@oh-my-pi/pi-catalog 17.2.9 → 17.2.10

package/CHANGELOG.md

@@ -2,6 +2,16 @@
 
 ## [Unreleased]
 
+## [17.2.10] - 2026-08-06
+
+### Changed
+
+- Removed the zod dependency by migrating GitLab Duo Workflow discovery schemas to omptype.
+
+### Fixed
+
+- Corrected thinking-effort tiers for deepseek-v4-flash to include the low tier alongside high and max.
+
 ## [17.2.9] - 2026-08-05
 
 ### Fixed

package/dist/types/identity/family.d.ts

@@ -34,6 +34,13 @@ export declare const isQwenModelId: (modelId: string) => boolean;
 export declare const isGemmaModelId: (modelId: string) => boolean;
 /** DeepSeek family by id or display name (proxies often rename the id but keep the name). */
 export declare const isDeepseekModelIdOrName: (modelId: string) => boolean;
+/**
+ * DeepSeek V4 Flash SKU in any host/namespace form (`deepseek-v4-flash`, dated
+ * `deepseek-v4-flash-0731`, `deepseek-ai/DeepSeek-V4-Flash`). Flash is the only
+ * V4 model whose `reasoning_effort` accepts the `low` tier; V4 Pro tops out at
+ * `high`/`max`. See https://api-docs.deepseek.com/api/create-chat-completion.
+ */
+export declare const isDeepseekV4FlashModelId: (modelId: string) => boolean;
 /** Xiaomi MiMo family by id or display name. */
 export declare const isMimoModelIdOrName: (modelId: string) => boolean;
 /**

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "17.2.9",
+	"version": "17.2.10",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -35,12 +35,11 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.1",
-		"@oh-my-pi/omptype": "17.2.9",
-		"@oh-my-pi/pi-utils": "17.2.9",
-		"zod": "^4"
+		"@oh-my-pi/omptype": "17.2.10",
+		"@oh-my-pi/pi-utils": "17.2.10"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "17.2.9",
+		"@oh-my-pi/pi-ai": "17.2.10",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/discovery/gitlab-duo-workflow.ts

@@ -1,6 +1,6 @@
 import * as fs from "node:fs/promises";
 import * as path from "node:path";
-import { z } from "zod/v4";
+import { type } from "@oh-my-pi/omptype";
 import type { FetchImpl, ModelSpec } from "../types";
 import { discoveryFetch, isRecord } from "../utils";
 
@@ -54,20 +54,28 @@ const ProjectRootNamespaceQuery = `query omp_gitlabDuoWorkflowProjectRootNamespa
   }
 }`;
 
-const modelRefSchema = z
-	.object({
-		name: z.string().optional().catch(undefined),
-		ref: z.string().optional().catch(undefined),
-	})
-	.loose();
-
-const aiChatAvailableModelsSchema = z
-	.object({
-		defaultModel: z.unknown().nullable().optional(),
-		selectableModels: z.array(z.unknown()).nullable().optional().catch([]),
-		pinnedModel: z.unknown().nullable().optional(),
-	})
-	.loose();
+const resilientString = type("unknown").pipe(value => {
+	if (value === undefined) return undefined;
+	const parsed = type("string")(value);
+	return parsed instanceof type.errors ? undefined : parsed;
+});
+
+const resilientUnknownArray = type("unknown").pipe(value => {
+	if (value === undefined || value === null) return value;
+	const parsed = type("unknown[]")(value);
+	return parsed instanceof type.errors ? [] : parsed;
+});
+
+const modelRefSchema = type({
+	"name?": resilientString,
+	"ref?": resilientString,
+});
+
+const aiChatAvailableModelsSchema = type({
+	"defaultModel?": "unknown",
+	"selectableModels?": resilientUnknownArray,
+	"pinnedModel?": "unknown",
+});
 
 type GitLabDuoWorkflowCandidateSource = "override" | "project" | "remote" | "group";
 
@@ -540,17 +548,15 @@ async function postGraphQL(
 }
 
 function parseAvailability(value: unknown): GitLabDuoWorkflowAvailability | null {
-	const parsed = aiChatAvailableModelsSchema.safeParse(value);
-	if (!parsed.success) {
-		return null;
-	}
+	const parsed = aiChatAvailableModelsSchema(value);
+	if (parsed instanceof type.errors) return null;
 	return {
-		defaultModel: parseModelRef(parsed.data.defaultModel),
-		selectableModels: (parsed.data.selectableModels ?? []).flatMap(model => {
+		defaultModel: parseModelRef(parsed.defaultModel),
+		selectableModels: (parsed.selectableModels ?? []).flatMap(model => {
 			const parsedModel = parseModelRef(model);
 			return parsedModel ? [parsedModel] : [];
 		}),
-		pinnedModel: parseModelRef(parsed.data.pinnedModel),
+		pinnedModel: parseModelRef(parsed.pinnedModel),
 	};
 }
 
@@ -558,15 +564,13 @@ function parseModelRef(value: unknown): GitLabDuoWorkflowModelRef | null {
 	if (value === null || value === undefined) {
 		return null;
 	}
-	const parsed = modelRefSchema.safeParse(value);
-	if (!parsed.success) {
-		return null;
-	}
-	const ref = normalizeIdentifier(parsed.data.ref);
+	const parsed = modelRefSchema(value);
+	if (parsed instanceof type.errors) return null;
+	const ref = normalizeIdentifier(parsed.ref);
 	if (!ref) {
 		return null;
 	}
-	const name = normalizeIdentifier(parsed.data.name) ?? ref;
+	const name = normalizeIdentifier(parsed.name) ?? ref;
 	return { name, ref };
 }
 

package/src/identity/family.ts

@@ -83,6 +83,16 @@ export const isDeepseekModelIdOrName = memo((value: string): boolean => {
 	return value.toLowerCase().includes("deepseek");
 });
 
+/**
+ * DeepSeek V4 Flash SKU in any host/namespace form (`deepseek-v4-flash`, dated
+ * `deepseek-v4-flash-0731`, `deepseek-ai/DeepSeek-V4-Flash`). Flash is the only
+ * V4 model whose `reasoning_effort` accepts the `low` tier; V4 Pro tops out at
+ * `high`/`max`. See https://api-docs.deepseek.com/api/create-chat-completion.
+ */
+export const isDeepseekV4FlashModelId = memo((modelId: string): boolean => {
+	return bareModelId(modelId).toLowerCase().includes("deepseek-v4-flash");
+});
+
 /** Xiaomi MiMo family by id or display name. */
 export const isMimoModelIdOrName = memo((value: string): boolean => {
 	return value.toLowerCase().includes("mimo");

package/src/model-thinking.ts

@@ -24,6 +24,7 @@ import {
 import {
 	findThinkingVariantToken,
 	isDeepseekModelIdOrName,
+	isDeepseekV4FlashModelId,
 	isGlm52ReasoningEffortModelId,
 	isKimiK3ModelId,
 	isMimoModelIdOrName,
@@ -62,9 +63,9 @@ const GEMINI_3_FLASH_EFFORTS: readonly Effort[] = [Effort.Minimal, Effort.Low, E
 const GPT_5_2_PLUS_EFFORTS: readonly Effort[] = [Effort.Low, Effort.Medium, Effort.High, Effort.XHigh];
 const GPT_5_1_CODEX_MINI_EFFORTS: readonly Effort[] = [Effort.Medium, Effort.High];
 const LOW_MEDIUM_HIGH_REASONING_EFFORTS: readonly Effort[] = [Effort.Low, Effort.Medium, Effort.High];
-/** Wire-exact `low`/`high`/`max` scale used by Kimi K3 and OpenRouter DeepSeek V4 Flash 0731. */
+/** Wire-exact `low`/`high`/`max` scale used by Kimi K3 and DeepSeek V4 Flash (direct API and aggregators). */
 const LOW_HIGH_MAX_REASONING_EFFORTS: readonly Effort[] = [Effort.Low, Effort.High, Effort.Max];
-/** Wire-exact two-tier scale (`high`/`max`): GLM-5.2 on Z.ai/Umans/Ollama Cloud/Baseten, Sakana Fugu, DeepSeek. */
+/** Wire-exact two-tier scale (`high`/`max`): GLM-5.2 on Z.ai/Umans/Ollama Cloud/Baseten, Sakana Fugu, DeepSeek V4 Pro. */
 const HIGH_MAX_REASONING_EFFORTS: readonly Effort[] = [Effort.High, Effort.Max];
 /** OpenRouter's DeepSeek route accepts only `high`. */
 const HIGH_ONLY_REASONING_EFFORTS: readonly Effort[] = [Effort.High];
@@ -366,14 +367,14 @@ function getModelDefinedEfforts<TApi extends Api>(
 		return OLLAMA_REASONING_EFFORTS;
 	}
 	if (isOpenAICompatReasoningApi(spec.api) && isDeepseekReasoningModel(spec)) {
-		// OpenRouter generally exposes only high for DeepSeek, but V4 Flash 0731
-		// advertises and accepts the wire-exact low/high/max ladder.
-		if (isOpenRouterThinkingFormat(compat)) {
-			return bareModelId(spec.id) === "deepseek-v4-flash-0731"
-				? LOW_HIGH_MAX_REASONING_EFFORTS
-				: HIGH_ONLY_REASONING_EFFORTS;
+		// DeepSeek V4 Flash accepts the wire-exact low/high/max ladder on every
+		// host — the direct API and aggregators alike (medium/xhigh map to
+		// high). V4 Pro and the older reasoners top out at high/max, and
+		// OpenRouter's non-flash DeepSeek route exposes only high.
+		if (isDeepseekV4FlashModelId(spec.id)) {
+			return LOW_HIGH_MAX_REASONING_EFFORTS;
 		}
-		return HIGH_MAX_REASONING_EFFORTS;
+		return isOpenRouterThinkingFormat(compat) ? HIGH_ONLY_REASONING_EFFORTS : HIGH_MAX_REASONING_EFFORTS;
 	}
 	if (spec.provider === "baseten" && isOpenAIGptOssModelId(spec.id)) {
 		// Baseten's gpt-oss router mirrors its GLM route: high/max only.