@bitkyc08/opencodex 2.91.0 → 2.92.0

package/src/providers/quota-decision-snapshot.ts

@@ -3,12 +3,23 @@ import { createHash } from "node:crypto";
 import { getConfigDir } from "../config/paths";
 import { captureConfigGeneration } from "../lib/state-store-sweeper";
 import type { OcxProviderConfig } from "../types";
+import { invalidateAntigravityQuota, readAntigravityQuota } from "./quota-decision-antigravity";
+import type { DecisionKeyObservation } from "../lib/decision-quota-sink";
 
 export interface DecisionQuotaWindow {
   window: "5h" | "weekly" | "monthly" | "fable" | "opus" | "sonnet";
   percent: number;
   observedAt: number;
   resetAt?: number;
+  /** Model family this window meters (Antigravity Gemini vs Claude/3P buckets); absent means every model of the provider. */
+  family?: DecisionQuotaFamily;
+}
+export type DecisionQuotaFamily = "gem" | "cla";
+/** True when the provider row can carry sole-credential decision evidence under `name`. */
+function keyEvidenceEligible(name: string, provider: OcxProviderConfig): boolean {
+  if (provider.disabled || (provider.apiKeyPool?.length ?? 0) > 1) return false;
+  const mode = provider.authMode ?? "key";
+  return mode === "key";
 }
 export interface DecisionQuotaAccount {
   id: string;
@@ -54,18 +65,24 @@ function keyPolicy(provider: OcxProviderConfig): string {
   ])).digest("hex");
 }
 /** Drop all key-based decision evidence and advance the epoch so earlier publications cannot be read. */
-export function invalidateDecisionKeyQuotas(): void { keyEpoch++; keys.clear(); }
+export function invalidateDecisionKeyQuotas(): void { keyEpoch++; keys.clear(); invalidateAntigravityQuota(); }
+/** Collector replacement does not retire live OAuth producer receipts or their original clocks. */
+export function replaceDecisionKeyQuotas(rows: readonly DecisionKeyObservation[]): void {
+  keyEpoch++; keys.clear();
+  for (const row of rows) publishDecisionKeyQuota(row.name, row.provider, row.windows);
+}
 /** Publish decision windows for a single-key provider; pooled, disabled or non-key providers are never published, and the table is bounded. */
 export function publishDecisionKeyQuota(name: string, provider: OcxProviderConfig, windows: readonly DecisionQuotaWindow[]): void {
   writesForTests++;
-  if (name.length > 1024 || provider.disabled || (provider.authMode ?? "key") !== "key" || (provider.apiKeyPool?.length ?? 0) > 1) return;
+  if (name.length > 1024 || !keyEvidenceEligible(name, provider)) return;
   if (keys.size >= 256 && !keys.has(name)) keys.delete(keys.keys().next().value!);
   keys.set(name, { root: getConfigDir(), epoch: keyEpoch, configGeneration: captureConfigGeneration(), policy: keyPolicy(provider), windows: copyWindows(windows) });
 }
 /** Read key-based evidence only when the config root, epoch, generation and provider policy still match what was published. */
 export function readLoadedDecisionKeyQuota(name: string, provider: OcxProviderConfig): readonly DecisionQuotaWindow[] | undefined {
+  if (provider.authMode === "oauth") return readAntigravityQuota(name, provider);
   const row = keys.get(name);
-  if (!row || provider.disabled || (provider.authMode ?? "key") !== "key" || (provider.apiKeyPool?.length ?? 0) > 1
+  if (!row || !keyEvidenceEligible(name, provider)
     || row.root !== getConfigDir() || row.epoch !== keyEpoch || row.configGeneration !== captureConfigGeneration()
     || row.policy !== keyPolicy(provider)) return undefined;
   return row.windows.map(row => ({ ...row }));
@@ -141,5 +158,5 @@ export function clearDecisionAccountQuotas(provider?: string, id?: string): void
 /** Copy only known window kinds into the bounded evidence table, so callers cannot retain a mutable reference. */
 function copyWindows(windows: readonly DecisionQuotaWindow[]): DecisionQuotaWindow[] {
   if (windows.length > MAX_WINDOWS) return [];
-  return windows.filter(row => ["5h", "weekly", "monthly", "fable", "opus", "sonnet"].includes(row.window)).map(row => ({ window: row.window, percent: row.percent, observedAt: row.observedAt, ...(row.resetAt !== undefined ? { resetAt: row.resetAt } : {}) }));
+  return windows.filter(row => ["5h", "weekly", "monthly", "fable", "opus", "sonnet"].includes(row.window)).map(row => ({ window: row.window, percent: row.percent, observedAt: row.observedAt, ...(row.resetAt !== undefined ? { resetAt: row.resetAt } : {}), ...(row.family === "gem" || row.family === "cla" ? { family: row.family } : {}) }));
 }

package/src/providers/quota-types.ts

@@ -53,6 +53,8 @@ export interface ProviderQuotaCreditsUsd {
 }
 
 export interface ProviderQuota {
+  /** Display-only subscription credits; not a routing or exhaustion window. */
+  workbuddyCredits?: import("./workbuddy-credit-balance").WorkBuddyCreditBalance;
   fiveHourPercent?: number;
   fiveHourResetAt?: number;
   weeklyPercent?: number;

package/src/providers/quota.ts

@@ -1,5 +1,6 @@
 import { listCodexAuthAccountsSnapshot } from "../codex/auth-api";
 import { isAnthropicInstanceId, configuredAnthropicInstance } from "./anthropic-instance";
+import { ANTIGRAVITY_OAUTH_STORE_KEY } from "./antigravity-quota-identity";
 import { fetchAnthropicAccountQuota } from "./quota/anthropic-account-quota";
 import { thrownQuotaFailure } from "./quota/failure-codes";
 import { resolveEnvValue } from "../config";
@@ -124,7 +125,7 @@ async function maybeFetchProviderQuota(
     }
     if (provider.authMode === "oauth" && explicitAccountReader(name)) return await fetchExplicitCurrentQuota(name, provider, config);
     if (provider.authMode === "oauth" && configuredAnthropicInstance(config, name)) return fetchAnthropicQuota(name, config);
-    if (provider.authMode === "oauth" && name === "google-antigravity") return await fetchAntigravityQuota(name);
+    if (provider.authMode === "oauth" && name === ANTIGRAVITY_OAUTH_STORE_KEY) return await fetchAntigravityQuota(name, { ...provider });
     if (provider.authMode === "oauth" && name === "kiro") return fetchKiroQuota(name);
     // meta-muse: a device-logged-in account can be probed at the key endpoint; an
     // imported or pasted one cannot, and falls back to its last in-band observation.
@@ -368,6 +369,7 @@ async function readExplicitAccountQuota(provider: string, accountId: string, con
     case "kimi": result = await fetchKimiQuota(provider, config, accessToken); break;
     case "command-code": result = await fetchCommandCodeQuota(provider, config, accessToken); break;
     case "devin": result = await (await import("./quota/devin")).fetchDevinQuota(provider, accessToken, credential.apiBaseUrl, config.baseUrl); break;
+    case "workbuddy-ai": result = await (await import("./quota/workbuddy")).fetchWorkBuddyQuota(provider, credential); break;
     case "dsh-account": result = await (await import("./quota/dsh-account")).fetchDshAccountQuota(provider, accessToken, { accountId, credential }); break;
     default: return null;
   }

package/src/providers/registry/entries-core.ts

@@ -516,6 +516,15 @@ export const PROVIDER_REGISTRY_CORE: readonly ProviderRegistryEntry[] = [
     // The proprietary generate wire has no verified per-request serialization flag.
     parallelToolCalls: false,
   },
+  {
+    id: "workbuddy-ai", label: "WorkBuddy AI", adapter: "workbuddy-ai",
+    baseUrl: "https://www.workbuddy.ai/v2", authKind: "oauth", oauthId: "workbuddy-ai",
+    featured: false, dashboardPreset: true, liveModels: true,
+    defaultModel: "default-model", models: ["default-model"],
+    modelDiscovery: { path: "/v3/config", format: "workbuddy" },
+    dashboardUrl: "https://www.workbuddy.ai",
+    note: "Use your WorkBuddy AI international subscription. Sign in in your browser; models follow your account catalog.",
+  },
   {
     id: "dsh-account",
     label: "DeepSeek Account (DSH)",

package/src/providers/registry/entries-extended.ts

@@ -483,6 +483,10 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
     noPenaltyModels: KIMI_API_MODELS,
     autoToolChoiceOnlyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
     preserveReasoningContentModels: KIMI_API_MODELS,
+    // api.moonshot.ai answers `400 Invalid request: role 'developer' is not allowed`. The
+    // translated route already folds when nothing is recorded, but the native
+    // /v1/chat/completions passthrough only folds on an explicit record, so record it here.
+    foldDeveloperRoleToSystem: true,
     note: "International default (api.moonshot.ai). China accounts: choose China (.cn) or Custom for api.moonshot.cn.",
   },
   { id: "huggingface", label: "Hugging Face", baseUrl: "https://router.huggingface.co/v1", adapter: "openai-chat", authKind: "key", dashboardUrl: "https://huggingface.co/settings/tokens" },
@@ -837,6 +841,8 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
   {
     id: "alibaba-token-plan",
     label: "Alibaba Token Plan (Beijing)",
+    // Official Personal quick start (2026-10-23) still documents this host:
+    // https://help.aliyun.com/en/model-studio/token-plan-personal-quick-start
     baseUrl: "https://token-plan.cn-beijing.maas.aliyuncs.com/compatible-mode/v1",
     adapter: "openai-chat",
     authKind: "key",
@@ -887,19 +893,23 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
       "deepseek-v4-pro-0813": deepseekThinkingEffortsFor("deepseek-v4-pro-0813"),
       "deepseek-v4-flash-0731": deepseekThinkingEffortsFor("deepseek-v4-flash-0731"),
       "deepseek-v4.1-flash": deepseekThinkingEffortsFor("deepseek-v4.1-flash"),
+      "deepseek-v4.1-flash-1017": deepseekThinkingEffortsFor("deepseek-v4.1-flash-1017"),
     },
     modelReasoningEffortMap: {
       "deepseek-v4-pro": deepseekReasoningMapFor("deepseek-v4-pro"),
       "deepseek-v4-pro-0813": deepseekReasoningMapFor("deepseek-v4-pro-0813"),
       "deepseek-v4-flash-0731": deepseekReasoningMapFor("deepseek-v4-flash-0731"),
       "deepseek-v4.1-flash": deepseekReasoningMapFor("deepseek-v4.1-flash"),
+      "deepseek-v4.1-flash-1017": deepseekReasoningMapFor("deepseek-v4.1-flash-1017"),
     },
     // Probed 260915 on the plan gateway: json_object returns valid JSON, strict
     // json_schema is rejected 400 ("This response_format type is unavailable now")
     // in both thinking modes, so requests downgrade to json_object rather than
     // sending a schema the gateway refuses.
     noJsonSchemaModels: ["deepseek-v4.1-flash"],
-    modelDefaultReasoningEfforts: Object.fromEntries(QWEN38_FAMILY.map(id => [id, "xhigh"])),
+    // 261020 strict schema rejected; json_object unverified. Omit rather than invent a fallback.
+    noStructuredOutputModels: ["deepseek-v4.1-flash-1017"],
+    modelDefaultReasoningEfforts: { ...Object.fromEntries(QWEN38_FAMILY.map(id => [id, "xhigh"])), "deepseek-v4.1-flash-1017": "high" },
     directReasoningEffortModels: QWEN38_FAMILY,
     thinkingBudgetModels: ALIBABA_TOKEN_PLAN_QWEN_MODELS.filter(id => !QWEN38_FAMILY.includes(id)),
     preserveReasoningContentModels: ALIBABA_TOKEN_PLAN_PRESERVE_REASONING,
@@ -935,25 +945,27 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
       "glm-5.3": ZAI_GLM_53_REASONING_EFFORTS,
       "deepseek-v4-pro": deepseekThinkingEffortsFor("deepseek-v4-pro"),
       "deepseek-v4-pro-0813": deepseekThinkingEffortsFor("deepseek-v4-pro-0813"),
-      "deepseek-v4-flash": deepseekThinkingEffortsFor("deepseek-v4-flash"),
       "deepseek-v4-flash-0731": deepseekThinkingEffortsFor("deepseek-v4-flash-0731"),
       "deepseek-v4.1-flash": deepseekThinkingEffortsFor("deepseek-v4.1-flash"),
+      "deepseek-v4.1-flash-1017": deepseekThinkingEffortsFor("deepseek-v4.1-flash-1017"),
     },
     modelReasoningEffortMap: {
       "deepseek-v4-pro": deepseekReasoningMapFor("deepseek-v4-pro"),
       "deepseek-v4-pro-0813": deepseekReasoningMapFor("deepseek-v4-pro-0813"),
-      "deepseek-v4-flash": deepseekReasoningMapFor("deepseek-v4-flash"),
       "deepseek-v4-flash-0731": deepseekReasoningMapFor("deepseek-v4-flash-0731"),
       "deepseek-v4.1-flash": deepseekReasoningMapFor("deepseek-v4.1-flash"),
+      "deepseek-v4.1-flash-1017": deepseekReasoningMapFor("deepseek-v4.1-flash-1017"),
     },
     // Same 260915 json_schema rejection probe as the Beijing entry.
     noJsonSchemaModels: ["deepseek-v4.1-flash"],
+    // 261020 strict schema rejected; json_object unverified. Omit rather than invent a fallback.
+    noStructuredOutputModels: ["deepseek-v4.1-flash-1017"],
     directReasoningEffortModels: QWEN38_FAMILY,
     thinkingBudgetModels: ALIBABA_INTL_TOKEN_PLAN_QWEN_MODELS.filter(id => !QWEN38_FAMILY.includes(id)),
     preserveReasoningContentModels: ALIBABA_TOKEN_PLAN_PRESERVE_REASONING,
     noVisionModels: ALIBABA_TOKEN_PLAN_NO_VISION,
-    noReasoningModels: ["kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5", "deepseek-v3.2", "glm-5.1", "glm-5", "MiniMax-M2.5"],
-    modelDefaultReasoningEfforts: Object.fromEntries(QWEN38_FAMILY.map(id => [id, "xhigh"])),
+    noReasoningModels: ["kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5", "glm-5.1", "glm-5", "MiniMax-M2.5"],
+    modelDefaultReasoningEfforts: { ...Object.fromEntries(QWEN38_FAMILY.map(id => [id, "xhigh"])), "deepseek-v4.1-flash-1017": "high" },
     promptCacheKey: true,
   },
   // NEEDS_HUMAN 2026-07-10: kept for config compatibility, but this is a dashboard URL,
@@ -1303,14 +1315,15 @@ export const PROVIDER_REGISTRY_EXTENDED: readonly ProviderRegistryEntry[] = [
     baseUrl: "https://api.xiaomimimo.com/api/free-ai/openai/chat",
     authKind: "key",
     keyOptional: true,
-    featured: true,
+    featured: false,
+    deprecated: "Deprecated: requests through this entry fail. Bootstrap may still return a JWT, but /api/free-ai/openai/chat rejects the configured model \"mimo-auto\" with 400 \"Unsupported model\". Kept only so existing saved providers keep resolving; new setups should use the keyed xiaomi provider.",
     liveModels: true,
     dashboardUrl: "https://xiaomimimo.com",
     defaultModel: "mimo-auto",
     models: ["mimo-auto"],
     reasoningEfforts: ["low", "medium", "high"],
     reasoningEffortMap: { xhigh: "high", max: "high", ultra: "high" },
-    note: "No key needed — uses Xiaomi MiMo's free public tier (limited-time offer). A JWT is bootstrapped automatically with an anonymous random client id stored locally. The endpoint contract mirrors the official MiMoCode client and is not publicly documented — Xiaomi may change or restrict it at any time. Prompts may be processed/retained by Xiaomi; do not send confidential material.",
+    note: "Deprecated: the anonymous MiMo free lane stopped completing requests for this entry in 2026. Bootstrap may still issue a JWT, but /api/free-ai/openai/chat rejects the configured model \"mimo-auto\" with 400 \"Unsupported model\", so turns cannot finish. This entry remains only so existing saved providers keep resolving; add the keyed Xiaomi provider at https://platform.xiaomimimo.com instead. Xiaomi may resume or change the service at any time.",
   },
   // Xiaomi MiMo paid token plan. Separate host and wire from both `xiaomi` (Anthropic) and
   // `mimo-free` (free tier, bespoke adapter), so it needs its own entry rather than a variant.

package/src/providers/registry/model-ids.ts

@@ -61,6 +61,7 @@ export const REGISTRY_FIELD_MODEL_ID_ROLES = {
   featured: NONE,
   sponsor: NONE,
   dashboardPreset: NONE,
+  deprecated: NONE,
   note: NONE,
   dashboardUrl: NONE,
   defaultModel: NONE,
@@ -114,10 +115,12 @@ export const REGISTRY_FIELD_MODEL_ID_ROLES = {
   noStopModels: NONE,
   noPenaltyModels: NONE,
   noJsonSchemaModels: NONE,
+  noStructuredOutputModels: NONE,
   parallelToolCalls: NONE,
   promptCacheKey: NONE,
   chatServiceTier: NONE,
   openaiChatEofTolerance: NONE,
+  foldDeveloperRoleToSystem: NONE,
   autoToolChoiceOnlyModels: NONE,
   preserveReasoningContentModels: NONE,
   requiresReasoningPlaceholderModels: NONE,

package/src/providers/registry/model-seeds.ts

@@ -573,22 +573,22 @@ export const deepseekReasoningMapFor = (modelId: string): Record<string, string>
 // text-only input, strict json_schema accepted. glm-5.3-flash REMAINS OUT:
 // still never served by the Token Plan gateway (docs.z.ai VLM id, not plan
 // entitlement).
-// The Beijing preset keeps the Personal Edition subset; non-chat ids (audio/image/
-// video families) stay out: they answer only on async endpoints openai-chat cannot
-// reach. deepseek-v4-pro-0813 is callable but NOT listed by /models, which is the
-// reason liveModels must stay false for this provider. deepseek-v4.1-flash is the
-// 260910 DeepSeek rename row: listed on /models on both tiers and regions from 260915,
-// hybrid thinking, vision via user message and tool result, json_object but not
-// json_schema (see noJsonSchemaModels on the entries).
-// Beijing serves the Personal Edition, so this is the Personal-tier roster probed
-// 260909 (a strict subset of Team). deepseek-v4-pro-0813 stays out of the Beijing
-// entry: its callability is only proven on Team keys, and no Personal key has been
-// shown to reach it. The Beijing entry also shares the intl maps, so it carries a
-// few orphan keys (kimi/glm-5/MiniMax rows); harmless, and one map beats two
-// drifting ones.
+// Both regions support Personal and Team subscriptions. Keep the existing Beijing
+// Personal and International Team model rosters. Discovery stays disabled because
+// the upstream list includes non-chat models.
+// 261020: 0813 is now listed on Global Personal/Team and CN Team, and the CN Personal
+// overview documents entitlement. Add it to the existing Personal curation without
+// claiming new Personal inference. Retired 261011: deepseek-v3.2 and bare v4-flash.
+// 1017 evidence: https://www.qianwenai.com/models/deepseek-v4.1-flash-1017 and
+// https://docs.qwencloud.com/api-reference/chat/openai-chat (checked 2026-10-20).
+// Global Personal probes: completion, thinking-off, low effort, tools, user/tool images
+// and Responses succeed; strict json_schema rejects 400. json_object and replay are
+// untested on this exact id. Published 1M context and 393K output use 393216 here,
+// consistent with the existing plan budget convention, not a measured boundary;
+// the lab's 384000 precision differs. Do not inherit the bare model's JSON fallback.
 export const ALIBABA_TOKEN_PLAN_MODELS = [
   "qwen3.8-max", "qwen3.8-flash", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash",
-  "deepseek-v4-pro", "deepseek-v4-flash-0731", "deepseek-v4.1-flash", "glm-5.2", "glm-5.3",
+  "deepseek-v4-pro", "deepseek-v4-pro-0813", "deepseek-v4-flash-0731", "deepseek-v4.1-flash", "deepseek-v4.1-flash-1017", "glm-5.2", "glm-5.3",
 ];
 export const ALIBABA_TOKEN_PLAN_QWEN_MODELS = [
   "qwen3.8-max", "qwen3.8-flash", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash",
@@ -604,12 +604,13 @@ export const ALIBABA_TOKEN_PLAN_INPUT_MODALITIES: Record<string, string[]> = {
   "deepseek-v4-flash-0731": ["text"],
   // Vision probed on the plan gateway 260915 (user message and tool result, both 200).
   "deepseek-v4.1-flash": ["text", "image"],
+  "deepseek-v4.1-flash-1017": ["text", "image"],
   "glm-5.2": ["text"],
   "glm-5.3": ["text"],
 };
 
 // 260721 Alibaba Token Plan International (ap-southeast-1 / Singapore, hardened 260721).
-// Multi-vendor lineup distinct from Beijing — includes DeepSeek V4 flash, Kimi K2.7, MiniMax.
+// International Team model roster; tier and region are independent.
 // Evidence: https://www.alibabacloud.com/help/en/model-studio/token-plan-overview
 //           https://qwencloud.com/pricing/token-plan (qwen3.8 metadata)
 // The Team Edition roster (Singapore), verified identical to the CN Team set on 260909.
@@ -617,7 +618,7 @@ export const ALIBABA_TOKEN_PLAN_INPUT_MODALITIES: Record<string, string[]> = {
 // listed on /models on both regions) after being dropped as "retired" upstream.
 export const ALIBABA_INTL_TOKEN_PLAN_MODELS = [
   "qwen3.8-max", "qwen3.8-flash", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash",
-  "deepseek-v4-pro", "deepseek-v4-pro-0813", "deepseek-v4-flash", "deepseek-v4-flash-0731", "deepseek-v4.1-flash", "deepseek-v3.2",
+  "deepseek-v4-pro", "deepseek-v4-pro-0813", "deepseek-v4-flash-0731", "deepseek-v4.1-flash", "deepseek-v4.1-flash-1017",
   "kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5",
   "glm-5.2", "glm-5.3", "glm-5.1", "glm-5",
   "MiniMax-M2.5",
@@ -702,8 +703,6 @@ export const VOLCENGINE_PLAN_TEXT_ONLY_MODELS = [
 export const ALIBABA_INTL_TOKEN_PLAN_INPUT_MODALITIES: Record<string, string[]> = {
   ...ALIBABA_TOKEN_PLAN_INPUT_MODALITIES,
   "qwen3.6-plus": ["text", "image"],
-  "deepseek-v4-flash": ["text"],
-  "deepseek-v3.2": ["text"],
   "kimi-k2.7-code": ["text", "image"],
   "kimi-k2.6": ["text", "image"],
   "kimi-k2.5": ["text", "image"],
@@ -712,14 +711,14 @@ export const ALIBABA_INTL_TOKEN_PLAN_INPUT_MODALITIES: Record<string, string[]>
   "MiniMax-M2.5": ["text"],
 };
 
-// Shared Token Plan metadata (260909 gateway probes; output ceilings are max_tokens
-// boundary probes: accept at N, reject at N+1).
+// Shared Token Plan metadata: legacy rows use 260909 max_tokens boundary probes.
+// The 1017 row uses published card normalization, not a boundary probe (see above).
 export const QWEN38_FAMILY = ["qwen3.8-max", "qwen3.8-flash"];
 export const ALIBABA_TOKEN_PLAN_CONTEXT_WINDOWS: Record<string, number> = {
   "qwen3.8-max": 1_000_000, "qwen3.8-flash": 1_000_000, "qwen3.7-max": 1_000_000, "qwen3.7-plus": 1_000_000,
   "qwen3.6-plus": 1_000_000, "qwen3.6-flash": 1_000_000,
-  "deepseek-v4-pro": 1_000_000, "deepseek-v4-pro-0813": 1_000_000, "deepseek-v4-flash": 1_000_000,
-  "deepseek-v4-flash-0731": 1_000_000, "deepseek-v4.1-flash": 1_000_000, "deepseek-v3.2": 131_072,
+  "deepseek-v4-pro": 1_000_000, "deepseek-v4-pro-0813": 1_000_000,
+  "deepseek-v4-flash-0731": 1_000_000, "deepseek-v4.1-flash": 1_000_000, "deepseek-v4.1-flash-1017": 1_000_000,
   "kimi-k2.7-code": 262_144, "kimi-k2.6": 262_144, "kimi-k2.5": 262_144,
   "glm-5.2": 1_000_000, "glm-5.3": 1_000_000, "glm-5.1": 202_752, "glm-5": 202_752,
   "MiniMax-M2.5": 196_608,
@@ -727,20 +726,20 @@ export const ALIBABA_TOKEN_PLAN_CONTEXT_WINDOWS: Record<string, number> = {
 export const ALIBABA_TOKEN_PLAN_MAX_OUTPUT_TOKENS: Record<string, number> = {
   "qwen3.8-max": 131_072, "qwen3.8-flash": 131_072, "qwen3.7-max": 131_072, "qwen3.7-plus": 131_072,
   "qwen3.6-plus": 65_536, "qwen3.6-flash": 65_536,
-  "deepseek-v4-pro": 393_216, "deepseek-v4-pro-0813": 393_216, "deepseek-v4-flash": 393_216,
-  "deepseek-v4-flash-0731": 393_216, "deepseek-v4.1-flash": 393_216, "deepseek-v3.2": 65_536,
+  "deepseek-v4-pro": 393_216, "deepseek-v4-pro-0813": 393_216,
+  "deepseek-v4-flash-0731": 393_216, "deepseek-v4.1-flash": 393_216, "deepseek-v4.1-flash-1017": 393_216,
   "kimi-k2.7-code": 262_144, "kimi-k2.6": 262_144, "kimi-k2.5": 98_304,
   "glm-5.2": 131_072, "glm-5.3": 131_072, "glm-5.1": 128_000, "glm-5": 16_384,
   "MiniMax-M2.5": 32_768,
 };
 export const ALIBABA_TOKEN_PLAN_NO_VISION = [
-  "qwen3.7-max", "deepseek-v4-pro", "deepseek-v4-pro-0813", "deepseek-v4-flash",
-  "deepseek-v4-flash-0731", "deepseek-v3.2", "glm-5.2", "glm-5.3", "glm-5.1", "glm-5", "MiniMax-M2.5",
+  "qwen3.7-max", "deepseek-v4-pro", "deepseek-v4-pro-0813",
+  "deepseek-v4-flash-0731", "glm-5.2", "glm-5.3", "glm-5.1", "glm-5", "MiniMax-M2.5",
 ];
 export const ALIBABA_TOKEN_PLAN_PRESERVE_REASONING = [
   "qwen3.8-max", "qwen3.8-flash", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash",
-  "deepseek-v4-pro", "deepseek-v4-pro-0813", "deepseek-v4-flash", "deepseek-v4-flash-0731",
-  "deepseek-v4.1-flash", "glm-5.2", "glm-5.3",
+  "deepseek-v4-pro", "deepseek-v4-pro-0813", "deepseek-v4-flash-0731",
+  "deepseek-v4.1-flash", "deepseek-v4.1-flash-1017", "glm-5.2", "glm-5.3",
 ];
 
 // 260717 Kimi K3: the subscription endpoint uses one upstream id (`k3`) for both

package/src/providers/registry/types.ts

@@ -63,6 +63,8 @@ export interface ProviderModelDiscoveryFilter {
 }
 
 interface ProviderModelDiscoverySharedSpec {
+  /** Product-config envelope used by WorkBuddy AI instead of an OpenAI model list. */
+  format?: "workbuddy";
   /** Query parameters applied to the resolved discovery URL. */
   query?: Readonly<Record<string, string>>;
   /** Top-level response key containing model rows; defaults to `data`. */
@@ -160,6 +162,14 @@ export interface ProviderRegistryEntry {
    */
   sponsor?: { tier: "main" | "standard"; url: string };
   dashboardPreset?: boolean;
+  /**
+   * The upstream route or service this entry depends on is gone. Deprecated entries stay in the
+   * registry so saved providers keep resolving and a pinned adapter keeps loading, but they are
+   * excluded from the add-provider picker, the init menu, the featured list, and new key logins.
+   * The value is the operator-facing reason; it is registry-only and never seeded into saved
+   * config.
+   */
+  deprecated?: string;
   note?: string;
   dashboardUrl?: string;
   defaultModel?: string;
@@ -334,6 +344,8 @@ export interface ProviderRegistryEntry {
    * `directReasoningEffortModels` above is registry-owned.
    */
   noJsonSchemaModels?: string[];
+  /** Exact models with no established structured-output wire format. Registry-owned at routing. */
+  noStructuredOutputModels?: string[];
   /** Opt this provider into parallel tool calls (see OcxProviderConfig.parallelToolCalls). */
   parallelToolCalls?: boolean;
   /** Opt this provider into forwarding prompt_cache_key (OpenAI-specific; strict backends reject it). */
@@ -346,6 +358,12 @@ export interface ProviderRegistryEntry {
   chatServiceTier?: boolean;
   /** OpenAI Chat EOF policy for gateways that omit terminal frames after complete tool calls. */
   openaiChatEofTolerance?: boolean;
+  /**
+   * Record that this destination rejects the Chat `developer` role (`400 role 'developer' is not
+   * allowed`), so both the translated route and the native `/v1/chat/completions` passthrough send
+   * `system` instead. Seeds `foldDeveloperRoleToSystem` on the provider config; a user value wins.
+   */
+  foldDeveloperRoleToSystem?: boolean;
   autoToolChoiceOnlyModels?: string[];
   preserveReasoningContentModels?: string[];
   requiresReasoningPlaceholderModels?: string[];
@@ -382,6 +400,6 @@ export type ProviderConfigSeed = Pick<
   | "modelMaxInputTokens" | "defaultMaxOutputTokens" | "modelMaxOutputTokens"
   | "reasoningEfforts" | "modelReasoningEfforts" | "modelDefaultReasoningEfforts" | "reasoningEffortMap" | "modelReasoningEffortMap" | "reasoningWireFormat"
   | "noVisionModels" | "noReasoningModels" | "noTemperatureModels" | "noTopPModels" | "noStopModels" | "noPenaltyModels"
-  | "autoToolChoiceOnlyModels" | "preserveReasoningContentModels" | "requiresReasoningPlaceholderModels" | "reasoningSplitModels" | "inlineThinkTagModels" | "reasoningDetailsModels" | "thinkingToggleModels" | "thinkingBudgetModels" | "escapeBuiltinToolNames" | "openaiChatEofTolerance" | "showThinkingSummary"
+  | "autoToolChoiceOnlyModels" | "preserveReasoningContentModels" | "requiresReasoningPlaceholderModels" | "reasoningSplitModels" | "inlineThinkTagModels" | "reasoningDetailsModels" | "thinkingToggleModels" | "thinkingBudgetModels" | "escapeBuiltinToolNames" | "openaiChatEofTolerance" | "foldDeveloperRoleToSystem" | "showThinkingSummary"
   | "googleMode" | "project" | "location" | "headers"
 >;

package/src/providers/resolved-model-policy.ts

@@ -31,8 +31,8 @@ export type StaticProviderPolicyField =
   | "modelMaxOutputTokens" | "reasoningEfforts" | "modelReasoningEfforts" | "modelReasoningEffortsAuthoritative"
   | "modelDefaultReasoningEfforts" | "reasoningEffortMap" | "modelReasoningEffortMap"
   | "reasoningWireFormat" | "noVisionModels" | "noReasoningModels" | "noTemperatureModels"
-  | "noTopPModels" | "noStopModels" | "noPenaltyModels" | "noJsonSchemaModels" | "parallelToolCalls"
-  | "promptCacheKey" | "chatServiceTier" | "openaiChatEofTolerance" | "statelessResponses"
+  | "noTopPModels" | "noStopModels" | "noPenaltyModels" | "noJsonSchemaModels" | "noStructuredOutputModels" | "parallelToolCalls"
+  | "promptCacheKey" | "chatServiceTier" | "openaiChatEofTolerance" | "foldDeveloperRoleToSystem" | "statelessResponses"
   | "requiresAdjacentResponsesToolResults" | "requiresPairedResponsesToolResults" | "annotateEmptyToolOutputs"
   | "fastWire" | "supportsServiceTier" | "modelSupportsServiceTier" | "supportsOpenAiWebSearchToolFields"
   | "supportsResponsesCustomTools" | "preserveResponsesReasoningContent" | "dropResponsesReasoningItems"
@@ -173,7 +173,7 @@ export function resolveModelPolicy(input: ResolveModelPolicyInput): ResolvedMode
     "responsesPath", "chatCompletionsPath", "keyOptional", "freeTier", "modelSuffixBracketStrip",
     "defaultModel", "models", "liveModels", "contextWindow", "defaultMaxOutputTokens",
     "reasoningWireFormat", "parallelToolCalls", "promptCacheKey",
-    "openaiChatEofTolerance", "statelessResponses",
+    "openaiChatEofTolerance", "foldDeveloperRoleToSystem", "statelessResponses",
     "requiresAdjacentResponsesToolResults", "requiresPairedResponsesToolResults",
     "annotateEmptyToolOutputs", "fastWire", "supportsServiceTier",
     "supportsOpenAiWebSearchToolFields", "supportsResponsesCustomTools",
@@ -231,7 +231,7 @@ export function resolveModelPolicy(input: ResolveModelPolicyInput): ResolvedMode
   for (const key of [
     "noVisionModels", "noReasoningModels", "noTemperatureModels", "noTopPModels",
     "noStopModels",
-    "noPenaltyModels", "noJsonSchemaModels", "autoToolChoiceOnlyModels",
+    "noPenaltyModels", "noJsonSchemaModels", "noStructuredOutputModels", "autoToolChoiceOnlyModels",
     "preserveReasoningContentModels", "requiresReasoningPlaceholderModels",
     "reasoningSplitModels", "reasoningDetailsModels", "thinkingToggleModels", "thinkingBudgetModels",
   ] as const) putUnion(key, entry?.[key]);

package/src/providers/routed-config.ts

@@ -21,6 +21,7 @@ import {
   providerSupportsLiveModelDiscovery,
 } from "./static-model-discovery";
 import { resolveModelPolicy } from "./resolved-model-policy";
+import { mapFill, nestedMapFill, stableUnion } from "./resolved-model-policy-merge";
 
 /** Same endpoint modulo surrounding space and trailing slashes — matches `matchBaseUrlChoice`. */
 function isSameEndpoint(a: string, b: string): boolean {
@@ -124,6 +125,15 @@ export function routedProviderConfig(providerName: string, provider: OcxProvider
     return {
       ...provider,
       apiKey: usableResolvedApiKey(provider.apiKey),
+      ...(destination?.modelDefaultReasoningEfforts ? {
+        modelDefaultReasoningEfforts: mapFill(destination.modelDefaultReasoningEfforts, provider.modelDefaultReasoningEfforts)[0],
+      } : {}),
+      ...(destination?.modelReasoningEffortMap ? {
+        modelReasoningEffortMap: nestedMapFill(destination.modelReasoningEffortMap, provider.modelReasoningEffortMap)[0],
+      } : {}),
+      ...(destination?.noStructuredOutputModels?.length ? {
+        noStructuredOutputModels: stableUnion(destination.noStructuredOutputModels, provider.noStructuredOutputModels)[0],
+      } : {}),
       ...(provider.dropResponsesReasoningItems === undefined && destination?.dropResponsesReasoningItems !== undefined
         ? { dropResponsesReasoningItems: destination.dropResponsesReasoningItems }
         : {}),
@@ -172,6 +182,7 @@ export function routedProviderConfig(providerName: string, provider: OcxProvider
   const noStopModels = staticPolicy.noStopModels;
   const noPenaltyModels = staticPolicy.noPenaltyModels;
   const noJsonSchemaModels = staticPolicy.noJsonSchemaModels;
+  const noStructuredOutputModels = staticPolicy.noStructuredOutputModels;
   const autoToolChoiceOnlyModels = staticPolicy.autoToolChoiceOnlyModels;
   const preserveReasoningContentModels = staticPolicy.preserveReasoningContentModels;
   const requiresReasoningPlaceholderModels = staticPolicy.requiresReasoningPlaceholderModels;
@@ -302,6 +313,9 @@ export function routedProviderConfig(providerName: string, provider: OcxProvider
     ...(provider.openaiChatEofTolerance === undefined && registryEntry.openaiChatEofTolerance !== undefined
       ? { openaiChatEofTolerance: registryEntry.openaiChatEofTolerance }
       : {}),
+    ...(provider.foldDeveloperRoleToSystem === undefined && registryEntry.foldDeveloperRoleToSystem !== undefined
+      ? { foldDeveloperRoleToSystem: registryEntry.foldDeveloperRoleToSystem }
+      : {}),
     ...(provider.reasoningWireFormat === undefined && registryEntry.reasoningWireFormat !== undefined
       ? { reasoningWireFormat: registryEntry.reasoningWireFormat }
       : {}),
@@ -325,6 +339,7 @@ export function routedProviderConfig(providerName: string, provider: OcxProvider
     ...(noStopModels ? { noStopModels } : {}),
     ...(noPenaltyModels ? { noPenaltyModels } : {}),
     ...(noJsonSchemaModels ? { noJsonSchemaModels } : {}),
+    ...(noStructuredOutputModels ? { noStructuredOutputModels } : {}),
     ...(autoToolChoiceOnlyModels ? { autoToolChoiceOnlyModels } : {}),
     ...(preserveReasoningContentModels ? { preserveReasoningContentModels } : {}),
     ...(requiresReasoningPlaceholderModels ? { requiresReasoningPlaceholderModels } : {}),