URI turns the finished asset into inline data for a self-contained email that renders without external requests. The catalogue and the campaign use the same assets, prepared once.

Inbound enquiry triage before a ticket is created

Contact form submissions arrive faster than anyone reads them. The agent runs Detect Spam or Phishing to filter the credential-harvesting and fake-invoice traffic before it becomes a ticket in Zendesk. Extract Contact Details then pulls names, email addresses, and phone numbers out of the message body and the signature block, and Extract Entities picks up the entity types the business actually cares about, such as product line, competitor, or contract reference. Analyse Sentiment sets the queue priority. What reaches a support agent is a structured record rather than a paragraph, and what does not reach them is the traffic that was never a customer.

Human-in-Loop Highlight

Moderate Content is where this connector stops and asks, and the reason is asymmetry. A missed piece of objectionable content is visible, embarrassing, and fixable in minutes. A suppressed review from a real, unhappy, paying customer is invisible: they see their words never appear, they do not hear why, and the company never learns what they said. Only one of those failures gets reported. Schogini returns a judgment on the text, and its own documentation notes that the API publishes no response schemas, so the verdict is a signal rather than a finding. So the agent does not delete anything. It holds the submission unpublished, keeps the customer's original wording untouched, and posts to the community channel: "Review from order 88214 was flagged by Moderate Content. Full text attached, sentiment negative, language English. Publish, publish with the flagged passage edited, or suppress and open a support ticket?" A moderator answers. If the answer is suppress, the agent opens the ticket, because a review bad enough to hide is a customer worth calling. That is human-in-the-loop used the way it should be on a moderation queue: the classifier narrows the pile, and a person decides what happens to a customer's words.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

34 actions

Image Editing

Resize Image Resizes an image fetched from a public URL to a given pixel width, preserving aspect ratio, and returns it in the requested output format. The format is free text because the API publishes no fixed list, so verify unusual formats against your account first.
Crop Image Crops to a rectangle defined by two corner points, with all four coordinates given as percentages between 0 and 100 rather than pixels, so the same crop works across images of different sizes.
Rotate Image Rotates by a given number of degrees between 0 and 360. Used to correct orientation on scans and user-uploaded photographs before publication.
Blur Image Applies a Gaussian-style blur at a given radius. Used to obscure sensitive detail in screenshots or to generate soft background variants of a product photograph.
Convert Image to Grayscale Converts a colour image to grayscale. Used for monochrome print variants, for reducing visual noise in document processing, and for normalising images before optical analysis.
Create Image Thumbnail Generates a thumbnail that fits inside a maximum width and height while preserving aspect ratio. Unlike Resize Image, which fixes the width exactly, this bounds both dimensions so the result never exceeds the box.
Add Border to Image Draws a solid border of a given thickness and colour. The colour is free text with no published list, so plain names or a hex value are the safest inputs to test first.
Add Watermark to Image Overlays a watermark image at a chosen position and opacity, both given as percentages, so the same settings place the mark consistently regardless of image size.
Add Text to Image Renders a text message onto an image, optionally with a topic heading and a channel label. Only the message is required, so it can also generate a plain text card when no source image is supplied.
Combine Two Images Overlays a second image onto a first at a position and size expressed as percentages of the base, producing a single composite. Used for logo placement and simple social cards.
Convert Image Format Converts an image from one format to another, for example PNG to JPEG for size or JPEG to PNG for transparency. The target format is free text, so confirm unusual formats before relying on them.
Convert SVG to Image Reads an SVG from a URL and rasterises it into a bitmap. Used when an email client or social platform cannot display vector graphics.

Image Information

Get Image Info Returns basic properties such as dimensions and format. Used to validate that an upload meets size requirements before publishing, or to branch a flow on image shape.
Get Image Metadata Reads embedded metadata such as EXIF fields recorded by cameras and phones. Used to audit capture dates and camera settings, and to spot assets carrying metadata you would rather not publish.
Convert Image URL to Data URI Fetches an image and returns it as inline data so it can be embedded directly in HTML or JSON. Used for self-contained emails and documents that must render without external requests.
Convert Data URI to Image URL Takes inline image data, stores it, and returns a hosted reference. The inverse of the above, used when a prior step produced inline data that a downstream system needs as a plain URL.

QR Codes

Generate QR Code Generates a QR code image encoding supplied text, which can be a URL, contact details, Wi-Fi credentials, or any short string. Used for tickets, invoices, packaging labels, and event badges produced inside a flow.
Read QR Code Decodes a QR code from an image at a given URL and returns the text it contains. Used to process scanned documents, inbound label photographs, and customer-submitted ticket images.

Text Analysis

Analyse Sentiment Assesses whether a block of text reads as positive, negative, or neutral. Used to triage support tickets, prioritise angry reviews, and route survey responses.
Moderate Content Checks text for material that may be unsuitable for publication, such as abusive, explicit, or otherwise objectionable language. Used to screen comments, reviews, and form submissions before they reach a public channel.
Detect Spam or Phishing Evaluates text for signs of spam or phishing such as deceptive urgency, credential harvesting, or suspicious offers. Used to filter inbound contact-form submissions before they create tickets.
Detect Language Identifies the language a passage is written in. Used to route multilingual enquiries and to decide whether a translation step is needed later in the flow.
Translate Text Translates text into a target language. The target is free text, so a plain language name is the safest input because the API publishes no fixed list of supported languages or codes.
Extract Contact Details Pulls names, email addresses, and phone numbers out of unstructured text. Used to turn email signatures, enquiry messages, and scanned business card text into structured records.
Extract Entities Extracts named entities of your choosing, with the entity types supplied per call. That makes it adaptable to domain-specific extraction without training a custom model.
Generate YouTube Metadata Produces a title, description, and tags from a summary of a video, steered towards chosen keywords. Used in publishing pipelines that need consistent listing copy.

Encoding and Utilities

Hash String Computes a cryptographic hash of a string. Used for stable deduplication keys, for comparing values without storing them in the clear, and for idempotency identifiers.
Encode Base64 Encodes a string as Base64. Used to prepare values for systems expecting encoded payloads.
Decode Base64 Decodes a Base64 string back to plain text. Used to read encoded fields arriving from webhooks, queues, and third-party systems.
Fetch Web Page Retrieves the contents of a page through Schogini's servers. Used to pull reference content into a flow or to confirm a page is reachable without hosting a fetcher.
Pick Random URL Selects one URL at random from a supplied list. Used to rotate creative assets or spread traffic across mirrors.
Create Short URL Shortens a long URL and returns the shortened form. Used to trim links before they go into SMS, printed material, or a QR code generated elsewhere in the same flow.
Resolve Short URL Looks up a short code and returns the original URL. Returns a not-found error when the code does not exist. Used to expand links captured from messages or scanned codes before acting on them.
Check Credentials Verifies that the configured API key is accepted. Used as a connection health check before a live run. A rejected key raises an authentication error rather than returning a negative result.

Frequently Asked Questions

What can FlowRunner do with Schogini?

FlowRunner agents can run Resize Image, Crop Image, and Rotate Image in Schogini, plus 31 more actions.

Does connecting Schogini to FlowRunner require OAuth?

No. Schogini connects to FlowRunner with an API key, no OAuth flow required.

Can Schogini trigger a FlowRunner workflow automatically?

Schogini doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with Schogini

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us