auto-model-router 0.3.4 → 0.4.1

package/test/turn.test.ts

@@ -29,7 +29,7 @@ import type {
 
 function mkConfig(escalation: Partial<EscalationConfig> = {}): RouterConfig {
 	return {
-		server: { host: "127.0.0.1", port: 8787, maxConcurrentTurns: 24 },
+		server: { host: "127.0.0.1", port: 8787, maxConcurrentTurns: 24, subagentProfile: "auto-sub" },
 		openrouter: { baseUrl: "https://openrouter.ai/api/v1", apiKey: "", title: "test", timeoutMs: 30_000, catalogTtlMs: 3_600_000, catalogRefreshMs: 0 },
 		ollama: { enabled: false, baseUrl: "http://127.0.0.1:11434/v1", apiKey: "", timeoutMs: 30_000, catalogTtlMs: 300_000, includeLocal: false, prices: {}, twins: {}, costBias: 1, biasUntilUsage: 0.9, usagePollMs: 0, quotaCooldownMs: 0, rateLimitCooldownMs: 0, planCreditsUsd: 0 },
 		benchmarks: { enabled: false, artificialAnalysisApiKey: "", benchlm: true, refreshMs: 86_400_000, timeoutMs: 30_000, useLocalScores: false },
@@ -46,10 +46,10 @@ function mkConfig(escalation: Partial<EscalationConfig> = {}): RouterConfig {
 			data: { axis: "intelligence", minQuality: 0 },
 			chat: { axis: "intelligence", minQuality: 0 },
 		},
-		filters: { allow: [], deny: [], includeFree: false, requireToolSupport: true, minTrust: 0.6, minTrustSamples: 5, trustScopedByHarness: false, trustWindowDays: 0, contextHeadroom: 1.2, latencyWeight: 0, latencyReferenceMs: 5000, latencyReferenceTokensPerSec: 30, latencyMinSamples: 20, escalationCostWeight: 0 },
+		filters: { allow: [], deny: [], includeFree: false, requireToolSupport: true, minTrust: 0.6, feedbackWeight: 0, minTrustSamples: 5, trustScopedByHarness: false, trustWindowDays: 0, contextHeadroom: 1.2, latencyWeight: 0, latencyReferenceMs: 5000, latencyReferenceTokensPerSec: 30, cacheReliabilityMinSamples: 10, latencyMinSamples: 20, escalationCostWeight: 0 },
 		classifier: {
 			ambiguityThreshold: 0,
-			model: "test/adjudicator",
+			model: "test/adjudicator", learnedModelPath: "",
 			maxCostFraction: 0.1,
 			maxCostUsd: 0.01,
 			timeoutMs: 5000,
@@ -57,7 +57,7 @@ function mkConfig(escalation: Partial<EscalationConfig> = {}): RouterConfig {
 			toolAxis: "coding",
 			chatAxis: "intelligence",
 			agenticLoopDepth: 3,
-			mechanicalRetryFactor: 0.2,
+			mechanicalRetryFactor: 0.2, readOnlyToolWeight: 0,
 			reasoningWeights: { medium: 0.14, high: 0.24, xhigh: 0.3, max: 0.34 },
 		},
 		escalation: {
@@ -76,6 +76,7 @@ function mkConfig(escalation: Partial<EscalationConfig> = {}): RouterConfig {
 		context: { enabled: false, baseUrl: "", token: "", defaultScope: "", timeoutMs: 3_000, maxStalenessMs: 900_000, maxBlockChars: 24_000, memoryLimit: 8, docsLimit: 2, sessionLimit: 6, briefChars: 0, recordTurns: false, maxQueue: 64 },
 		compaction: { enabled: false, budgetTokens: 40_000, floorRatio: 1, fitToWindow: true, protectRecentTurns: 4, maxToolResultBytes: 4_096, keepHeadBytes: 512, keepTailBytes: 512, elideSupersededReads: true, collapseDuplicateResults: true, replanGrowthRatio: 1 },
 		budget: { onExceeded: "downgrade" },
+		report: { baselines: [] },
 		profiles: [],
 		ledger: { path: ":memory:", blendWindowDays: 7, blendMinSamples: 20, fallbackBlend: { inputPerMtok: 1, outputPerMtok: 4 }, conversationTtlMs: 86_400_000 },
 		adaptiveTierFloors: true,
@@ -91,6 +92,7 @@ function mkReq(): NormRequest {
 		harnessId: "",
 		ompSessionId: "",
 		agentdoxScope: "",
+		isSubagent: false,
 		requestedModel: "auto",
 		messages: [{ role: "user", text: "hi", images: 0, textBytes: 2, toolCalls: [] }],
 		tools: [],
@@ -896,4 +898,34 @@ describe("latency measurement covers the work the router actually does", () => {
 		expect(map.get("conv-test")!.cacheWarmSlug).toBe(ollamaModel.slug);
 	});
 
+	test("an Ollama estimate is scaled by the ledger-vs-meter calibration", async () => {
+		const ollamaModel: CatalogModel = {
+			slug: "ollama/glm-5.3-flash",
+			provider: "ollama",
+			canonicalSlug: "ollama/glm-5.3-flash",
+			name: "glm",
+			contextLength: 1_000_000,
+			supportsTools: true,
+			supportsReasoning: true,
+			reasoningMandatory: false,
+			supportsToolChoice: false,
+			inputModalities: ["text"],
+			price: { prompt: 0.15 / 1e6, cacheRead: 0.03 / 1e6, completion: 0.5 / 1e6 },
+			priceTiers: [],
+			quality: {},
+			tokenizer: "Other",
+			isFree: false,
+			createdAtMs: 0,
+			author: "ollama",
+		};
+		const priced = { ...catalog, find: (slug: string) => (slug === ollamaModel.slug ? ollamaModel : undefined) };
+		const { router } = mkRouter([mkDecision("simple", ollamaModel.slug)]);
+		const { upstream } = mkUpstream([{ kind: "chunks", chunks: [startChunk(ollamaModel.slug), textChunk("ok"), finishChunk("stop"), usageChunk({ promptTokens: 100_000, completionTokens: 0 }, null)] }]);
+		const { ledger, entries } = mkLedger();
+		const { store } = mkConversations();
+		await runTurn(mkReq(), mkSink().sink, { config: mkConfig(), router, upstream, ledger, conversations: store, catalog: priced, context: createDisabledBridge(), ollamaCostScale: () => 1.25 }, new AbortController().signal);
+		// 100k × $0.15/M = $0.015, scaled ×1.25.
+		expect(entries[0]!.reportedUsd).toBeCloseTo(0.01875, 6);
+	});
+
 });