
Read and compare the current authenticated model catalog before releasing unsent work that depends on a model or resolution. Keep a dated, validated snapshot in your application and alert on meaningful differences. A failed discovery request is not evidence that every model disappeared, and a catalog update must not rewrite already submitted task history.
Key Takeaways
- Compare semantic capabilities and points, not JSON object order.
- Pause affected unsent operations for review instead of inventing an automatic fallback.
- Preserve the original model, input and quote on accepted operations.
Validate discovery before comparing it
The Imagild API guide documents GET /models, whose response contains data.models. It requires a valid key and active membership. Current entries represent configured available choices.
Check the HTTP outcome and the expected envelope before replacing a known snapshot. A 401, 403, timeout or malformed response should produce a discovery problem. It should not be converted into an empty successful model list.
Fetch at an application-appropriate frequency and cache the result for settings screens. Catalog reads share the account's read budget with task checks and downloads; there is little value in rediscovering it for every rendered task card.
Compare the actual contract fields
For image entries, compare IDs, supported resolution tiers and point values. For video entries, relevant fields include ratios, minimum and maximum seconds, supported resolutions, enhancement capability and current point rates.
This image-only fingerprint assumes you already validated a catalog entry:
function imageContractFingerprint(model) {
return JSON.stringify({
id: model.id,
resolutions: [...model.resolutions].sort(),
points: Object.entries(model.points)
.sort(([left], [right]) => left.localeCompare(right))
});
}
Sorting avoids a false change caused merely by ordering. Do not use this image helper as a complete validator for video entries or the entire response.
Runware's schema documentation illustrates why capabilities belong to the receiving model contract. Its fields and routing IDs are not additional Imagild options.
Classify changes before changing the interface
A newly advertised model can become an available draft choice after review. A removed choice should stop new submissions that depend on it. A point change should trigger a fresh quote for unsent work.
Do not silently translate an unavailable ID to a similarly named alias. Do not change a saved request to another model simply because a catalog diff exists. Those actions create new creative and spending decisions.
Check the model hub when explaining current public choices. Keep internal channel assumptions out of the selector; a routing name does not prove equivalence of every provider implementation.
Preserve accepted tasks while repairing drafts
Accepted operations retain their original task ID, submitted parameters and stored point quote. Changing the current dropdown should not relabel previous results or restart processing.
When an unsent draft is affected, show the user what changed: model availability, resolution support or the estimated points. Ask for a deliberate revised generation decision through your own application workflow rather than an automatic cross-provider fallback.
Use the points and jobs guide to communicate budget changes. Use the failed-job checks if a task itself fails; that is separate from discovery drift.
Test detection without manufacturing an outage
Create local snapshots with reordered resolutions, one removed model, one changed point value and a failed catalog read. Reordering alone should not alert. Meaningful changes should identify affected drafts, while the failed read retains the last known snapshot with a visible stale label.
Pass the monitor when it can explain the difference and preserve job history. Periodic discovery improves awareness; it does not guarantee that no request will ever encounter a change between its last check and submission.
Related tasks
- Imagild Points, Resolution and Jobs: Plan a Creation Before You Spend
- What Should You Check Before Retrying a Failed Image or Video Job?
Sources and scope
- Imagild API console: Actual request and task envelopes, strict parameters, references, membership, points usage and authenticated result retrieval.
- Imagild model hub: Current advertised model capabilities, resolutions and generation point rates.
- Runware model schemas: Model-specific capabilities belong to the receiving contract; Runware identifiers and fields do not become Imagild options.
Product instructions were checked against the current Imagild API contract and the public model catalog. Code, budgets and acceptance checks are illustrative; this article does not claim a new paid generation test, an uptime measurement, a search-volume study or a ranking of AI citations.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: Original Imagild product screenshot captured on 28 October 2026; model availability and prices may change. This image is not an API response or a performance benchmark.