@prestyj/core 5.28.1 → 5.29.1
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/dist/{chunk-CK6UIITE.js → chunk-N73LFQ3N.js} +39 -25
- package/dist/chunk-N73LFQ3N.js.map +1 -0
- package/dist/index.cjs +70 -33
- package/dist/index.cjs.map +1 -1
- package/dist/index.d.cts +12 -2
- package/dist/index.d.ts +12 -2
- package/dist/index.js +35 -12
- package/dist/index.js.map +1 -1
- package/dist/model-registry.cjs +11 -17
- package/dist/model-registry.cjs.map +1 -1
- package/dist/model-registry.d.cts +11 -19
- package/dist/model-registry.d.ts +11 -19
- package/dist/model-registry.js +1 -3
- package/package.json +2 -2
- package/dist/chunk-CK6UIITE.js.map +0 -1
|
@@ -134,25 +134,17 @@ declare function getDefaultThinkingLevel(modelId: string, options?: {
|
|
|
134
134
|
baseUrl?: string;
|
|
135
135
|
}): ThinkingLevel;
|
|
136
136
|
/**
|
|
137
|
-
* Get the model to use for compaction summarization
|
|
138
|
-
*
|
|
139
|
-
*
|
|
140
|
-
* - Gemini: use the current model
|
|
141
|
-
* - GLM: GLM-5.3-Flash (the registered low-cost sibling)
|
|
142
|
-
* - Moonshot: use the current model (no cheap alternative registered)
|
|
143
|
-
*/
|
|
144
|
-
declare function getSummaryModel(provider: Provider, currentModelId: string): ModelInfo;
|
|
145
|
-
/**
|
|
146
|
-
* Fastest/cheapest sibling within the SAME provider, for scout-style read-only
|
|
147
|
-
* sub-agents (recon, research) where a low-latency model is enough and the
|
|
148
|
-
* frontier model is wasted spend + latency.
|
|
137
|
+
* Get the model to use for compaction summarization — the ONLY place a
|
|
138
|
+
* cheaper model is used. Everything else (sub-agents, reviewers, judges)
|
|
139
|
+
* runs on the parent's model.
|
|
149
140
|
*
|
|
150
|
-
*
|
|
151
|
-
*
|
|
152
|
-
*
|
|
153
|
-
*
|
|
154
|
-
*
|
|
141
|
+
* Picks the provider's first model tagged with its summary tier
|
|
142
|
+
* ({@link SUMMARY_COST_TIER}). When none is registered — the tier was removed,
|
|
143
|
+
* or the provider has no cheap sibling — it summarizes on the current model.
|
|
144
|
+
* Never throws. The caller still retries on the current model when the
|
|
145
|
+
* provider rejects the chosen one at runtime (retired upstream, or not on the
|
|
146
|
+
* user's plan), since the registry cannot know that.
|
|
155
147
|
*/
|
|
156
|
-
declare function
|
|
148
|
+
declare function getSummaryModel(provider: Provider, currentModelId: string): ModelInfo;
|
|
157
149
|
|
|
158
|
-
export { type ContextWindowOptions, DEFAULT_MAX_VIDEO_BYTES, MODELS, type ModelInfo, clearRuntimeModels, getAllModels, getAuthStorageKey, getAuthStorageKeys, getContextWindow, getDefaultModel, getDefaultThinkingLevel,
|
|
150
|
+
export { type ContextWindowOptions, DEFAULT_MAX_VIDEO_BYTES, MODELS, type ModelInfo, clearRuntimeModels, getAllModels, getAuthStorageKey, getAuthStorageKeys, getContextWindow, getDefaultModel, getDefaultThinkingLevel, getMaxThinkingLevel, getModel, getModelsForProvider, getSummaryModel, getToolResultCharLimit, getVideoByteLimit, registerRuntimeModels, usesOpenAICodexTransport };
|
package/dist/model-registry.d.ts
CHANGED
|
@@ -134,25 +134,17 @@ declare function getDefaultThinkingLevel(modelId: string, options?: {
|
|
|
134
134
|
baseUrl?: string;
|
|
135
135
|
}): ThinkingLevel;
|
|
136
136
|
/**
|
|
137
|
-
* Get the model to use for compaction summarization
|
|
138
|
-
*
|
|
139
|
-
*
|
|
140
|
-
* - Gemini: use the current model
|
|
141
|
-
* - GLM: GLM-5.3-Flash (the registered low-cost sibling)
|
|
142
|
-
* - Moonshot: use the current model (no cheap alternative registered)
|
|
143
|
-
*/
|
|
144
|
-
declare function getSummaryModel(provider: Provider, currentModelId: string): ModelInfo;
|
|
145
|
-
/**
|
|
146
|
-
* Fastest/cheapest sibling within the SAME provider, for scout-style read-only
|
|
147
|
-
* sub-agents (recon, research) where a low-latency model is enough and the
|
|
148
|
-
* frontier model is wasted spend + latency.
|
|
137
|
+
* Get the model to use for compaction summarization — the ONLY place a
|
|
138
|
+
* cheaper model is used. Everything else (sub-agents, reviewers, judges)
|
|
139
|
+
* runs on the parent's model.
|
|
149
140
|
*
|
|
150
|
-
*
|
|
151
|
-
*
|
|
152
|
-
*
|
|
153
|
-
*
|
|
154
|
-
*
|
|
141
|
+
* Picks the provider's first model tagged with its summary tier
|
|
142
|
+
* ({@link SUMMARY_COST_TIER}). When none is registered — the tier was removed,
|
|
143
|
+
* or the provider has no cheap sibling — it summarizes on the current model.
|
|
144
|
+
* Never throws. The caller still retries on the current model when the
|
|
145
|
+
* provider rejects the chosen one at runtime (retired upstream, or not on the
|
|
146
|
+
* user's plan), since the registry cannot know that.
|
|
155
147
|
*/
|
|
156
|
-
declare function
|
|
148
|
+
declare function getSummaryModel(provider: Provider, currentModelId: string): ModelInfo;
|
|
157
149
|
|
|
158
|
-
export { type ContextWindowOptions, DEFAULT_MAX_VIDEO_BYTES, MODELS, type ModelInfo, clearRuntimeModels, getAllModels, getAuthStorageKey, getAuthStorageKeys, getContextWindow, getDefaultModel, getDefaultThinkingLevel,
|
|
150
|
+
export { type ContextWindowOptions, DEFAULT_MAX_VIDEO_BYTES, MODELS, type ModelInfo, clearRuntimeModels, getAllModels, getAuthStorageKey, getAuthStorageKeys, getContextWindow, getDefaultModel, getDefaultThinkingLevel, getMaxThinkingLevel, getModel, getModelsForProvider, getSummaryModel, getToolResultCharLimit, getVideoByteLimit, registerRuntimeModels, usesOpenAICodexTransport };
|
package/dist/model-registry.js
CHANGED
|
@@ -8,7 +8,6 @@ import {
|
|
|
8
8
|
getContextWindow,
|
|
9
9
|
getDefaultModel,
|
|
10
10
|
getDefaultThinkingLevel,
|
|
11
|
-
getFastModel,
|
|
12
11
|
getMaxThinkingLevel,
|
|
13
12
|
getModel,
|
|
14
13
|
getModelsForProvider,
|
|
@@ -17,7 +16,7 @@ import {
|
|
|
17
16
|
getVideoByteLimit,
|
|
18
17
|
registerRuntimeModels,
|
|
19
18
|
usesOpenAICodexTransport
|
|
20
|
-
} from "./chunk-
|
|
19
|
+
} from "./chunk-N73LFQ3N.js";
|
|
21
20
|
import "./chunk-CRU3SSNX.js";
|
|
22
21
|
export {
|
|
23
22
|
DEFAULT_MAX_VIDEO_BYTES,
|
|
@@ -29,7 +28,6 @@ export {
|
|
|
29
28
|
getContextWindow,
|
|
30
29
|
getDefaultModel,
|
|
31
30
|
getDefaultThinkingLevel,
|
|
32
|
-
getFastModel,
|
|
33
31
|
getMaxThinkingLevel,
|
|
34
32
|
getModel,
|
|
35
33
|
getModelsForProvider,
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@prestyj/core",
|
|
3
|
-
"version": "5.
|
|
3
|
+
"version": "5.29.1",
|
|
4
4
|
"type": "module",
|
|
5
5
|
"description": "Provider-agnostic, UI-free shared foundation: model registry, auth, paths, telegram, voice",
|
|
6
6
|
"license": "MIT",
|
|
@@ -33,7 +33,7 @@
|
|
|
33
33
|
"dist"
|
|
34
34
|
],
|
|
35
35
|
"dependencies": {
|
|
36
|
-
"@prestyj/ai": "5.
|
|
36
|
+
"@prestyj/ai": "5.29.1"
|
|
37
37
|
},
|
|
38
38
|
"optionalDependencies": {
|
|
39
39
|
"@huggingface/transformers": "4.2.0",
|