@toloka-platform/sdk 0.12.2 → 0.13.1

package/dist/cli/toloka.js.jsc.sha256

@@ -1 +1 @@
-8ed8ac9a70b9d78e84e61a5ab6f59647b65043707f6fd5a09e394e648151df5f  toloka.js.jsc
+279a8411c9d81b84cedaf92105db4990e733707701c0f78ef7a40b08787a7d71  toloka.js.jsc

package/dist/cli/toloka.js.sha256

@@ -1 +1 @@
-b5ed5a56f22841b7d917e9f4f7e974cb1a7d3c9fe4a0e58a1df5f8967a4be250  toloka.js
+9dede722e09fc4a46a7e7efcc968508799b374e50ea2b75731c6c3817a13b24c  toloka.js

package/docs/authoring.md

@@ -311,10 +311,10 @@ toloka validate
 
 A pipeline that is not on the platform yet is skipped until `toloka apply` creates it. The published JSON contract is still available as `validate(graph)` for importer and fixture tests; that helper also runs the local entity checks described under [generation](#generation) and makes no network call.
 
-A successful validate is necessary, not sufficient for launch. Gates outside the graph contract — an underpriced audience, a Training course that is enabled but unbuilt, a task UI that was never validated — can still block deployment. See [limitations](./limitations.md).
+A successful validate is necessary, not sufficient for launch. Gates outside the graph contract — a Training course that is enabled but unbuilt, a task UI that was never validated — can still block deployment. See [limitations](./limitations.md).
 
 ## Next
 
-- [Nodes](./nodes.md) — choosing a node, and the pricing floors
+- [Nodes](./nodes.md) — choosing a node, and the price range it must stay within
 - [Datasets](./datasets.md) — columns and immutability
 - [CLI](./cli.md) — the reconcile loop

package/docs/cli.md

@@ -350,6 +350,8 @@ toloka pipeline results <runId> [--limit <n>] [--failed]
 
 `pipeline run` without `--wait` prints `Started run <id> is <status> for <name>.` and a follow-up that uses the pipeline id. Exit code `0` means only that the run was accepted. With `--wait`, it prints `Run <id> succeeded.` or `Run <id> failed.` and sets the exit code accordingly; see [Exit codes](#exit-codes).
 
+A run needs at least one dataset item that matches the root node's dataset and filter. Otherwise `pipeline run` fails with `No dataset items match, so a run would have nothing to process.` and starts nothing. `apply` and `validate` still succeed over an empty dataset, so upload items or widen the filter, then run again.
+
 `pipeline status` reports the latest run of that pipeline. `pipeline runs` lists recent runs (default 10). `pipeline cost` forecasts what a run would bill. `pipeline cancel` and `pipeline results` take a run id.
 
 ## base-model
@@ -400,6 +402,23 @@ Hyperparameters are optional flags on `training create` and `training update`: `
 
 A Training learns on a ready dataset slice that freezes a split column, and `training create` needs `--dataset` or `--slice`. `--dataset <name-or-id>` fills the empty cells of the dataset's split column (a new `split` column, 80/10/10, when it has none), freezes a slice named after the Training, and trains on it; it waits up to 10 minutes for each step and prints one line per step before `Created training`. A split column adopted from an ordinary column has no shares, so `--dataset` does not fill it; give it shares first with `toloka dataset split <dataset> --column <name> --parts <name>=<roles>:<percent>,...`. `--slice <id>` takes a slice frozen with `toloka dataset slice create`. With several split columns, `--split-column` picks the one to train with. The Training learns on the column's train parts, or the ones `--parts` names, and validates on its validate parts, including a part that both validates and tests; without validation rows, the run holds out 10% of the training rows. Rows without a part are never trained on. `training update` keeps the slice, split column and train parts. A new `--dataset` or `--slice` resets the split column and train parts unless you pass them too; `--dataset` splits again and freezes a new slice named after the Training. `training get` prints the slice, split column and train parts. `training cost` leaves out rows that any split column now puts in a test part, a validate + test part of the Training's own column apart, and, when there are any, prints how many as `Left out as test`. The test parts of the Training's own column are outside the train parts to begin with, so they are not counted there. A Training saved before split slices were required is refused by `training cost` until `training update` gives it one.
 
+## batch-inference
+
+Answer a dataset's chat rows with a catalog Base Model or a Model Version of the project.
+
+```sh
+toloka batch-inference run --dataset <id> [--filter field=value] [--slice <slice-id>] [--parts <names> [--split-column <name>]] (--base-model <id> | --model-version <mdv.id|Model@name>) (--output-field <name> | --output-dataset <name>) --max-output-tokens <n> [--temperature 0..2] [--top-p <p>] [--stop <text>]... [--seed <n>] [--samples-per-row <n>] [--idempotency-key <key>] [--yes] [--json]
+toloka batch-inference list [--json]
+toloka batch-inference status <batch-inference-run-id> [--wait] [--json]
+toloka batch-inference confirm <batch-inference-run-id> [--json]
+toloka batch-inference resume <batch-inference-run-id> [--json]
+toloka batch-inference cancel <batch-inference-run-id> [--json]
+```
+
+Each selected row sends its `messages` (and `tools`, if any) as the prompt; a trailing assistant turn and the label column are ignored, so a training or holdout dataset works as-is. Rows with no user turn to answer, and prompts over about 16,384 tokens (estimated the way training estimates examples: each character counts by its script, about 5 Latin letters or 2.5 Cyrillic letters to a token and every digit one, plus the chat template's markup per turn and tool), are skipped and counted in the quote; when no row is left, the run fails with `no_requests` before anything is held. `--slice` answers only the rows of a ready slice, with the prompts it froze; `--parts test` answers only the rows in those parts of the dataset's split column, and `--split-column` picks the column when there are several. With both, a split column the slice froze keeps the parts it had when the slice was taken. `run` freezes the selection, waits for the quote and prints the number of rows and the maximum charge; nothing is held or spent until you confirm, either with `--yes` on `run` or later with `confirm` within 24 hours. `--output-field` writes each answer as a new value of a text column on the same dataset, creating the column if it does not exist. An existing column must allow strings, including a string column limited to an enum, a pattern or a length; a column whose schema allows no string (a number, a boolean, an object, or an enum, const, anyOf, oneOf or allOf with no string option) is refused with "does not accept text" before anything is quoted, and an answer that breaks a column's constraint fails that row. `--output-dataset` creates a new chat-sft dataset and adds a row for each answer as it arrives, keeping the prompt, the answer and the source row; `--samples-per-row` asks for several answers to each row (with a temperature above 0), one dataset row per answer with its `sample` number. A sample that fails adds no row. Answers appear in the dataset as they are generated, and while the run writes, the column shows "Updating · done/total", and the field's `operations` in the API list the write, so a slice, a split or another run that reads the column waits until it ends. Each answer is a new value of the column; rows the run left without one are found with the column filter for rows missing a value from the run, and `status` counts the failures by reason. The run is charged for the GPU time it used, at most the quoted maximum; a cancel keeps the answers already generated and charges the time used, and a platform failure charges nothing. `status --wait` follows a run until it finishes or needs you.
+
+Batch inference runs on qwen3-4b and qwen3-8b: a catalog Base Model, or a Model Version trained from one of them (LoRA, full weights or gisting). `run` refuses other models before anything is quoted. The charge is the GPU time used at the quoted price per GPU-hour: at least $0.50 per GPU once the job has used any GPU time, and never more than the quoted maximum. A cancel pays for the time used so far. A run that fails on the platform's side, times out, or whose results cannot be read at all charges nothing; when only some answers cannot be read, the run completes, charges the GPU time used and counts those rows as failed. A `blocked` run says why in `status`: `funds` when the balance cannot cover the hold (add funds, then `resume`); `payment` when the charge could not be collected (add funds and `resume` within seven days, after which the run fails with `payment_window_expired`; the answers already written stay); `billing_review` when support has to reconcile the charge. With a scoped API token, `list` and `status` need `inference:read`, and `run`, `confirm`, `resume` and `cancel` need `inference:deploy`: the `deploy` preset grants both, `read` and `write` grant only `inference:read`, and the labeling presets grant neither.
+
 ## dataset
 
 Catalog and row commands for datasets in the selected project. Name or id is accepted; duplicate names must be addressed by id.

package/docs/getting-started.md

@@ -141,7 +141,7 @@ Exit code `0` means the run reached a successful terminal state, `1` a failed on
 ## Next
 
 - [Authoring](./authoring.md) — the typed graph API
-- [Nodes](./nodes.md) — which node to reach for, and the pricing floors you must clear
+- [Nodes](./nodes.md) — which node to reach for, and the price range you must stay within
 - [CLI](./cli.md) — every command and flag
 - [Troubleshooting](./troubleshooting.md) — the failures you will hit
 - [Limitations](./limitations.md) — current constraints and `0.x` compatibility expectations

package/docs/nodes.md

@@ -63,7 +63,24 @@ Two defaults worth knowing before you author one:
 
 ## Pricing and time
 
-`audienceAndPricing` carries `pricePerTask` (USD) and `timeForTask` (seconds). `timeForTask` is a real duration, not a display hint. The shipped example uses `pricePerTask: 0.2` with `timeForTask: 60`.
+`audienceAndPricing` carries `pricePerTask` (USD) and `timeForTask` (seconds). `timeForTask` is a real duration, not a display hint: a whole number of seconds from 10 to 359,999 (99:59:59). The shipped example uses `pricePerTask: 0.2` with `timeForTask: 60`, the lowest price an ID-verified audience allows.
+
+You choose `pricePerTask`, and any value inside the audience's range is accepted. The range is set per hour of work and converted to one task as `rate × timeForTask / 3600`:
+
+| Audience | `pricePerTask` per hour of work | The editor suggests |
+|---|---|---|
+| ID-verified experts, no specialization | $12–$90 | $21 |
+| ID-verified experts with a specialization | from $12; the top follows the specialization's current rates | a rate set by the specialization |
+| Broad crowd (`verifiedExperts: false`) | $3–$24 | $6 |
+
+For a 120-second task with ID-verified experts and no specialization, that is $0.40–$3.00 per task.
+
+The platform checks the price and the audience on `toloka apply`, `toloka validate`, and every run:
+
+- A `pricePerTask` outside the range fails with an error that names the exact range for your audience and `timeForTask`.
+- The audience must match at least one expert, and every language, specialization, and country code must exist in the audience catalog. Countries are ISO 3166-1 alpha-2 codes such as `US`.
+
+With Human QA on (`qaMode: 'human_only'` or `'hybrid'`), `qualityAssurance.humanQaTimeForTask` is the time for one review: a whole number of seconds, at least 30 and at least 3 per quality criterion, up to 359,999 (99:59:59).
 
 ## Quorum data shape
 

package/examples/product-moderation/nodes/moderate-product/node.ts

@@ -46,7 +46,7 @@ export const moderateProduct = generation("moderate_product", {
       specializations: { type: "and", values: [] },
       countries: { type: "or", values: [] },
     },
-    pricePerTask: 0.2,
+    pricePerTask: 0.4,
     timeForTask: 120,
     perExpertCapEnabled: false,
     perExpertCapValue: 1,

package/examples/video-pair-rating/nodes/adjudicate/node.ts

@@ -227,7 +227,7 @@ export const adjudicate = generation("adjudicate", {
       specializations: { type: "and", values: [] },
       countries: { type: "or", values: [] },
     },
-    pricePerTask: 0.4,
+    pricePerTask: 0.6,
     timeForTask: 180,
     perExpertCapEnabled: false,
     perExpertCapValue: 1,

package/examples/video-pair-rating/nodes/adjudicate_op/node.ts

@@ -167,7 +167,7 @@ export const adjudicateOp = generation("adjudicate_op", {
       specializations: { type: "and", values: [] },
       countries: { type: "or", values: [] },
     },
-    pricePerTask: 0.4,
+    pricePerTask: 0.6,
     timeForTask: 180,
     perExpertCapEnabled: false,
     perExpertCapValue: 1,

package/examples/video-pair-rating/nodes/overall_preference_pick/node.ts

@@ -160,7 +160,7 @@ export const overallPreferencePick = quorum("overall_preference_pick", {
       specializations: { type: "and", values: [] },
       countries: { type: "or", values: [] },
     },
-    pricePerTask: 0.2,
+    pricePerTask: 0.4,
     timeForTask: 120,
     perExpertCapEnabled: false,
     perExpertCapValue: 1,

package/examples/video-pair-rating/nodes/rate_pair/node.ts

@@ -243,7 +243,7 @@ export const ratePair = quorum("rate_pair", {
       specializations: { type: "and", values: [] },
       countries: { type: "or", values: [] },
     },
-    pricePerTask: 0.2,
+    pricePerTask: 0.4,
     timeForTask: 120,
     perExpertCapEnabled: false,
     perExpertCapValue: 1,

package/openapi/gen/api.ts

@@ -4,9 +4,12 @@ import type {
   AttachPipelineDatasetBody,
   BaseModel,
   BaseModelId,
+  BatchInferenceRun,
+  BatchInferenceRunId,
   CheckJob,
   CheckRunId,
   CopyPipelineBody,
+  CreateBatchInferenceRunBody,
   CreateDatasetBody,
   CreateDatasetExportBody,
   CreateDatasetField200,
@@ -41,6 +44,8 @@ import type {
   ListAudienceSpecializations200,
   ListAudienceSpecializationsParams,
   ListBaseModels200,
+  ListBatchInferenceRuns200,
+  ListBatchInferenceRunsParams,
   ListDatasetFields200,
   ListDatasetFieldsParams,
   ListDatasetItems200,
@@ -2144,3 +2149,140 @@ export const getPipelineCheck = async (checkRunId: CheckRunId, options?: Request
     
   }
 );}
+
+
+
+export const getCreateBatchInferenceRunUrl = (projectId: ProjectId,) => {
+
+
+  
+
+  return `/api/v2-beta/projects/${projectId}/batch-inference-runs`
+}
+
+export const createBatchInferenceRun = async (projectId: ProjectId,
+    createBatchInferenceRunBody: CreateBatchInferenceRunBody, options?: RequestInit): Promise<BatchInferenceRun> => {
+  
+  return potatoFetch<BatchInferenceRun>(getCreateBatchInferenceRunUrl(projectId),
+  {      
+    ...options,
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json', ...options?.headers },
+    body: JSON.stringify(
+      createBatchInferenceRunBody,)
+  }
+);}
+
+
+
+export const getListBatchInferenceRunsUrl = (projectId: ProjectId,
+    params?: ListBatchInferenceRunsParams,) => {
+  const normalizedParams = new URLSearchParams();
+
+  Object.entries(params || {}).forEach(([key, value]) => {
+    
+    if (value !== undefined) {
+      normalizedParams.append(key, value === null ? 'null' : value.toString())
+    }
+  });
+
+  const stringifiedParams = normalizedParams.toString();
+
+  return stringifiedParams.length > 0 ? `/api/v2-beta/projects/${projectId}/batch-inference-runs?${stringifiedParams}` : `/api/v2-beta/projects/${projectId}/batch-inference-runs`
+}
+
+export const listBatchInferenceRuns = async (projectId: ProjectId,
+    params?: ListBatchInferenceRunsParams, options?: RequestInit): Promise<ListBatchInferenceRuns200> => {
+  
+  return potatoFetch<ListBatchInferenceRuns200>(getListBatchInferenceRunsUrl(projectId,params),
+  {      
+    ...options,
+    method: 'GET'
+    
+    
+  }
+);}
+
+
+
+export const getGetBatchInferenceRunUrl = (batchInferenceRunId: BatchInferenceRunId,) => {
+
+
+  
+
+  return `/api/v2-beta/batch-inference-runs/${batchInferenceRunId}`
+}
+
+export const getBatchInferenceRun = async (batchInferenceRunId: BatchInferenceRunId, options?: RequestInit): Promise<BatchInferenceRun> => {
+  
+  return potatoFetch<BatchInferenceRun>(getGetBatchInferenceRunUrl(batchInferenceRunId),
+  {      
+    ...options,
+    method: 'GET'
+    
+    
+  }
+);}
+
+
+
+export const getConfirmBatchInferenceRunUrl = (batchInferenceRunId: BatchInferenceRunId,) => {
+
+
+  
+
+  return `/api/v2-beta/batch-inference-runs/${batchInferenceRunId}/confirm`
+}
+
+export const confirmBatchInferenceRun = async (batchInferenceRunId: BatchInferenceRunId, options?: RequestInit): Promise<BatchInferenceRun> => {
+  
+  return potatoFetch<BatchInferenceRun>(getConfirmBatchInferenceRunUrl(batchInferenceRunId),
+  {      
+    ...options,
+    method: 'POST'
+    
+    
+  }
+);}
+
+
+
+export const getResumeBatchInferenceRunUrl = (batchInferenceRunId: BatchInferenceRunId,) => {
+
+
+  
+
+  return `/api/v2-beta/batch-inference-runs/${batchInferenceRunId}/resume`
+}
+
+export const resumeBatchInferenceRun = async (batchInferenceRunId: BatchInferenceRunId, options?: RequestInit): Promise<BatchInferenceRun> => {
+  
+  return potatoFetch<BatchInferenceRun>(getResumeBatchInferenceRunUrl(batchInferenceRunId),
+  {      
+    ...options,
+    method: 'POST'
+    
+    
+  }
+);}
+
+
+
+export const getCancelBatchInferenceRunUrl = (batchInferenceRunId: BatchInferenceRunId,) => {
+
+
+  
+
+  return `/api/v2-beta/batch-inference-runs/${batchInferenceRunId}/cancel`
+}
+
+export const cancelBatchInferenceRun = async (batchInferenceRunId: BatchInferenceRunId, options?: RequestInit): Promise<BatchInferenceRun> => {
+  
+  return potatoFetch<BatchInferenceRun>(getCancelBatchInferenceRunUrl(batchInferenceRunId),
+  {      
+    ...options,
+    method: 'POST'
+    
+    
+  }
+);}

package/openapi/gen/models/batchInferenceMode.ts

@@ -0,0 +1,3 @@
+import type { BatchInferenceModeOneOf } from './batchInferenceModeOneOf';
+
+export type BatchInferenceMode = BatchInferenceModeOneOf;

package/openapi/gen/models/batchInferenceModeOneOf.ts

@@ -0,0 +1,4 @@
+
+export type BatchInferenceModeOneOf = {
+  type: 'single_turn';
+};

package/openapi/gen/models/batchInferenceModel.ts

@@ -0,0 +1,4 @@
+import type { BatchInferenceModelOneOf } from './batchInferenceModelOneOf';
+import type { BatchInferenceModelOneOfThree } from './batchInferenceModelOneOfThree';
+
+export type BatchInferenceModel = BatchInferenceModelOneOf | BatchInferenceModelOneOfThree;

package/openapi/gen/models/batchInferenceModelOneOf.ts

@@ -0,0 +1,6 @@
+import type { BatchInferenceModelOneOfBaseModelId } from './batchInferenceModelOneOfBaseModelId';
+
+export type BatchInferenceModelOneOf = {
+  type: 'base_model';
+  baseModelId: BatchInferenceModelOneOfBaseModelId;
+};

package/openapi/gen/models/batchInferenceModelOneOfBaseModelId.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceModelOneOfBaseModelId = 'qwen3-4b' | 'qwen3-8b' | 'qwen3-32b' | 'qwen3-235b';

package/openapi/gen/models/batchInferenceModelOneOfThree.ts

@@ -0,0 +1,5 @@
+
+export type BatchInferenceModelOneOfThree = {
+  type: 'model_version';
+  modelVersionId: string;
+};

package/openapi/gen/models/batchInferenceRun.ts

@@ -0,0 +1,35 @@
+import type { BatchInferenceRunId } from './batchInferenceRunId';
+import type { ProjectId } from './projectId';
+import type { DatasetId } from './datasetId';
+import type { BatchInferenceModel } from './batchInferenceModel';
+import type { BatchInferenceMode } from './batchInferenceMode';
+import type { BatchInferenceRunSampling } from './batchInferenceRunSampling';
+import type { BatchInferenceRunOutput } from './batchInferenceRunOutput';
+import type { BatchInferenceRunStatus } from './batchInferenceRunStatus';
+import type { BatchInferenceRunProgress } from './batchInferenceRunProgress';
+import type { BatchInferenceRunBlockedReason } from './batchInferenceRunBlockedReason';
+import type { BatchInferenceRunMessage } from './batchInferenceRunMessage';
+import type { BatchInferenceRunError } from './batchInferenceRunError';
+import type { BatchInferenceRunQuote } from './batchInferenceRunQuote';
+import type { BatchInferenceRunChargedAmount } from './batchInferenceRunChargedAmount';
+import type { BatchInferenceRunResult } from './batchInferenceRunResult';
+
+export interface BatchInferenceRun {
+  id: BatchInferenceRunId;
+  projectId: ProjectId;
+  datasetId: DatasetId;
+  model: BatchInferenceModel;
+  mode: BatchInferenceMode;
+  sampling: BatchInferenceRunSampling;
+  output: BatchInferenceRunOutput;
+  status: BatchInferenceRunStatus;
+  progress: BatchInferenceRunProgress;
+  blockedReason: BatchInferenceRunBlockedReason;
+  message: BatchInferenceRunMessage;
+  error: BatchInferenceRunError;
+  quote: BatchInferenceRunQuote;
+  chargedAmount: BatchInferenceRunChargedAmount;
+  result: BatchInferenceRunResult;
+  createdAt: string;
+  updatedAt: string;
+}

package/openapi/gen/models/batchInferenceRunBlockedReason.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunBlockedReason = 'funds' | 'payment' | 'billing_review' | null;

package/openapi/gen/models/batchInferenceRunChargedAmount.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunChargedAmount = string | null;

package/openapi/gen/models/batchInferenceRunError.ts

@@ -0,0 +1,3 @@
+import type { BatchInferenceRunErrorAnyOf } from './batchInferenceRunErrorAnyOf';
+
+export type BatchInferenceRunError = BatchInferenceRunErrorAnyOf | null;

package/openapi/gen/models/batchInferenceRunErrorAnyOf.ts

@@ -0,0 +1,6 @@
+import type { BatchInferenceRunErrorAnyOfCode } from './batchInferenceRunErrorAnyOfCode';
+
+export type BatchInferenceRunErrorAnyOf = {
+  code: BatchInferenceRunErrorAnyOfCode;
+  message: string;
+};

package/openapi/gen/models/batchInferenceRunErrorAnyOfCode.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunErrorAnyOfCode = 'requests_export_failed' | 'no_requests' | 'selection_too_large' | 'billing_rejected' | 'provider_failed' | 'provider_timeout' | 'invalid_provider_output' | 'output_rejected' | 'payment_window_expired' | 'canceled';

package/openapi/gen/models/batchInferenceRunId.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunId = string;

package/openapi/gen/models/batchInferenceRunMessage.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunMessage = string | null;

package/openapi/gen/models/batchInferenceRunOutput.ts

@@ -0,0 +1,9 @@
+import type { BatchInferenceRunOutputType } from './batchInferenceRunOutputType';
+import type { DatasetId } from './datasetId';
+import type { DatasetFieldId } from './datasetFieldId';
+
+export type BatchInferenceRunOutput = {
+  type: BatchInferenceRunOutputType;
+  datasetId: DatasetId;
+  fieldId: DatasetFieldId;
+};

package/openapi/gen/models/batchInferenceRunOutputType.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunOutputType = 'field' | 'dataset';

package/openapi/gen/models/batchInferenceRunProgress.ts

@@ -0,0 +1,3 @@
+import type { BatchInferenceRunProgressAnyOf } from './batchInferenceRunProgressAnyOf';
+
+export type BatchInferenceRunProgress = BatchInferenceRunProgressAnyOf | null;

package/openapi/gen/models/batchInferenceRunProgressAnyOf.ts

@@ -0,0 +1,7 @@
+import type { BatchInferenceRunProgressAnyOfPhase } from './batchInferenceRunProgressAnyOfPhase';
+import type { BatchInferenceRunProgressAnyOfRequests } from './batchInferenceRunProgressAnyOfRequests';
+
+export type BatchInferenceRunProgressAnyOf = {
+  phase: BatchInferenceRunProgressAnyOfPhase;
+  requests?: BatchInferenceRunProgressAnyOfRequests;
+};

package/openapi/gen/models/batchInferenceRunProgressAnyOfPhase.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunProgressAnyOfPhase = 'queued' | 'preparing' | 'generating' | 'finishing';

package/openapi/gen/models/batchInferenceRunProgressAnyOfRequests.ts

@@ -0,0 +1,5 @@
+
+export type BatchInferenceRunProgressAnyOfRequests = {
+  done: number;
+  total: number;
+};

package/openapi/gen/models/batchInferenceRunQuote.ts

@@ -0,0 +1,3 @@
+import type { BatchInferenceRunQuoteAnyOf } from './batchInferenceRunQuoteAnyOf';
+
+export type BatchInferenceRunQuote = BatchInferenceRunQuoteAnyOf | null;

package/openapi/gen/models/batchInferenceRunQuoteAnyOf.ts

@@ -0,0 +1,9 @@
+import type { BatchInferenceRunQuoteAnyOfConfirmBefore } from './batchInferenceRunQuoteAnyOfConfirmBefore';
+
+export type BatchInferenceRunQuoteAnyOf = {
+  requests: number;
+  skippedRows: number;
+  maximumCharge: string;
+  currency: 'USD';
+  confirmBefore: BatchInferenceRunQuoteAnyOfConfirmBefore;
+};

package/openapi/gen/models/batchInferenceRunQuoteAnyOfConfirmBefore.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunQuoteAnyOfConfirmBefore = string | null;

package/openapi/gen/models/batchInferenceRunResult.ts

@@ -0,0 +1,3 @@
+import type { BatchInferenceRunResultAnyOf } from './batchInferenceRunResultAnyOf';
+
+export type BatchInferenceRunResult = BatchInferenceRunResultAnyOf | null;

package/openapi/gen/models/batchInferenceRunResultAnyOf.ts

@@ -0,0 +1,12 @@
+import type { BatchInferenceRunResultAnyOfErrorCounts } from './batchInferenceRunResultAnyOfErrorCounts';
+import type { BatchInferenceRunResultAnyOfErrorSamplesItem } from './batchInferenceRunResultAnyOfErrorSamplesItem';
+
+export type BatchInferenceRunResultAnyOf = {
+  written: number;
+  failed: number;
+  truncated: number;
+  inputTokens: number;
+  outputTokens: number;
+  errorCounts: BatchInferenceRunResultAnyOfErrorCounts;
+  errorSamples: BatchInferenceRunResultAnyOfErrorSamplesItem[];
+};

package/openapi/gen/models/batchInferenceRunResultAnyOfErrorCounts.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunResultAnyOfErrorCounts = {[key: string]: number};

package/openapi/gen/models/batchInferenceRunResultAnyOfErrorSamplesItem.ts

@@ -0,0 +1,6 @@
+
+export type BatchInferenceRunResultAnyOfErrorSamplesItem = {
+  itemId: string;
+  code: string;
+  message: string;
+};

package/openapi/gen/models/batchInferenceRunSampling.ts

@@ -0,0 +1,9 @@
+
+export type BatchInferenceRunSampling = {
+  maxOutputTokens: number;
+  temperature?: number;
+  topP?: number;
+  stop?: string[];
+  seed?: number;
+  samplesPerRow?: number;
+};

package/openapi/gen/models/batchInferenceRunStatus.ts

@@ -0,0 +1,2 @@
+
+export type BatchInferenceRunStatus = 'preparing' | 'awaiting_confirmation' | 'queued' | 'running' | 'writing_outputs' | 'stopping' | 'blocked' | 'completed' | 'failed' | 'canceled';

package/openapi/gen/models/createBatchInferenceRunBody.ts

@@ -0,0 +1,14 @@
+import type { CreateBatchInferenceRunBodyItemFilter } from './createBatchInferenceRunBodyItemFilter';
+import type { BatchInferenceModel } from './batchInferenceModel';
+import type { BatchInferenceMode } from './batchInferenceMode';
+import type { CreateBatchInferenceRunBodySampling } from './createBatchInferenceRunBodySampling';
+import type { CreateBatchInferenceRunBodyOutput } from './createBatchInferenceRunBodyOutput';
+
+export type CreateBatchInferenceRunBody = {
+  datasetId: string;
+  itemFilter?: CreateBatchInferenceRunBodyItemFilter;
+  model: BatchInferenceModel;
+  mode?: BatchInferenceMode;
+  sampling: CreateBatchInferenceRunBodySampling;
+  output: CreateBatchInferenceRunBodyOutput;
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilter.ts

@@ -0,0 +1,14 @@
+import type { CreateBatchInferenceRunBodyItemFilterFieldValuesItem } from './createBatchInferenceRunBodyItemFilterFieldValuesItem';
+import type { CreateBatchInferenceRunBodyItemFilterFieldSourcesItem } from './createBatchInferenceRunBodyItemFilterFieldSourcesItem';
+import type { CreateBatchInferenceRunBodyItemFilterSlice } from './createBatchInferenceRunBodyItemFilterSlice';
+import type { CreateBatchInferenceRunBodyItemFilterParts } from './createBatchInferenceRunBodyItemFilterParts';
+import type { CreateBatchInferenceRunBodyItemFilterExcludePartsItem } from './createBatchInferenceRunBodyItemFilterExcludePartsItem';
+
+export type CreateBatchInferenceRunBodyItemFilter = {
+  fieldValues?: CreateBatchInferenceRunBodyItemFilterFieldValuesItem[];
+  fieldSources?: CreateBatchInferenceRunBodyItemFilterFieldSourcesItem[];
+  missingRevisionFields?: string[];
+  slice?: CreateBatchInferenceRunBodyItemFilterSlice;
+  parts?: CreateBatchInferenceRunBodyItemFilterParts;
+  excludeParts?: CreateBatchInferenceRunBodyItemFilterExcludePartsItem[];
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterExcludePartsItem.ts

@@ -0,0 +1,5 @@
+
+export type CreateBatchInferenceRunBodyItemFilterExcludePartsItem = {
+  fieldId: string;
+  names: string[];
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldSourcesItem.ts

@@ -0,0 +1,6 @@
+import type { CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItem } from './createBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItem';
+
+export type CreateBatchInferenceRunBodyItemFilterFieldSourcesItem = {
+  fieldId: string;
+  sources: CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItem[];
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItem.ts

@@ -0,0 +1,4 @@
+import type { CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOf } from './createBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOf';
+import type { CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOfTwo } from './createBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOfTwo';
+
+export type CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItem = CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOf | CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOfTwo;

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOf.ts

@@ -0,0 +1,5 @@
+
+export type CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOf = {
+  type: 'user';
+  userId: string;
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOfTwo.ts

@@ -0,0 +1,5 @@
+
+export type CreateBatchInferenceRunBodyItemFilterFieldSourcesItemSourcesItemOneOfTwo = {
+  type: 'system';
+  source: string;
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldValuesItem.ts

@@ -0,0 +1,12 @@
+import type { CreateBatchInferenceRunBodyItemFilterFieldValuesItemOperator } from './createBatchInferenceRunBodyItemFilterFieldValuesItemOperator';
+import type { CreateBatchInferenceRunBodyItemFilterFieldValuesItemValue } from './createBatchInferenceRunBodyItemFilterFieldValuesItemValue';
+import type { CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItem } from './createBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItem';
+
+export type CreateBatchInferenceRunBodyItemFilterFieldValuesItem = {
+  jsonPath: string;
+  operator: CreateBatchInferenceRunBodyItemFilterFieldValuesItemOperator;
+  value: CreateBatchInferenceRunBodyItemFilterFieldValuesItemValue;
+  fieldId: string;
+  latestOnly?: boolean;
+  sources?: CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItem[];
+};

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldValuesItemOperator.ts

@@ -0,0 +1,2 @@
+
+export type CreateBatchInferenceRunBodyItemFilterFieldValuesItemOperator = 'text_eq' | 'text_neq';

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItem.ts

@@ -0,0 +1,4 @@
+import type { CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOf } from './createBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOf';
+import type { CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOfTwo } from './createBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOfTwo';
+
+export type CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItem = CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOf | CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOfTwo;

package/openapi/gen/models/createBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOf.ts

@@ -0,0 +1,5 @@
+
+export type CreateBatchInferenceRunBodyItemFilterFieldValuesItemSourcesItemOneOf = {
+  type: 'user';
+  userId: string;
+};