@civitai/sdk 0.12.0 → 0.12.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md
CHANGED
|
@@ -492,15 +492,15 @@ Do **one of two things, and never mint a new key for that press**:
|
|
|
492
492
|
🔴 **For a `kind: 'training'` body and a step body carrying a `quoteId`, the
|
|
493
493
|
key is not the run's identity.** The host does not use the client key on those
|
|
494
494
|
two; the run is identified by the confirmed quote (a step) or by the training
|
|
495
|
-
run itself.
|
|
496
|
-
carries. So after an unknown outcome on one of these, check
|
|
495
|
+
run itself. So after an unknown outcome on one of these, check
|
|
497
496
|
`app.workflows.list()` **before** asking the viewer to confirm another quote.
|
|
498
497
|
|
|
499
498
|
What this client can promise stops there. Matching a repeated key to an
|
|
500
499
|
existing run is done by the host and the orchestrator, not here; it is not
|
|
501
|
-
guaranteed to match, and the host remembers a key only for a limited time.
|
|
502
|
-
`list()`
|
|
503
|
-
|
|
500
|
+
guaranteed to match, and the host remembers a key only for a limited time.
|
|
501
|
+
`list()` does not settle it either: a listed workflow carries no idempotency
|
|
502
|
+
key, so it cannot be tied to a press for certain, and a run missing from it is
|
|
503
|
+
not proof that nothing ran.
|
|
504
504
|
|
|
505
505
|
```ts
|
|
506
506
|
import { ApiError, SubmissionUnconfirmedError } from '@civitai/sdk';
|
|
@@ -565,7 +565,7 @@ server text — log it and show your own copy.
|
|
|
565
565
|
without one. The key is what lets the host recognise a submit that was sent
|
|
566
566
|
twice — a retry after a dropped connection, a double press — as one request
|
|
567
567
|
(for a `kind: 'training'` body or a step with a `quoteId`, see the note above:
|
|
568
|
-
there the
|
|
568
|
+
there the key is not the run's identity).
|
|
569
569
|
|
|
570
570
|
- Mint it **once per thing the viewer asked for**, keep it until that press has
|
|
571
571
|
a known outcome, and reuse it on every retry. A key minted inside the retry
|
|
@@ -1,2 +1,2 @@
|
|
|
1
1
|
/** This package's version, reported to the host as `sdkVersion` in `BLOCK_READY`. */
|
|
2
|
-
export declare const VERSION = "0.12.
|
|
2
|
+
export declare const VERSION = "0.12.1";
|
|
@@ -138,13 +138,15 @@ export interface WorkflowsClient {
|
|
|
138
138
|
* {@link WorkflowsClient.list}. Never mint a new key for the same viewer
|
|
139
139
|
* action: to the host a new key is a new generation. That matching is the
|
|
140
140
|
* host's and the orchestrator's, is not guaranteed, and is not something
|
|
141
|
-
* this client can verify
|
|
141
|
+
* this client can verify. Nor does `list` settle it: a listed workflow
|
|
142
|
+
* carries no idempotency key, and a run missing from it is not proof that
|
|
143
|
+
* nothing ran.
|
|
142
144
|
*
|
|
143
145
|
* 🔴 For a `kind: 'training'` body and a step body carrying a `quoteId` the
|
|
144
146
|
* key is NOT the run's identity: the host does not use it there, and
|
|
145
|
-
* identifies the run by the confirmed quote or the training run.
|
|
146
|
-
*
|
|
147
|
-
*
|
|
147
|
+
* identifies the run by the confirmed quote or the training run. After an
|
|
148
|
+
* unknown outcome on one of these, check `list` before asking the viewer to
|
|
149
|
+
* confirm another quote.
|
|
148
150
|
*
|
|
149
151
|
* This client does not retry a failed submit itself; it only re-sends once
|
|
150
152
|
* after refreshing an expired token, as every call does.
|
package/package.json
CHANGED