@oh-my-pi/pi-catalog 18.3.5 → 18.4.1

package/CHANGELOG.md

@@ -2,6 +2,26 @@
 
 ## [Unreleased]
 
+## [18.4.1] - 2026-09-28
+
+### Added
+
+- Added `resolveCatalogAxes` to `compat/resolve`, which resolves a model's catalog-axis policy without computing its full compat and thinking policy; context-window and catalog-policy lookups now use it, reducing model-catalog load time.
+
+### Fixed
+
+- Fixed `--thinking xhigh` on Amazon Bedrock Grok 4.6 (`us.xai.grok-4.6`, `global.xai.grok-4.6`, `xai.grok-4.6`), which was silently lowered to `high` ([#13515](https://github.com/can1357/oh-my-pi/pull/13515) by [@pgkt04](https://github.com/pgkt04)).
+- Fixed `--thinking xhigh` and `--thinking max` on Amazon Bedrock Claude Opus 5.5, which were silently lowered to `high` ([#13515](https://github.com/can1357/oh-my-pi/pull/13515) by [@pgkt04](https://github.com/pgkt04)).
+- Fixed GitHub Copilot GPT-5.6 and GPT-6 Astra starting on the 1.05M premium context window instead of the default tier ([#13017](https://github.com/can1357/oh-my-pi/pull/13017)).
+- Added Cursor model pricing so usage stats no longer record zero cost for Cursor turns ([#13302](https://github.com/can1357/oh-my-pi/pull/13302) by [@eggpeat](https://github.com/eggpeat)).
+- Devin requests now identify as Devin CLI 3000.11.3 instead of 3000.6.2, the CLI release that routes Fusion pairings ([#13527](https://github.com/can1357/oh-my-pi/pull/13527) by [@will-bogusz](https://github.com/will-bogusz)).
+
+## [18.4.0] - 2026-09-28
+
+### Fixed
+
+- Corrected Cerebras model pricing so cached input tokens are charged at the model’s input rate, consistent with Cerebras billing.
+
 ## [18.3.5] - 2026-09-27
 
 ### Added

package/dist/types/build.d.ts

@@ -9,8 +9,8 @@
 import { type Api, type Model, type ModelSpec } from "./types.js";
 /**
  * Applies reviewed catalog-data value corrections (`cost-patch`,
- * `limits-patch`, `long-context-cost`, `context-window-floor`,
- * `input-modalities`) onto an upstream-sourced spec. Applied by
+ * `cache-read-at-input-rate`, `limits-patch`, `long-context-cost`,
+ * `context-window-floor`, `input-modalities`) onto an upstream-sourced spec. Applied by
  * `buildModel` to every upstream-sourced spec; user-authored overrides are
  * recomposed after building by the override applicators, so explicit user
  * limits and pricing still win.

package/dist/types/compat/resolve.d.ts

@@ -10,6 +10,12 @@ export interface ResolvedModelPolicy<TApi extends Api = Api> {
 }
 /** Resolve the request adapter assigned to a discovery backend before materialization. */
 export declare function resolveDiscoveryApi(spec: ModelSpec<Api>, providerType: string): Api;
+/**
+ * Catalog-data axis assignments for one model spec — the `catalog` field of
+ * {@link resolveModelPolicy} without resolving the compat and thinking
+ * policies, which dominate its cost on catalog-wide scans.
+ */
+export declare function resolveCatalogAxes(spec: ModelSpec<Api>): Record<string, unknown>;
 /**
  * Resolves the full policy surface for one model spec: structured identity,
  * complete compat record, thinking metadata, and catalog-data corrections.

package/dist/types/compat/types.d.ts

@@ -240,6 +240,8 @@ export interface CompiledCursorParameter {
 /** One provider quota-scope table. */
 export interface CompiledQuotaRule {
     provider: string;
+    /** Tier used when no exact or fallback membership matches. */
+    defaultTier?: string;
     tiers: {
         label: string;
         models: string[];

package/dist/types/discovery/cursor-proto.d.ts

@@ -5070,6 +5070,11 @@ export interface TruncatedToolCallSuccess extends ProtoMessage {
 export declare const TruncatedToolCallSuccessSchema: MessageCodec<TruncatedToolCallSuccess>;
 /** Cursor agent message agent.v1.TurnEndedUpdate. */
 export interface TurnEndedUpdate extends ProtoMessage {
+    inputTokens?: bigint;
+    outputTokens?: bigint;
+    cacheReadTokens?: bigint;
+    cacheWriteTokens?: bigint;
+    reasoningTokens?: bigint;
 }
 export declare const TurnEndedUpdateSchema: MessageCodec<TurnEndedUpdate>;
 /** Cursor agent message agent.v1.TypeAction. */

package/dist/types/wire/devin.d.ts

@@ -8,17 +8,28 @@
 export declare const DEVIN_DEFAULT_BASE_URL = "https://server.codeium.com";
 /** Session token as the wire format carries it: the scheme prefix is required. */
 export declare function normalizeDevinSessionToken(apiKey: string | undefined): string;
+/** Released-CLI metadata with credential bytes already encoded for the wire. */
+export declare function devinWireMetadata(apiKey: string | undefined, userJwt?: string): {
+    ideName: "devin-cli";
+    ideType: "chisel";
+    ideVersion: "3000.11.3";
+    extensionName: "chisel";
+    extensionVersion: "3000.11.3";
+    locale: "en";
+    os: "darwin" | "linux" | "windows";
+    apiKey: string;
+    userJwt: string;
+};
 /**
  * Fields for `Metadata` on released-CLI calls (`GetUserJwt`, `AssignModel`,
- * `GetChatMessage`, `GetUserStatus`). `userJwt` stays empty for the calls the
- * CLI makes with the session token alone (auth, model assignment, usage).
+ * `GetChatMessage`, `GetUserStatus`) authenticated by a Devin session token.
  */
 export declare function devinCliMetadata(apiKey: string | undefined, userJwt?: string): {
     ideName: "devin-cli";
     ideType: "chisel";
-    ideVersion: "3000.6.2";
+    ideVersion: "3000.11.3";
     extensionName: "chisel";
-    extensionVersion: "3000.6.2";
+    extensionVersion: "3000.11.3";
     locale: "en";
     os: "darwin" | "linux" | "windows";
     apiKey: string;

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.3.5",
+	"version": "18.4.1",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"keywords": [
 		"ai",
@@ -93,12 +93,12 @@
 		"gen:proto": "bun scripts/generate-protocols.ts"
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.3.5",
-		"@oh-my-pi/pi-utils": "18.3.5"
+		"@oh-my-pi/omptype": "18.4.1",
+		"@oh-my-pi/pi-utils": "18.4.1"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.3.5",
+		"@oh-my-pi/pi-ai": "18.4.1",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/build.ts

@@ -140,8 +140,8 @@ function applyCatalogAssignments<TApi extends Api>(model: Model<TApi>, catalog:
 
 /**
  * Applies reviewed catalog-data value corrections (`cost-patch`,
- * `limits-patch`, `long-context-cost`, `context-window-floor`,
- * `input-modalities`) onto an upstream-sourced spec. Applied by
+ * `cache-read-at-input-rate`, `limits-patch`, `long-context-cost`,
+ * `context-window-floor`, `input-modalities`) onto an upstream-sourced spec. Applied by
  * `buildModel` to every upstream-sourced spec; user-authored overrides are
  * recomposed after building by the override applicators, so explicit user
  * limits and pricing still win.
@@ -220,6 +220,14 @@ export function applyCatalogCorrections(
 			applyEffectiveFallbackRates(model.cost, Reflect.get(fallback, "effectiveRates"));
 		}
 	}
+	if (catalog.cacheReadAtInputRate === true) {
+		const { longContext } = model.cost;
+		model.cost = {
+			...model.cost,
+			cacheRead: model.cost.input,
+			...(longContext && { longContext: { ...longContext, cacheRead: longContext.input } }),
+		};
+	}
 	if (catalog.timeBased !== undefined) {
 		model.cost = { ...model.cost, timeBased: materializeTimeBasedCost(catalog.timeBased) };
 	}

package/src/compat/axes.ts

@@ -321,6 +321,16 @@ export const AXES: Readonly<Record<string, AxisDef>> = {
 	"context-window-floor": { key: "contextWindowFloor", set: "catalog", shape: "scalar" },
 	"cost-patch": { key: "costPatch", set: "catalog", shape: "object" },
 	"cost-fallback": { key: "costFallback", set: "catalog", shape: "object" },
+	/**
+	 * The host bills cache-hit input tokens at the full input rate (no cache
+	 * discount), so the built row's `cacheRead` tracks its live `input` price.
+	 */
+	"cache-read-at-input-rate": {
+		key: "cacheReadAtInputRate",
+		set: "catalog",
+		shape: "scalar",
+		values: [true, false],
+	},
 	"delegation-bias": { key: "delegationBias", set: "catalog", shape: "scalar", values: DELEGATION_BIASES },
 	"discovery-api": { key: "discoveryApi", set: "catalog", shape: "scalar" },
 	"edit-prompt-variant": { key: "editPromptVariant", set: "catalog", shape: "scalar", values: ["full", "compact"] },

package/src/compat/behavior.ts

@@ -100,7 +100,7 @@ export function quotaTierFor(provider: string, model: string): string | undefine
 	for (const fallback of rule.fallbacks) {
 		if (model.includes(fallback.substring)) return fallback.label;
 	}
-	return undefined;
+	return rule.defaultTier;
 }
 
 /** UTC offset for a provider's timezone-naive absolute retry-reset timestamps. */

package/src/compat/catalog-policy.ts

@@ -1,6 +1,6 @@
 import { toModelSpec } from "../provider-models/bundled-references";
 import type { Model } from "../types";
-import { resolveModelPolicy } from "./resolve";
+import { resolveCatalogAxes } from "./resolve";
 
 /**
  * Rule-owned catalog-axis policy by model. Resolve once per process rather
@@ -21,7 +21,7 @@ export function resolveCatalogPolicy(model: Model): Readonly<Record<string, unkn
 	const key = `${model.provider}\u0000${model.providerType ?? ""}\u0000${model.api}\u0000${model.id}\u0000${model.reasoning ? 1 : 0}`;
 	const cached = catalogPolicyCache.get(key);
 	if (cached !== undefined) return cached;
-	const policy = Object.freeze({ ...resolveModelPolicy(toModelSpec(model)).catalog });
+	const policy = Object.freeze({ ...resolveCatalogAxes(toModelSpec(model)) });
 	if (catalogPolicyCache.size >= CATALOG_POLICY_CACHE_MAX) catalogPolicyCache.clear();
 	catalogPolicyCache.set(key, policy);
 	return policy;

package/src/compat/context-window.ts

@@ -1,6 +1,6 @@
 import { toModelSpec } from "../provider-models/bundled-references";
 import type { Model } from "../types";
-import { resolveModelPolicy } from "./resolve";
+import { resolveCatalogAxes } from "./resolve";
 
 /**
  * Rule-owned maxima by provider/id/api. Resolve once per process rather than
@@ -21,7 +21,7 @@ export function resolveMaxContextWindow(model: Model): number | undefined {
 	const key = `${model.provider} ${model.id} ${model.api}`;
 	let curated = ruleMaximumCache.get(key);
 	if (curated === undefined) {
-		const maximum = resolveModelPolicy(toModelSpec(model)).catalog.maxContextWindow;
+		const maximum = resolveCatalogAxes(toModelSpec(model)).maxContextWindow;
 		curated = typeof maximum === "number" && Number.isFinite(maximum) && maximum > 0 ? maximum : null;
 		if (ruleMaximumCache.size >= RULE_POLICY_CACHE_MAX) ruleMaximumCache.clear();
 		ruleMaximumCache.set(key, curated);
@@ -56,7 +56,7 @@ export function clampsContextOverride(model: Model): boolean {
 	const key = `${model.provider} ${model.id} ${model.api}`;
 	const cached = clampOverrideCache.get(key);
 	if (cached !== undefined) return cached;
-	const clamps = resolveModelPolicy(toModelSpec(model)).catalog.clampContextOverride === true;
+	const clamps = resolveCatalogAxes(toModelSpec(model)).clampContextOverride === true;
 	if (clampOverrideCache.size >= RULE_POLICY_CACHE_MAX) clampOverrideCache.clear();
 	clampOverrideCache.set(key, clamps);
 	return clamps;

package/src/compat/resolve.ts

@@ -1241,6 +1241,15 @@ export function resolveDiscoveryApi(spec: ModelSpec<Api>, providerType: string):
 	return typeof discoveryApi === "string" ? discoveryApi : spec.api;
 }
 
+/**
+ * Catalog-data axis assignments for one model spec — the `catalog` field of
+ * {@link resolveModelPolicy} without resolving the compat and thinking
+ * policies, which dominate its cost on catalog-wide scans.
+ */
+export function resolveCatalogAxes(spec: ModelSpec<Api>): Record<string, unknown> {
+	return resolveCascade(buildResolveTarget(spec, resolveIdentity(spec))).catalog;
+}
+
 /**
  * Resolves the full policy surface for one model spec: structured identity,
  * complete compat record, thinking metadata, and catalog-data corrections.

package/src/compat/rules/providers/amazon-bedrock.kdl

@@ -7,11 +7,6 @@ provider "amazon-bedrock" {
 	// (`amazon-bedrock.ts`), so a block can land above text already rendered.
 	stream-revision "possible"
 	class "anthropic" {
-		family "fable" {
-			revision ">=5.1 <5.2" {
-				thinking-efforts "low" "medium" "high" "xhigh" "max"
-			}
-		}
 		family "opus" {
 			revision ">=4.6 <4.7" {
 				thinking-mode "anthropic-adaptive"
@@ -22,6 +17,22 @@ provider "amazon-bedrock" {
 			stream-idle-timeout-ms 900000
 			thinking-efforts "low" "medium" "high" "max"
 		}
+		// First-party >=4.7 xhigh/max omits amazon-bedrock: the host rule
+		// above is a shorter ladder. From 5.1 the host accepts the full set.
+		revision ">=5.1" {
+			thinking-efforts "low" "medium" "high" "xhigh" "max"
+		}
+	}
+	class "xai" {
+		family "grok" {
+			// Converse accepts reasoning.effort for Grok 4.6; budget_tokens is
+			// the Anthropic path and the unknown-class default. Older Groks
+			// are unverified and keep budget.
+			revision ">=4.6" {
+				thinking-mode "effort"
+				thinking-efforts "low" "medium" "high" "xhigh"
+			}
+		}
 	}
 	class "openai" {
 		thinking-mode "effort"

package/src/compat/rules/providers/cerebras.kdl

@@ -10,6 +10,9 @@ provider "cerebras" {
 	// Replaces the Cerebras provider branches in the completions baseline.
 	supports-usage-in-streaming #false
 	tool-strict-mode "all_strict"
+	// Prompt caching carries no discount: cached input tokens bill at the
+	// standard input rate (https://inference-docs.cerebras.ai/capabilities/prompt-caching).
+	cache-read-at-input-rate #true
 	// Cerebras serves Gemma 4 with vision enabled; its /v1/models rows carry
 	// no modality metadata.
 	models "gemma-4-31b" {