Queue Storage (QQS) is powerful, but only if you can manage it confidently.

Operational teams need to answer:

• Which queue storages exist for this subscription?
• What are their durations and schedules?
• Are items piling up?
• Are batches being released as expected?
• Can we adjust thresholds without breaking pipelines?

Bringing QQS management into the Portal UI turns buffering from “a clever config” into “an operational tool”.

That matters a lot when you’re running high-volume offer or catalog syncs across multiple channels.

Credentials are not pipeline logic.

They’re security artifacts.

So the platform experience is moving toward a cleaner separation:

• pipeline definitions reference credentials by ID
• credentials are managed centrally (create/update/rotate)
• permissions control who can access or modify what

When credential management is visible and structured, teams can:

• rotate keys without redeploying pipelines
• separate agency responsibilities from merchant ownership
• reduce “who still has the production token?” moments (which should not exist)

As pipelines grow, they stop being linear.

They become decision trees:

• filter out invalid objects
• route based on marketplace/storefront
• switch-case based on complex conditions
• merge parallel branches
• loop through collections when needed
• enrich objects with rule-based actions

Qilin.Cloud has been building these primitives for a while.

February’s focus has been making them easier to configure and operate through the Portal UI, so teams can:

• build complex routing without hand-editing JSON
• review logic visually
• reduce configuration mistakes
• onboard new team members faster

This is how a platform stays usable as it gets more powerful.

Operational UI work might sound “non-technical”, but it’s actually developer leverage:

• fewer support tickets
• fewer manual interventions
• fewer misconfigurations
• faster iteration cycles

You spend less time being the “human control plane”.

• Merchants: more transparency and safer operation of pipelines in production.
• Agencies: easier delivery and maintenance; clearer boundaries of responsibility.
• Investors: this is maturity. Operational features reduce costs and unlock bigger customers with higher requirements.

The story from the past 18 months has been consistent:

• build strong pipeline primitives (filtering, enrichment, merge, buffering)
• make them observable (tracking and status)
• make them operable (UI control and safe recovery)

That’s how platforms earn trust.

Complex commerce workflows will always have complexity.

The goal isn’t to pretend it’s simple.

The goal is to make complexity manageable – predictable, observable, and controllable.

That’s what February’s work is about. And it’s where Qilin.Cloud is heading.

Testing Mode lets you run pipeline logic without affecting actual pipeline execution.

That means you can validate:

• routing
• filtering
• enrichment
• merge behavior
• connector payload structure

…without accidentally pushing real updates to a marketplace or polluting production logs with “test objects”.

The results are still observable (tracked like normal executions), but they’re flagged as test runs.

Pinned Data is a mechanism that replaces the live output of a processor with fixed mock data.

In normal runs:

• a processor does its real job (HTTP call, connector sync, etc.)
• output depends on external systems, network, timing

In test runs with pinned data:

• the processor’s real work is skipped
• the output becomes deterministic
• downstream steps can be tested with certainty

This is the classic integration engineer’s dream:

> “Let me test the next steps, even if the external API is flaky today.”

1) Test the full pipeline end-to-end (with pinned entry data)

You provide pinned data for the entry processor, simulating the trigger object.

Result: the pipeline runs end-to-end on a controlled input, producing deterministic outputs.

2) Test a single processor using real historical context

You provide a previous Pipeline Execution ID.

Qilin loads the execution context from that run and re-executes only the processor you’re testing.

This is perfect when:

• you changed an enrichment rule
• you adjusted a filter predicate
• you want to replay one step against real production-like data

3) Test a processor in isolation (no historical context needed)

You provide pinned input directly for the processor.

Result: fast, isolated, great for validating configs and behavior without waiting for upstream steps.

You want to verify enrichment logic that depends on a pricing service.

But the pricing API is unstable today.

Pinned data lets you fake the pricing output so you can test:

• calculations
• mapping
• downstream connector payloads

No external dependency. No waiting. No chaos.

Agencies live and die by delivery speed and reliability.

Testing Mode enables a cleaner workflow:

• build pipeline logic
• test with pinned scenarios
• validate edge cases
• only then turn on live execution

That’s how you reduce go-live risk without multiplying custom tooling.

• deterministic tests for pipeline logic
• faster iteration loops
• easier debugging (“same input, same output”)
• safer experimentation with advanced processors (merge, switch-case, enrichment, http calls)

• fewer production incidents caused by “untested config changes”
• faster onboarding because pipelines can be validated before real data flows
• stronger platform trust: changes are intentional and verifiable

February will continue the “operational control” theme:

• expanded UI management for queue storage and credentials
• more advanced routing and processor configuration in the portal

Because once pipelines grow, operations needs knobs – not prayers.

The traditional way of testing pipelines is stressful because it mixes two activities:

• validating logic
• operating production

Testing Mode separates them.

And honestly? That’s long overdue.

In marketplaces, “create” is risky.

It’s easy to create the wrong offer with:

• the wrong condition
• the wrong storefront
• missing compliance fields
• the wrong EAN mapping
• inconsistent identifiers

Those mistakes can be expensive.

So update-only sync is the disciplined, traditional approach:

> First, prove you can update reliably. Then expand scope.

In Kaufland, an offer (unit) is uniquely identified by the combination:

**EAN + Condition + Storefront**

That means the connector looks at these three pieces of information to locate the existing unit.

 Behavior

• If a matching unit is found → update price/stock/handling time
• If no matching unit is found → mark the sync as failed with:
  • `Cannot find existing offer`

This is the correct kind of strictness. It prevents “phantom offers” from appearing because a mapping was wrong.

You can export offers using either:

• `OutputConnector` (single item)
• `OutputConnectorByBatch` (bulk)

Batch mode supports up to 150 offers per batch, which is ideal for real-world offer volume and helps reduce request overhead.

A high-level flow looks like:

1. build a pipeline that produces Offer objects
2. add an output processor (single or batch)
3. configure a `targetId` that represents the Kaufland storefront destination
4. onboard the channel using that same `targetId` so the connector knows where to send updates

This keeps the pipeline definition and the channel onboarding aligned – no hidden mapping magic.

• Always include valid EAN and Condition on offers. They’re mandatory for correct unit matching.
• Treat `targetId` as a stable contract. If it changes, your connector routing changes.
• Prefer batch export for high-volume updates. It’s cheaper, faster, and friendlier to APIs.
• Use Data Flow Tracking to monitor failures and spot “offer not found” patterns early – those usually indicate mapping issues.

Developers

• clear identification logic (less ambiguity)
• safer connector behavior (less risk during rollout)
• batch export for operational throughput

Merchants

• safer price/stock updates without accidental new offers
• predictable behavior under load
• cleaner error signals when something isn’t mapped correctly

Agencies

• easier rollout strategy: start with updates, validate matching logic, then scale
• fewer marketplace “oops moments” during go-live

Investors

Connector depth expands addressable value. Update-only is a reliable first step toward full lifecycle sync while keeping risk low.

January will focus on the developer and operations experience again:

pipeline testing mode and expanded UI capabilities – so teams can validate flows safely before turning them loose on production channels.

Fast and reckless breaks things.

Slow and deliberate builds trust.

Qilin.Cloud is taking the deliberate path – because that’s how integrations survive in the real world.

Traditional integration recovery often looks like:

• re-run the whole job
• hope duplicates don’t happen
• manually reconcile partial updates
• dig through logs to guess what happened

It’s expensive, risky, and it doesn’t scale as you add more pipelines.

So Qilin.Cloud is moving toward a cleaner model:

> Treat executions as trackable artifacts you can inspect, classify, and retry.

Not every error deserves the same response.

Sometimes:

• a product is missing a non-critical field → warn and continue
• one optional enrichment service times out → warn and continue
• the output connector rejects the object → fail the object (or the pipeline) depending on policy
• a validation error occurs → stop, because continuing would produce bad data

So processors can be configured with settings like:

• continue on error (do we proceed downstream?)
• custom error status (how should this failure be classified?)

This allows a pipeline to finish with nuance:

• Completed (all good)
• Completed with warnings (action required, but business kept moving)
• Failed (hard stop)

That is exactly how experienced operations teams think.

Sometimes failure isn’t caused by your data or your pipeline logic.

Sometimes it’s just the world:

• an external API is down
• a token expired
• a marketplace has a temporary outage
• a partner system returns 500 for 20 minutes and then “recovers”

In those cases, the right response is often:

retry the execution once the dependency is healthy again.

Qilin.Cloud now supports manual retry of:

• a pipeline execution
• a processor execution

based on execution identifiers from Data Flow Tracking.

This turns recovery into a controlled operation:

• inspect what failed
• fix the root cause (credentials, upstream system, connectivity)
• retry only the relevant execution, without replaying everything blindly

Retries are only trustworthy when they’re reproducible.

If a pipeline definition changes while an execution is running, you get an ugly question:

> “Which version actually ran?”

So one of the operational safeguards is making pipeline definitions effectively stable during execution. That way, when you retry an execution, you’re retrying the same logic – unless you intentionally deploy a new version.

This is the kind of “boring correctness” that makes debugging and audits sane.

• explicit error semantics reduce debugging time
• retries become controlled operations, not guesswork
• execution identity becomes a first-class tool (“retry execution X”)
• fewer custom “recovery scripts” and manual reconciliations

• faster incident recovery
• fewer duplicate updates
• better transparency into what happened and what was retried
• easier operations handover (“here’s the execution ID and the status story”)

Operational maturity is revenue maturity:

• fewer support escalations
• higher trust from larger customers
• more complex use cases become feasible
• lower cost of operating at scale

In December we’ll zoom into a very concrete connector milestone:

Kaufland offer sync improvements – with a focus on update-only strategies that merchants can trust and agencies can implement cleanly.

Failures will happen.

The platform’s job isn’t to pretend they won’t.

The platform’s job is to make failure:

• observable
• classifiable
• recoverable

That’s the direction Qilin.Cloud is heading—so operations feel less like firefighting and more like engineering.

If you try to push binary payloads through every pipeline step, a few things go wrong fast:

• memory spikes
• request sizes explode
• retries become expensive
• logging and tracking become noisy (or impossible)
• transformations become unsafe (“did we accidentally re-encode it?”)

It’s the same lesson we learned years ago with large messages in messaging systems:

> Large payloads don’t belong in the pipe. They belong beside the pipe.

In Qilin.Cloud, binary content can be represented as a BlobReference object.

Instead of carrying raw bytes, a pipeline can carry metadata like:

– blob name / container name
– public or internal URL
– content type
– size
– optional tags
– original filename

So a flow object might include something like:

{
  "contentType": "image/jpeg",
  "fileName": "product_123.jpg",
  "size": 482193,
  "publicUrl": "https://…/product_123.jpg",
  "containerName": "assets"
}

Now every step downstream can refer to “the file” without transporting it.

That makes retries cheap, tracking clear, and throughput predictable.

1) Product image pipelines

• ingest images from a DAM or PIM
• store once
• publish to multiple destinations (Shopware, marketplaces, feeds)

2) Invoice / label workflows

• generate PDFs
• store references
• send links or forward to systems that fetch on demand

3) Bulk catalog exports

• assemble big files
• store as blobs
• send references to partners instead of pushing megabytes through APIs repeatedly

The HTTP Request Processor can handle multipart form uploads where the “file” is sourced from a URL (for example, a blob public URL).

That means you can build flows like:

1. receive an image URL from a webhook
2. store it as a blob (or reference it)
3. upload it to a destination API via multipart
4. continue processing with tracking intact

The pipeline stays stream-friendly and predictable.

Claim Check isn’t a new invention.

It’s one of those patterns seasoned integration engineers used long before “cloud native” became a phrase.

What’s new is that modern commerce workflows need it constantly – because media, documents, and binaries are now part of everyday commerce data.

Qilin.Cloud is making the pattern pipeline-native.

• Merchants: fewer failures when syncing images or documents; better reliability for catalog publishing.
• Agencies: cleaner implementations (no custom file-handling services), easier scaling across clients.
• Investors: binary handling is a capability multiplier—more workflows, bigger customers, less operational strain.

In November, we’ll zoom into the moment every integration fears:

something failed. Now what?

We’ll cover improvements around error handling, reproducibility, and manual retries—so operations teams can recover cleanly without rebuilding history from scratch.

Carry a claim ticket.

That’s what Qilin.Cloud is bringing to binary data workflows: clean references, clean tracking, and pipelines that don’t melt under file payloads.

Time Entry is an entry processor that starts a pipeline based on time.

It supports common scheduling styles:

• simple intervals (e.g., every 15 minutes)
• cron expressions (e.g., “every weekday at 08:00”)
• time zones (because commerce does not run on UTC feelings)

If you’ve been building scheduled integrations for years, you already know why this matters:

• fewer external schedulers
• fewer scattered jobs across servers
• schedules live with the pipeline definition
• everything is trackable and observable in one place

{
  "id": "time_entry",
  "type": "Qilin.TimeEntry",
  "config": {
    "scheduleType": "Cronjob",
    "cronExpression": "0 */15 * ? * *",
    "timeZone": "Europe/Berlin"
  }
}

That’s “every 15 minutes” in a predictable time zone.

No separate infrastructure. No mystery scheduler running on an old VM “because it always worked”.

The Email Processor does what it says:

• sends email to one or more recipients
• supports To / CC / BCC
• allows dynamic values from pipeline execution context
• supports templates via Scriban (so emails can be personalized and structured)
• uses stored credentials (so you don’t paste SMTP passwords into configs)

This becomes useful immediately for:

• failure alerts
• daily/weekly summaries
• “import finished” notifications
• “data quality issue detected” reports
• agency handover and operational transparency

A simplified config:

{
  "EmailProviderName": "CustomSmtpProvider",
  "credentialId": "credential-guid",
  "Receivers": { "raw": ["op*@*********ny.com"] },
  "From": { "raw": "qi**********@*********ny.com" },
  "Subject": "Qilin pipeline alert",
  "BodyContent": "Pipeline {{ FlowObjectAttributes.entry.pipelineId }} reported {{ FlowObjectAttributes.entry.status }}",
  "IsBodyHtml": false
}

You can also use templating to include richer context:

• object IDs
• execution IDs
• counts (how many items processed / failed)
• links to tracking pages

A scheduled pipeline without notifications is a silent machine.

A notification system without scheduling becomes reactive chaos.

Together, they enable a very traditional – but very effective – operational pattern:

1. run predictable jobs on predictable rhythms
2. surface exceptions immediately
3. let humans focus on decisions, not polling dashboards

• scheduling lives with the pipeline definition
• templates are versionable, testable, and dynamic
• credentials are managed centrally and permissioned
• fewer “sidecar services” just to send emails or run cron

• Merchants: get operational transparency without extra tooling
• Agencies: deliver “managed integrations” with built-in reporting and alerts
• Investors: these are features that reduce support cost and increase platform stickiness

In October we’ll tackle a very modern problem with a very classic pattern:

moving large binary data through pipelines without moving it through pipelines.

(Yes, that sentence is weird. It will make sense.)

Some of the best engineering work is turning “how we always did it” into platform primitives.

Time Entry and Email are exactly that: practical tools for real operations, baked into Qilin.Cloud.

If QQS + Buffer Entry is the “warehouse staging area”, the Batch Output Connector is the “truck”.

Instead of calling a connector once per object, it can:

• take a list of objects
• package them into a connector request
• send them in one call
• handle a response that contains success/error per item

A typical config looks like:

{
  "id": "batch_out",
  "type": "Qilin.BatchOutputConnector",
  "config": {
    "outputConnectorId": "connector-guid",
    "domainName": "Offer",
    "objectTypePath": "$.FlowObjectContents.entry.objectType",
    "objectIdsPath": "$.FlowObjectContents.entry.objectIds",
    "objectBodiesPath": "$.FlowObjectContents.entry.objectData"
  }
}

When paired with buffering, this gives you classic enterprise integration behavior:

• fewer HTTP calls
• better throughput control
• reduced connector overhead
• easier handling of platform quotas and rate limits

The next problem: not every object should go to the same place.

You might need to route by:

• marketplace
• storefront
• country
• product category
• pricing strategy
• stock availability
• compliance rules

In the old world, routing becomes a pile of nested if-statements.

In Qilin, the Switch-Case Processor gives you a structured approach:

• each outgoing connection can define a condition (Filter Predicate)
• Qilin activates the first matching case
• a default route can catch everything else

A simplified mental model:

• if case A matches → follow path A
• else if case B matches → follow path B
• else → follow default

This is the “switch statement” you already know – adapted to pipeline routing.

Multi-conditions are the difference between:

• “we can build a demo”
• and “we can build policies”

Because real routing often looks like:

• If it’s a hazardous material AND destination is EU → route to compliance enrichment
• If it’s a luxury brand AND price is missing → stop and alert
• If it’s marketplace X AND stock < 5 → route to low-stock strategy

Switch-case lets you express this cleanly, reuse it, and review it.

• batching reduces connector chatter and failure surfaces
• switch-case reduces custom code and makes routing auditable
• both are pipeline-native, meaning you get tracking, logs, and consistent error behavior

• Merchants: faster, more stable syncs—especially for high-volume catalog/offer updates
• Agencies: cleaner architectures and reusable routing templates across clients
• Investors: this is platform maturity that increases addressable use cases and lowers delivery cost

In September, we’ll shift from routing and batching into a different kind of control:

time.

Scheduling is a core part of integration work, and Qilin is adding pipeline primitives for time-based triggering and notifications.

Batches make shipping efficient.

Switch-case makes shipping intelligent.

Qilin.Cloud is building both—so your integrations can grow in volume and complexity without growing in chaos.

Think of QQS as a staging warehouse inside the platform.

Instead of pushing every object immediately to the next step, you can:

1. collect objects into a queue storage
2. wait until the queue reaches a condition (size or time)
3. trigger a downstream pipeline step with a batch

This is especially useful when destinations prefer bulk updates – or punish you with rate limits for chattiness.

At the conceptual level:

• a Queue Storage belongs to a subscription and has a configured duration (how long items live)
• a Queue Item stores:
  • object type
  • object ID
  • object data (serialized)
  • created timestamp
  • expiration timestamp

This gives you the platform equivalent of “put it in the staging area, and process it later”.

This processor takes the current object in your flow and pushes it into a chosen queue storage.

Example:

{
  "id": "push_to_queue",
  "type": "Qilin.PushQilinObjectToQueueStorage",
  "config": {
    "queueStorageId": "your-queue-storage-id",
    "objectType": "Offer",
    "domainName": "Offer",
    "domainObjectIdPath": "$.FlowObjectAttributes.entry.objectId"
  }
}

You can override the object ID or object type if needed, and you can write different object kinds to different queue storages.

Once objects are in the queue, you need a clean way to “release” them downstream.

That’s what Buffer Entry does.

Two common trigger patterns:

1) Threshold-based
Run when the queue contains at least N items:

{
  "id": "buffer_entry",
  "type": "Qilin.BufferEntry",
    "config": {
    "queueStorageId": "your-queue-storage-id",
    "minimumItemsToRelease": 150
  }
}

2) Schedule-based

Run on a schedule (cron expression), releasing what’s available.

This matches classic operational workflows: “every 15 minutes, push what changed.”

For developers

• fewer connector calls, less overhead
• easier throughput control with predictable bursts
• reduced retry complexity (retries happen per batch, not per object)
• clean separation between “capture changes” and “ship changes”

For merchants and agencies

• stable sync behavior under load
• predictable operational windows (e.g., nightly exports)
• fewer “API limit exceeded” incidents
• easier to align platform behavior with business rhythms

For investors

Queue storage unlocks higher-volume use cases without linear increases in operational cost. That’s leverage.

Offers (price, stock, handling time) can change frequently.

Instead of pushing every micro-change instantly:

• collect offer updates in QQS
• release them in batches every 10–15 minutes
• ship a predictable batch to the connector

This keeps data fresh enough for commerce… without turning your connector into a chatty stress test.

Once you can batch, the next step is:

smarter branching and routing.

August will introduce more around conditional execution and switch-case routing – so pipelines can decide what to do with a batch based on content, not just schedule.

Real-time is exciting.

Batching is profitable.

Qilin.Cloud is building both – so you can choose the right tool for each workflow, instead of forcing everything into one tempo.

In a composable pipeline, you often run things in parallel:

• fetch price from system A
• fetch description from system B
• fetch stock from system C

Eventually, you need to bring results back together.

That’s what the Merge Processor does.

But the tricky part is: how long do you wait, and what do you do when not everything arrives?

So Qilin supports different merge policies.

1) Wait for all input processors (default)

This is the “strict correctness” mode:

• Merge waits until all parent processors finish
• No timeout policy
• Best when every input is required for a valid result

Use this when incomplete data would be worse than delayed data.

2) Wait for any input processor

This is the “fast enough” mode:

• Merge starts a timeout countdown once any input finishes
• When the timeout ends, merge proceeds with whatever results arrived

Use this when you value speed and partial results are acceptable.

3) Wait for required input processors

This is the “hybrid” mode:

• define a list of required parents
• wait for those to complete
• start a timeout countdown afterward
• proceed even if non-required inputs didn’t arrive

Use this when some inputs are critical and others are optional enhancements.

A timeout is not always a failure. Sometimes it’s just a signal.

So processors can be configured with:

• TimeoutConfig (how long is “too long”)
• RetryConfig (how many times to retry and how long to wait between tries)
• OnErrorBehavior (continue vs stop)

A simple retry policy looks like:

"retryConfig": {
  "maximumRetries": 3,
  "retryAfterInSeconds": 3
}

This gives you structured resilience without rewriting your own scheduler logic.

Let’s say you enrich products using:

• ERP (core master data) → required
• PIM (marketing text) → optional
• Pricing service (dynamic pricing) → required

A “required inputs” merge policy lets you:

• wait for ERP + pricing
• proceed even if PIM is slow or unavailable
• still produce a usable product update

That’s a pipeline that keeps the business running without silently dropping correctness.

Alongside merge reliability, we’ve also been expanding processors that work directly with object content:

• content filtering based on reusable predicate logic
• content enrichment designed to combine or augment fields

The long-term goal: keep transformations inside the pipeline, where they can be tracked, tested, and maintained.

Developers

• clear semantics instead of “mystery retries”
• consistent timeout and retry behavior across processors
• less custom glue code for parallel fan-out / fan-in patterns

Merchants & agencies

• fewer brittle syncs that collapse when one dependency hiccups
• pipelines that degrade gracefully under partial failure
• easier troubleshooting because behavior is explicit, not implied

Investors

Reliability features reduce operational cost and support load. They’re not just technical niceties – they’re economic multipliers.

Once you can merge and retry reliably, you can start making a new trade:

batching vs real-time.

July will introduce more around queue-based buffering – because sometimes the smartest pipeline isn’t the fastest one, it’s the one that ships in stable, cost-efficient chunks.

The “old way” taught us the patterns.

Qilin.Cloud is turning those patterns into configurable primitives – so your pipelines can behave like experienced engineers wrote them… even when nobody is awake at 03:00 AM.