@oh-my-pi/pi-catalog 18.13.2 → 18.14.0

package/CHANGELOG.md

@@ -2,6 +2,46 @@
 
 ## [Unreleased]
 
+## [18.14.0] - 2026-10-29
+
+### Breaking Changes
+
+- Removed the raw `@oh-my-pi/pi-catalog/models.json` export; use the bundled model registry APIs instead.
+
+### Added
+
+- Added `nextBuildPriceChange(cost, now)` to identify when a model's dated pricing card may change, allowing callers to safely reuse earlier builds until that time.
+- Added the `selector-revision-suffix` catalog axis for Google Vertex, so `google-vertex/claude-opus-4-8` resolves exactly to `claude-opus-4-8@default`.
+
+### Changed
+
+- Reduced the bundled model catalog size and loading overhead.
+- Updated role defaults so Factory Droid image models are preferred for the `image` role, allowing `generate_image` to work with only a Factory login.
+- Updated `smol` and `slow` role defaults to use the currently bundled Cerebras and OpenAI Codex model identifiers.
+
+### Fixed
+
+- Fixed Mistral reasoning models so streamed thinking appears in the transcript and is preserved for subsequent turns.
+- Fixed later-turn HTTP 422 errors in Mistral-hosted DeepSeek conversations when assistant reasoning history is included.
+- Fixed DeepSeek V4.1 Flash image attachments being dropped for publisher-namespaced model IDs such as `ds/deepseek-flash`.
+
+## [18.13.3] - 2026-10-28
+
+### Added
+
+- Added DeepSeek V4.1 Flash 1017 to the Factory Droid provider for accounts with Factory's 1017 flag; turning thinking off on its Baseten route disables reasoning instead of falling back to low effort ([#17234](https://github.com/can1357/oh-my-pi/pull/17234) by [@skipper019](https://github.com/skipper019))
+
+### Changed
+
+- Made Kimi Code model discovery authoritative and cached per credential, added K3.1 Pro and K3.1 Pro Highspeed to the bundled Kimi Code catalog, and preserved live thinking metadata there ([#17224](https://github.com/can1357/oh-my-pi/pull/17224) by [@trukhinyuri](https://github.com/trukhinyuri))
+- Updated Bun engine requirement to ≥1.4.4
+
+### Fixed
+
+- Codex model refresh now lists models available to an entitled account even when other saved accounts return HTTP 402 ([#17198](https://github.com/can1357/oh-my-pi/issues/17198)).
+- LiteLLM discovery now reads every `/v2/model/info` page before selecting the proxy's model roster, rather than caching only the first 50 entries; malformed or incomplete pagination falls back without caching a partial page ([#17237](https://github.com/can1357/oh-my-pi/pull/17237) by [@maykonmichel](https://github.com/maykonmichel)).
+- Fixed hosted web search on OpenAI Responses and Codex Responses selecting GPT-5.6 Luna instead of GPT-6 Luna ([#17202](https://github.com/can1357/oh-my-pi/pull/17202) by [@mateo-bolanos](https://github.com/mateo-bolanos)).
+
 ## [18.13.2] - 2026-10-27
 
 ### Added

package/README.md

@@ -21,7 +21,7 @@ Import from subpaths (`@oh-my-pi/pi-catalog/<module>`) or the root barrel.
 
 ## models.json and rules.json are generated
 
-Never edit `src/models.json` or `src/compat/rules.json` by hand. `models.json` is produced from upstream sources (stencil.so, provider catalog discovery, OpenCode docs) by `scripts/generate-models.ts`; `rules.json` is compiled from the KDL tree in `src/compat/rules/`. Regenerate with:
+Never edit `src/models.json` or `src/compat/rules.json` by hand. `models.json` is a compact generated catalog; use `getBundledProviders()` and `getBundledModels()` instead of reading its internal format. It is produced from upstream sources (stencil.so, provider catalog discovery, OpenCode docs) by `scripts/generate-models.ts`; `rules.json` is compiled from the KDL tree in `src/compat/rules/`. Regenerate with:
 
 ```sh
 bun run gen:compat   # src/compat/rules/**/*.kdl -> src/compat/rules.json
@@ -36,7 +36,7 @@ Model- or provider-conditional policy (identity, effort ladders, wire quirks, mo
 bun add @oh-my-pi/pi-catalog
 ```
 
-Ships TypeScript source directly (no build step); requires Bun ≥ 1.3.14.
+Ships TypeScript source directly (no build step); requires Bun ≥ 1.4.4.
 
 ## Cost calculation
 
@@ -47,7 +47,7 @@ The `models` subpath (also exported from the root) provides timestamp-aware pric
 | `calculateCost(model, usage, timestamp?)`                         | Updates and returns `usage.cost` using `model.cost`.                                                      |
 | `calculateUsageCost(cost, usage, timestamp?)`                     | Updates and returns `usage.cost` using a `ModelCost`.                                                     |
 | `calculateUncachedInputCost(cost, promptInputTokens, timestamp?)` | Returns the cost of a fully uncached prompt.                                                              |
-| `getEffectiveTokenCost(cost, timestamp?)`                         | Returns the base card in force: the latest due effective rate, else the row's own legs.                  |
+| `getEffectiveTokenCost(cost, timestamp?)`                         | Returns the base card in force: the latest due effective rate, else the row's own legs.                   |
 | `getTimeBasedPricingPeriod(cost, timestamp?)`                     | Returns `"peak"`, `"off-peak"`, or `undefined` without a recurring tariff.                                |
 | `hasRecurringTariff(schedule)`                                    | Whether a `TimeBasedCost` carries a peak/off-peak tariff rather than only dated rate cards.               |
 | `getNextTimeBasedPricingTransition(cost, timestamp?)`             | Returns the next actual peak/off-peak change strictly after the timestamp, or `undefined` if none exists. |

package/dist/types/build.d.ts

@@ -6,7 +6,7 @@
  * Request handlers read fields — they never detect, parse ids, or allocate
  * compat per request.
  */
-import { type Api, type Model, type ModelSpec } from "./types.js";
+import { type Api, type Model, type ModelCost, type ModelSpec } from "./types.js";
 /**
  * Applies reviewed catalog-data value corrections (`cost-patch`,
  * `cache-read-at-input-rate`, `cache-write-rate`, `limits-patch`,
@@ -24,8 +24,16 @@ export declare function applyCatalogCorrections(model: Pick<ModelSpec<Api>, "cos
  */
 export declare function buildDiscoveredModel(spec: ModelSpec<Api>, providerType: string): Model<Api>;
 /**
- * Build one model from an authored spec. Bundled models.json rows are fully
- * materialized by the generator and consumed directly (see `models.ts`), so
- * this only runs for discovered/custom/override specs.
+ * The first instant after `now` at which {@link buildModel} could select a
+ * different dated price card for a spec carrying `cost`: the next
+ * `effectiveFrom` of the spec's own rate schedule or of any dated card in the
+ * compiled rules. A build of an unchanged spec made at or after `now` is
+ * identical to a fresh one until then. `Infinity` when no change is pending.
+ */
+export declare function nextBuildPriceChange(cost: ModelCost, now: number): number;
+/**
+ * Build one model from an authored spec. Bundled models are prebuilt by the
+ * generator and decoded by `models.ts`, so this only runs for discovered,
+ * custom, and override specs.
  */
 export declare function buildModel<TApi extends Api>(spec: ModelSpec<TApi>): Model<TApi>;

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

@@ -369,8 +369,10 @@ export type CompiledAuthValidation = {
     kind: "models-endpoint";
     label?: string;
     url: string;
-    /** Env var holding an alternate base URL; `/models` is appended to it. */
+    /** Env var holding an alternate base URL; `baseUrlPath` is appended to it. */
     baseUrlEnv?: string;
+    /** Path appended to the `baseUrlEnv` value (default `/models`), for bases that exclude the API version. */
+    baseUrlPath?: string;
     /** Hook returning extra request headers (may throw a configuration error). */
     headersHook?: string;
     optional?: boolean;

package/dist/types/discovery/codex.d.ts

@@ -27,12 +27,11 @@ export interface CodexModelDiscoveryResult {
     models: ModelSpec<"openai-codex-responses">[];
     etag?: string;
     /**
-     * Set when the backend rejected the credential itself (401/403, e.g.
-     * `token_revoked`); `models` is empty. A definitive per-account denial,
-     * unlike the `null` result for transport/parse failures, so multi-account
-     * discovery can skip the account instead of aborting.
+     * A definitive per-account denial (401/403 auth, 402 payment). `models` is
+     * empty; unlike transport/parse failures, sibling accounts can still supply
+     * an authoritative catalog.
      */
-    rejectedStatus?: 401 | 403;
+    rejectedStatus?: 401 | 402 | 403;
 }
 /**
  * Fetches model metadata from Codex backend and normalizes it for pi model management.

package/dist/types/model-cache.d.ts

@@ -48,6 +48,14 @@ export interface CacheEntry<TApi extends Api = Api> {
  * cache was opened under — Windows cannot remove a database that is still open.
  */
 export declare function closeModelCache(): void;
+/**
+ * Keeps one connection to a non-default cache path open until the returned
+ * handle is disposed. The default `models.db` already shares a long-lived
+ * connection; other paths otherwise open, initialize and close the database on
+ * every read and write, which a refresh pass repeats once per provider.
+ * Retains nest: the connection closes when the last one is disposed.
+ */
+export declare function retainModelCacheConnection(dbPath: string | undefined): Disposable;
 export declare function readModelCache<TApi extends Api>(providerId: string, ttlMs: number, now: () => number, dbPath?: string): CacheEntry<TApi> | null;
 /**
  * Persist a provider snapshot.

package/dist/types/model-catalog-format.d.ts

@@ -0,0 +1,37 @@
+import type { Api, Model } from "./types.js";
+type EncodedValue = string | number | boolean | null | [number];
+type ObjectNode = [0, EncodedValue[], number, number[]];
+type ArrayNode = [1, EncodedValue[]];
+type Node = ObjectNode | ArrayNode;
+type CompatDelta = [number[], EncodedValue];
+type ModelRow = [string, EncodedValue, CompatDelta?];
+type ProviderRow = [string, EncodedValue, ModelRow[]];
+interface CatalogData {
+    version: 1;
+    keys: string[];
+    booleanShapes: number[][];
+    nodes: Node[];
+    compat: [string, EncodedValue][];
+    providers: ProviderRow[];
+}
+/**
+ * Parsed compact catalog. Decoded nested objects and arrays can be shared by
+ * multiple models; callers must treat those values as immutable.
+ */
+export interface ModelCatalog {
+    readonly data: CatalogData;
+    readonly values: unknown[];
+    readonly decoded: boolean[];
+    readonly providers: Map<string, Record<string, Model<Api>>>;
+    readonly compatBases: Map<string, EncodedValue>;
+    readonly compatValues: Map<string, Map<number, Map<string, Record<string, unknown>>>>;
+}
+/** Encode a model catalog into a deterministic, compact JSON representation. */
+export declare function encodeModelCatalog(models: Record<string, Record<string, Model<Api>>>): string;
+/** Open a parsed catalog without materializing its provider models. */
+export declare function openModelCatalog(value: unknown): ModelCatalog;
+/** List catalog provider keys in deterministic order. */
+export declare function getCatalogProviders(catalog: ModelCatalog): string[];
+/** Materialize and cache one provider, or return undefined if it is absent. */
+export declare function getCatalogProvider(catalog: ModelCatalog, provider: string): Record<string, Model<Api>> | undefined;
+export {};

package/dist/types/models.d.ts

@@ -1,10 +1,13 @@
-import MODELS from "./models.json";
 import type { Api, KnownProvider, Model, ModelCost, ModelPricingStatus, TimeBasedCost, TokenCost, Usage } from "./types.js";
-export type GeneratedProvider = keyof typeof MODELS;
+/** Provider identifier accepted by the bundled model registry. */
+export type GeneratedProvider = string;
 /** Whether `provider` has bundled rows in models.json (e.g. a provider id authored in KDL). */
 export declare function isGeneratedProvider(provider: string): provider is GeneratedProvider;
+/** Look up a bundled model; nested catalog values are shared and must not be mutated. */
 export declare function getBundledModel<TApi extends Api = Api>(provider: GeneratedProvider, modelId: string): Model<TApi>;
-export declare function getBundledProviders(): KnownProvider[];
+/** List providers with bundled rows; the array is shared and frozen. */
+export declare function getBundledProviders(): readonly KnownProvider[];
+/** List a provider's bundled models; nested catalog values are shared and must not be mutated. */
 export declare function getBundledModels(provider: GeneratedProvider): Model<Api>[];
 /** Resolve display semantics without mistaking an absent rate card for a free model. */
 export declare function getModelPricingStatus(model: Pick<Model<Api>, "cost" | "pricingStatus">): ModelPricingStatus;

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.13.2",
+	"version": "18.14.0",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"keywords": [
 		"ai",
@@ -28,7 +28,6 @@
 		"README.md",
 		"CHANGELOG.md",
 		"LICENSE",
-		"THIRD-PARTY-NOTICES.txt",
 		"dist/types"
 	],
 	"type": "module",
@@ -39,10 +38,6 @@
 			"types": "./dist/types/index.d.ts",
 			"import": "./src/index.ts"
 		},
-		"./models.json": {
-			"types": "./dist/types/models.json.d.d.ts",
-			"import": "./src/models.json"
-		},
 		"./provider-models": {
 			"types": "./dist/types/provider-models/index.d.ts",
 			"import": "./src/provider-models/index.ts"
@@ -93,15 +88,15 @@
 		"gen:proto": "bun scripts/generate-protocols.ts"
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.13.2",
-		"@oh-my-pi/pi-utils": "18.13.2"
+		"@oh-my-pi/omptype": "18.14.0",
+		"@oh-my-pi/pi-utils": "18.14.0"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.13.2",
+		"@oh-my-pi/pi-ai": "18.14.0",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {
-		"bun": ">=1.3.14"
+		"bun": ">=1.4.4"
 	}
 }

package/src/build.ts

@@ -7,6 +7,7 @@
  * compat per request.
  */
 
+import RULES from "./compat/rules.json" with { type: "json" };
 import { resolveDiscoveryApi, resolveModelPolicy } from "./compat/resolve";
 import type { ModelIdentity } from "./compat/types";
 import { resolveModelTokenizer } from "./model-tokenizer";
@@ -463,10 +464,56 @@ export function buildDiscoveredModel(spec: ModelSpec<Api>, providerType: string)
 	return buildModel({ ...spec, api, providerType });
 }
 
+/** Every dated card instant in the compiled rules, ascending; collected on first use. */
+let ruleCardInstants: number[] | undefined;
+
+function collectRuleCardInstants(value: unknown, out: number[]): void {
+	if (Array.isArray(value)) {
+		for (const item of value) collectRuleCardInstants(item, out);
+		return;
+	}
+	if (typeof value !== "object" || value === null) return;
+	for (const key in value) {
+		const child: unknown = Reflect.get(value, key);
+		if (key === "effectiveFrom" && typeof child === "string") {
+			const instant = Date.parse(child);
+			if (Number.isFinite(instant)) out.push(instant);
+		} else {
+			collectRuleCardInstants(child, out);
+		}
+	}
+}
+
+/**
+ * The first instant after `now` at which {@link buildModel} could select a
+ * different dated price card for a spec carrying `cost`: the next
+ * `effectiveFrom` of the spec's own rate schedule or of any dated card in the
+ * compiled rules. A build of an unchanged spec made at or after `now` is
+ * identical to a fresh one until then. `Infinity` when no change is pending.
+ */
+export function nextBuildPriceChange(cost: ModelCost, now: number): number {
+	if (ruleCardInstants === undefined) {
+		const instants: number[] = [];
+		collectRuleCardInstants(RULES, instants);
+		ruleCardInstants = instants.sort((a, b) => a - b);
+	}
+	let next = Number.POSITIVE_INFINITY;
+	for (const instant of ruleCardInstants) {
+		if (instant > now) {
+			next = instant;
+			break;
+		}
+	}
+	for (const rate of cost.timeBased?.effectiveRates ?? []) {
+		if (rate.effectiveFrom > now && rate.effectiveFrom < next) next = rate.effectiveFrom;
+	}
+	return next;
+}
+
 /**
- * Build one model from an authored spec. Bundled models.json rows are fully
- * materialized by the generator and consumed directly (see `models.ts`), so
- * this only runs for discovered/custom/override specs.
+ * Build one model from an authored spec. Bundled models are prebuilt by the
+ * generator and decoded by `models.ts`, so this only runs for discovered,
+ * custom, and override specs.
  */
 export function buildModel<TApi extends Api>(spec: ModelSpec<TApi>): Model<TApi> {
 	const policy = resolveModelPolicy(spec);

package/src/compat/axes.ts

@@ -469,6 +469,12 @@ export const AXES: Readonly<Record<string, AxisDef>> = {
 		values: ["gemini", "anthropic", "codex", "xai", "openrouter", "openai"],
 	},
 	"web-search-model": { key: "webSearchModel", set: "catalog", shape: "scalar" },
+	/**
+	 * Revision suffix the host appends to the model's id (Vertex `@default`): a
+	 * `provider/id` selector written without it names this model exactly.
+	 * Unassigned: the id has no implied revision and must be written in full.
+	 */
+	"selector-revision-suffix": { key: "selectorRevisionSuffix", set: "catalog", shape: "scalar" },
 	"hosted-image": { key: "hostedImage", set: "catalog", shape: "scalar", values: [true, false] },
 	/** How the model line bills an input image; shape and formulas in `./image-tokenization`. */
 	"image-tokenization": { key: "imageTokenization", set: "catalog", shape: "object" },

package/src/compat/rules/README.md

@@ -179,16 +179,16 @@ provider "openrouter" {
 
 ### Selectors and nesting
 
-| Selector   | Form                                     | Matching semantics                                                                                                                                                             |
-| ---------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `class`    | `class "id" { ... }`                     | Exact class ID. At document root it may contain `on`, `on-api`, `family`, `revision`, and `models`. Under `provider` it may contain `family`, `revision`, and `models`.        |
-| `provider` | `provider "id" { ... }`                  | Exact provider ID. It is root-only and may contain `class`, `on-api`, and `models`.                                                                                           |
-| `on`       | `on "provider-a" "provider-b" { ... }`   | One or more provider IDs, combined as OR. It is allowed under a root `class`, and may contain `on-api` to conjoin an API scope plus `family`, `revision`, and `models`. |
-| `on-api`   | `on-api "adapter-a" "adapter-b" { ... }` | One or more request adapter IDs, combined as OR. At document root it may contain `class` and `models`; under a root `class` (directly or nested inside `on`) or `provider` it may contain `family`, `revision`, and `models`. |
-| `family`   | `family "id" { ... }`                    | Exact classified family ID. It may contain `revision` and `models`. A target with no family does not match.                                                                    |
-| `revision` | `revision ">=2.5 <4" { ... }`            | A non-empty, whitespace-separated conjunction of comparisons. It may contain `models`. A target with no revision does not match.                                               |
-| `models`   | `models "id" "vendor/*" { ... }`         | One or more alternatives, combined as OR. It may contain only `on-upstream`. `token="name"` matches an ASCII-case-insensitive token bounded by non-alphanumerics.                 |
-| `on-upstream` | `on-upstream "a" "b" { ... }`        | Exact selected upstream IDs, combined as OR. Allowed inside any selector scope once; preserves the containing scope's other permitted children. Absent upstream never matches. |
+| Selector      | Form                                     | Matching semantics                                                                                                                                                                                                            |
+| ------------- | ---------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `class`       | `class "id" { ... }`                     | Exact class ID. At document root it may contain `on`, `on-api`, `family`, `revision`, and `models`. Under `provider` it may contain `family`, `revision`, and `models`.                                                       |
+| `provider`    | `provider "id" { ... }`                  | Exact provider ID. It is root-only and may contain `class`, `on-api`, and `models`.                                                                                                                                           |
+| `on`          | `on "provider-a" "provider-b" { ... }`   | One or more provider IDs, combined as OR. It is allowed under a root `class`, and may contain `on-api` to conjoin an API scope plus `family`, `revision`, and `models`.                                                       |
+| `on-api`      | `on-api "adapter-a" "adapter-b" { ... }` | One or more request adapter IDs, combined as OR. At document root it may contain `class` and `models`; under a root `class` (directly or nested inside `on`) or `provider` it may contain `family`, `revision`, and `models`. |
+| `family`      | `family "id" { ... }`                    | Exact classified family ID. It may contain `revision` and `models`. A target with no family does not match.                                                                                                                   |
+| `revision`    | `revision ">=2.5 <4" { ... }`            | A non-empty, whitespace-separated conjunction of comparisons. It may contain `models`. A target with no revision does not match.                                                                                              |
+| `models`      | `models "id" "vendor/*" { ... }`         | One or more alternatives, combined as OR. It may contain only `on-upstream`. `token="name"` matches an ASCII-case-insensitive token bounded by non-alphanumerics.                                                             |
+| `on-upstream` | `on-upstream "a" "b" { ... }`            | Exact selected upstream IDs, combined as OR. Allowed inside any selector scope once; preserves the containing scope's other permitted children. Absent upstream never matches.                                                |
 
 Every selector scope may additionally contain `on-upstream`; it cannot replace an already constrained upstream. Class, provider/`on`, `on-api`, `on-upstream`, and family values are compared exactly and case-sensitively to the structured resolve target. Revision operators are `>=`, `>`, `<=`, `<`, and `=`; operands have one to three dot-separated unsigned 8-bit components, omitted components zero.
 
@@ -415,7 +415,7 @@ auth "anthropic" {
 
 Login kinds:
 
-- `login "api-key" { auth-url "…"; instructions "…"; prompt "…" placeholder="…"; empty-fallback "…"; normalize "strip-bearer"; validate … }` — `validate "chat-completions" base-url= model= tolerate-model-denied= max-tokens-field= max-tokens=`, `validate "anthropic-messages" base-url= model=`, or `validate "models-endpoint" url= base-url-env= headers-hook=`; all accept `label=` (error-message label, defaults to `name`), `optional=#true` (only auth failures reject), and `trust-forbidden=#true` (requires `optional=#true`; a 403 no longer rejects, so only a 401 does).
+- `login "api-key" { auth-url "…"; instructions "…"; prompt "…" placeholder="…"; empty-fallback "…"; normalize "strip-bearer"; validate … }` — `validate "chat-completions" base-url= model= tolerate-model-denied= max-tokens-field= max-tokens=`, `validate "anthropic-messages" base-url= model=`, or `validate "models-endpoint" url= base-url-env= base-url-path= headers-hook=` (`base-url-path`, default `/models`, is appended to the `base-url-env` value and needs it); all accept `label=` (error-message label, defaults to `name`), `optional=#true` (only auth failures reject), and `trust-forbidden=#true` (requires `optional=#true`; a 403 no longer rejects, so only a 401 does).
 - `login "oauth-code" { … }` as above; `token`/`refresh` `params` values may use `{code}`, `{state}`, `{redirect_uri}`, `{code_verifier}`, `{client_id}`, `{client_secret}`, `{refresh_token}`, `{scope}`; the standard grant parameters are sent unless `standard=#false`.
 - `login "device-code" { client-id …; base-url "…"; scopes …; headers-hook "…"; device url="{base}/…" body="form" { params {…} headers {…} }; token url="…" url-hook="…"; response user-code= device-code= verification-uri= verification-uri-complete= interval= expires-in=; instructions "Enter code: {user_code}"; credential {…}; userinfo …; after-exchange hook=… }`.
 - `login "custom" hook="name"` — the whole flow is a named `@oh-my-pi/pi-ai` hook (`src/registry/hooks/custom.ts`).

package/src/compat/rules/auth/typesafe.kdl

@@ -8,6 +8,7 @@ auth "typesafe" {
 		auth-url "https://console.typesafe.ai/"
 		instructions "Create or copy your API key from the TypeSafe console."
 		prompt "Paste your TypeSafe API key" placeholder="API key"
-		validate "models-endpoint" label="TypeSafe" url="https://api.typesafe.ai/v1/models" base-url-env="TYPESAFE_BASE_URL"
+		// TYPESAFE_BASE_URL is an origin without `/v1` (the SDK appends `/v1/systemone`).
+		validate "models-endpoint" label="TypeSafe" url="https://api.typesafe.ai/v1/models" base-url-env="TYPESAFE_BASE_URL" base-url-path="/v1/models"
 	}
 }

package/src/compat/rules/classes/deepseek.kdl

@@ -90,10 +90,11 @@ class "deepseek" {
 	}
 	// V4.1 Flash is natively multimodal (vision encoder + projector, image-text
 	// pre-training) but its release id carries no `vision` token to key on.
+	// The leading glob also covers publisher-namespaced ids (`ds/deepseek-flash`).
 	// Same priority as the V4.1 Flash carve-out: the OpenCode Go provider rule
 	// declares the same value for its `deepseek-flash` lane and would otherwise
 	// tie at equal rank.
-	models "deepseek-flash" priority=10 {
+	models "*deepseek-flash" priority=10 {
 		strip-image-input #false
 	}
 	// DSML is the model's own tool-call grammar, not a host quirk: any server

package/src/compat/rules/classes/openai.kdl

@@ -84,14 +84,14 @@ class "openai" {
 	on-api "openai-responses" {
 		revision ">=5" {
 			web-search "openai"
-			web-search-model "gpt-5.6-luna"
+			web-search-model "gpt-6-luna"
 			hosted-image #true
 			image-model "gpt-image-2"
 		}
 	}
 	on-api "openai-codex-responses" {
 		revision ">=5" {
-			web-search-model "gpt-5.6-luna"
+			web-search-model "gpt-6-luna"
 			hosted-image #true
 			image-model "gpt-image-2"
 		}

package/src/compat/rules/providers/factory-droid.kdl

@@ -345,6 +345,12 @@ provider "factory-droid" {
 			cost input=0 output=0 cache-read=0 cache-write=0
 			limits context=136464 max-tokens=65536
 		}
+		model "deepseek-v4.1-flash-1017" name="DeepSeek V4.1 Flash 1017" {
+			reasoning #true
+			input "text" "image"
+			cost input=0 output=0 cache-read=0 cache-write=0
+			limits context=908928 max-tokens=131072
+		}
 		model "deepseek-v4.1-flash" name="DeepSeek V4.1 Flash" {
 			reasoning #true
 			input "text" "image"
@@ -975,6 +981,17 @@ provider "factory-droid" {
 		thinking-requires-effort #false
 		thinking-default-level "high"
 	}
+	models "deepseek-v4.1-flash-1017" {
+		routing-family "factory"
+		upstream-rotation "fireworks" "baseten" "databricks"
+		credit-rates { input 0.12; output 4; cache-read 0.024 }
+		// No bundled Fireworks row for the 1017 slug; Fireworks prices it like V4.1 Flash.
+		list-price-from "fireworks" "deepseek-v4.1-flash"
+		thinking-efforts "low" "high" "max"
+		thinking-requires-effort #false
+		thinking-default-level "high"
+		entitlement { feature-flag "deepseek_v4_1_flash_1017" }
+	}
 	models "deepseek-v4.1-flash" {
 		routing-family "factory"
 		upstream-rotation "fireworks" "baseten" "databricks"
@@ -1131,6 +1148,13 @@ provider "factory-droid" {
 			synthetic-reasoning-content-fallback " "
 		}
 	}
+	// Droid 0.250.0 configures the 1017 checkpoint's Baseten route with
+	// thinkingMode "reasoning-effort": thinking off is sent as `none`.
+	models "deepseek-v4.1-flash-1017" {
+		on-upstream "baseten" {
+			completions-reasoning-mode "effort"
+		}
+	}
 	// The replay codec follows the selected route, including EU GLM routes.
 	on-upstream "mistral" {
 		models "glm-*" "mistral-medium-3.5" "mistral-large-4" {

package/src/compat/rules/providers/google-vertex.kdl

@@ -4,6 +4,9 @@ provider "google-vertex" {
 	default-model "gemini-3.1-pro-preview"
 	allow-unauthenticated #true
 	image-count-budget 200
+	// Vertex publishes Anthropic ids pinned to a revision (`claude-opus-4-8@default`);
+	// `google-vertex/claude-opus-4-8` names that model exactly.
+	selector-revision-suffix "@default"
 
 	// Replaces google-shared.ts stripping functionCall/functionResponse ids for Vertex.
 	supports-function-part-id #false

package/src/compat/rules/providers/kimi-code.kdl

@@ -2,6 +2,7 @@
 
 provider "kimi-code" {
 	default-model "kimi-for-coding"
+	dynamic-models-authoritative #true
 	discovery label="Kimi Code" {
 		env "KIMI_API_KEY"
 	}
@@ -39,6 +40,11 @@ provider "kimi-code" {
 		thinking-efforts "minimal" "low" "medium" "high"
 		thinking-format "zai"
 	}
+	// K2.7 Code Highspeed always thinks (`supports_thinking_type: "only"`) but
+	// exposes no adjustable effort ladder (`think_efforts` is omitted).
+	models "kimi-for-coding-highspeed" {
+		thinking-efforts
+	}
 
 	// /coding/v1/models carries no pricing. `cost-fallback` seeds the public
 	// Moonshot list price only when discovery reports $0, so a future priced

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

@@ -29,6 +29,14 @@ provider "mistral" {
 	// without it, requests only hit replica-local implicit cache. Same shape as
 	// the kimi-code rule (#6049).
 	supports-prompt-cache-key #true
+	// The gateway accepts assistant reasoning as text, not DeepSeek's native
+	// reasoning_content field, even when it serves DeepSeek-family models.
+	requires-reasoning-content-for-tool-calls #false
+	requires-reasoning-content-for-all-assistant-turns #false
+	// Reasoning streams and replays as typed `thinking` content parts
+	// (https://docs.mistral.ai/studio/conversations/reasoning) for every
+	// reasoning model the host serves, including GLM and DeepSeek (#17375).
+	mistral-reasoning-content-parts #true
 	class "unknown" {
 		thinking-efforts "minimal" "low" "medium" "high" "xhigh"
 	}

package/src/compat/rules/runtime/behavior.kdl

@@ -100,8 +100,8 @@ behavior {
 	// (flat rate) or Standard Credits. Unknown ids have no inferred entitlement.
 	quota-tiers provider="factory-droid" {
 		tier "core" "inkling" "mistral-large-4" "mistral-medium-3.5" "glm-5.3-flash" "glm-5.3" "glm-5.2" "glm-5.2-fast" \
-			"kimi-k3" "qwen3.8-max" "nemotron-3-ultra" "deepseek-v4.1-flash" "deepseek-v4-flash-0731" \
-			"deepseek-v4-pro" "minimax-m3" "minimax-m2.7"
+			"kimi-k3" "qwen3.8-max" "nemotron-3-ultra" "deepseek-v4.1-flash" "deepseek-v4.1-flash-1017" \
+			"deepseek-v4-flash-0731" "deepseek-v4-pro" "minimax-m3" "minimax-m2.7"
 		tier "standard" "claude-fable-5.1" "claude-fable-5" "claude-opus-5-5" "claude-opus-5-5-fast" \
 			"claude-opus-5" "claude-opus-5-fast" "claude-opus-4-8" "claude-opus-4-8-fast" "claude-opus-4-7" \
 			"claude-opus-4-6" "claude-opus-4-5-20251101" "claude-sonnet-5-5" "claude-sonnet-5" \
@@ -315,8 +315,8 @@ behavior {
 			exact="gemini-3-flash-preview" exact="gemini-3-pro-image-preview"
 		route "openai-completions" exact="inkling" exact="mistral-large-4" exact="mistral-medium-3.5" exact="glm-5.3-flash" \
 			exact="glm-5.3" exact="glm-5.2" exact="glm-5.2-fast" exact="kimi-k3" exact="qwen3.8-max" \
-			exact="nemotron-3-ultra" exact="deepseek-v4.1-flash" exact="deepseek-v4-flash-0731" \
-			exact="deepseek-v4-pro" exact="minimax-m3"
+			exact="nemotron-3-ultra" exact="deepseek-v4.1-flash" exact="deepseek-v4.1-flash-1017" \
+			exact="deepseek-v4-flash-0731" exact="deepseek-v4-pro" exact="minimax-m3"
 	}
 
 	// Copilot serves these models with tighter effective limits than its
@@ -428,7 +428,6 @@ behavior {
 		model "oss-120b"
 		model "cerebras/zai-glm-4.7"
 		model "cerebras/zai-glm-4.6"
-		model "cerebras/zai-glm"
 		model "claude-haiku-4-5"
 		model "claude-haiku-4.5"
 		model "haiku-4-5"
@@ -463,7 +462,6 @@ behavior {
 		model "anthropic/claude-opus-5"
 		model "claude-opus-5"
 		model "opus-5"
-		model "openai-codex/gpt-5.5"
 		model "gpt-5.5"
 		model "claude-opus-4-8"
 		model "claude-opus-4.8"
@@ -479,6 +477,11 @@ behavior {
 		model "openrouter/google/gemini-3-pro-image"
 		model "google/gemini-3-pro-image"
 		model "deepinfra/black-forest-labs/FLUX-2-pro"
+		// Factory Droid last, in the CLI's own preference order: a Factory
+		// login alone can generate, and no other provider's turn moves.
+		model "factory-droid/gpt-image-2.5-sunburst"
+		model "factory-droid/gpt-image-2.5-flare"
+		model "factory-droid/gemini-3-pro-image-preview"
 	}
 	role-defaults "web" {
 		model "web/parallel"