@caupulican/pi-adaptative 0.81.20 → 0.81.21
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/CHANGELOG.md +15 -0
- package/dist/core/agent-session.d.ts.map +1 -1
- package/dist/core/agent-session.js +17 -0
- package/dist/core/agent-session.js.map +1 -1
- package/dist/core/doctor.d.ts.map +1 -1
- package/dist/core/doctor.js +4 -1
- package/dist/core/doctor.js.map +1 -1
- package/dist/core/local-runtime-controller.d.ts +2 -0
- package/dist/core/local-runtime-controller.d.ts.map +1 -1
- package/dist/core/local-runtime-controller.js +92 -6
- package/dist/core/local-runtime-controller.js.map +1 -1
- package/dist/core/models/context-sizing.d.ts.map +1 -1
- package/dist/core/models/context-sizing.js +18 -26
- package/dist/core/models/context-sizing.js.map +1 -1
- package/dist/core/models/local-runtime.d.ts +56 -4
- package/dist/core/models/local-runtime.d.ts.map +1 -1
- package/dist/core/models/local-runtime.js +286 -19
- package/dist/core/models/local-runtime.js.map +1 -1
- package/dist/core/models/perf-profile.d.ts +3 -0
- package/dist/core/models/perf-profile.d.ts.map +1 -1
- package/dist/core/models/perf-profile.js +42 -13
- package/dist/core/models/perf-profile.js.map +1 -1
- package/dist/core/models/runtime-arbiter.d.ts +45 -0
- package/dist/core/models/runtime-arbiter.d.ts.map +1 -1
- package/dist/core/models/runtime-arbiter.js +115 -0
- package/dist/core/models/runtime-arbiter.js.map +1 -1
- package/dist/modes/interactive/local-model-commands.d.ts.map +1 -1
- package/dist/modes/interactive/local-model-commands.js +24 -6
- package/dist/modes/interactive/local-model-commands.js.map +1 -1
- package/examples/extensions/custom-provider-anthropic/package-lock.json +2 -2
- package/examples/extensions/custom-provider-anthropic/package.json +1 -1
- package/examples/extensions/custom-provider-gitlab-duo/package.json +1 -1
- package/examples/extensions/sandbox/package-lock.json +2 -2
- package/examples/extensions/sandbox/package.json +1 -1
- package/examples/extensions/with-deps/package-lock.json +2 -2
- package/examples/extensions/with-deps/package.json +1 -1
- package/npm-shrinkwrap.json +12 -12
- package/package.json +4 -4
|
@@ -31,6 +31,7 @@ import { deriveModelCapabilityProfile, filterToolNamesForCapability, } from "./m
|
|
|
31
31
|
import { formatModelRouterModel, ModelRouterController } from "./model-router-controller.js";
|
|
32
32
|
import { ModelSelectionController } from "./model-selection-controller.js";
|
|
33
33
|
import { ModelAdaptationStore } from "./models/adaptation-store.js";
|
|
34
|
+
import { HF_TRANSFORMERS_PROVIDER, OLLAMA_PROVIDER } from "./models/local-registration.js";
|
|
34
35
|
import { DEFAULT_ADAPTIVE_STREAM_IDLE_CEILING_MS, estimateContextPromptTokens, resolveAdaptiveStreamIdleOptions, withModelPerfProfile, } from "./models/perf-profile.js";
|
|
35
36
|
import { ProfileFilterController } from "./profile-filter-controller.js";
|
|
36
37
|
import { expandPromptTemplate } from "./prompt-templates.js";
|
|
@@ -266,6 +267,7 @@ export class AgentSession {
|
|
|
266
267
|
base: { ...DEFAULT_STREAM_IDLE, ...configured },
|
|
267
268
|
profile,
|
|
268
269
|
promptTokens: estimateContextPromptTokens(context),
|
|
270
|
+
localClass: this._isWarmableLocalModel(model),
|
|
269
271
|
ceilingMs: httpIdleTimeoutMs === 0
|
|
270
272
|
? DEFAULT_ADAPTIVE_STREAM_IDLE_CEILING_MS
|
|
271
273
|
: Math.max(DEFAULT_STREAM_IDLE.quietIdleMs, httpIdleTimeoutMs - 60_000),
|
|
@@ -659,6 +661,21 @@ export class AgentSession {
|
|
|
659
661
|
options.apiKey = auth.apiKey;
|
|
660
662
|
options.headers = auth.headers;
|
|
661
663
|
}
|
|
664
|
+
if (controller.signal.aborted)
|
|
665
|
+
return;
|
|
666
|
+
if (model.provider === OLLAMA_PROVIDER || model.provider === HF_TRANSFORMERS_PROVIDER) {
|
|
667
|
+
await this._localRuntimeController.ensureRouteModelReady({
|
|
668
|
+
decision: {
|
|
669
|
+
tier: "cheap",
|
|
670
|
+
risk: "read-only",
|
|
671
|
+
confidence: 1,
|
|
672
|
+
reasonCode: "prefix_warmer",
|
|
673
|
+
reasons: ["prefix warmer residency"],
|
|
674
|
+
model: `${model.provider}/${model.id}`,
|
|
675
|
+
},
|
|
676
|
+
model,
|
|
677
|
+
});
|
|
678
|
+
}
|
|
662
679
|
if (controller.signal.aborted)
|
|
663
680
|
return;
|
|
664
681
|
const stream = await this.agent.streamFn(model, {
|