@oh-my-pi/pi-catalog 18.1.14 → 18.1.16

package/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+## [18.1.16] - 2026-09-09
+
+- Updated Fire Pass (`firepass`) login validation probe to `accounts/fireworks/routers/glm-5p2-fast` and bundled `glm-5.2-fast` and `kimi-k3-fast` models in place of decommissioned `kimi-k2.6-turbo` ([#10859](https://github.com/can1357/oh-my-pi/pull/10859) by [@olegpulatov](https://github.com/olegpulatov)).
+
 ## [18.1.14] - 2026-09-07
 
 ### Fixed

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

@@ -180,6 +180,8 @@ export interface CompiledRule {
     source: string;
     class?: string;
     providers?: string[];
+    /** Request adapter identifiers matched by an `on-api` selector. */
+    apis?: string[];
     family?: string;
     revision?: CompiledRevisionTerm[];
     models?: CompiledSelector[];
@@ -549,6 +551,8 @@ export interface ModelIdentity {
 export interface ResolveTarget {
     /** Deployment provider hosting the model. */
     provider: string;
+    /** Request adapter used to serialize the model. */
+    api: string;
     /** Centrally classified vendor lineage. */
     class: string;
     /** Classified product family within the class, when known. */

package/dist/types/fireworks-model-id.d.ts

@@ -1,10 +1,10 @@
 export declare function toFireworksPublicModelId(modelId: string): string;
 export declare function toFireworksWireModelId(modelId: string): string;
 /**
- * Fire Pass exposes its Kimi K2.6 Turbo subscription through a dedicated router
- * endpoint at `accounts/fireworks/routers/<id>` rather than the `models/` namespace.
- * We keep a friendly public id (e.g. `kimi-k2.6-turbo`) in the catalog and translate
- * to the wire form (`accounts/fireworks/routers/kimi-k2p6-turbo`) at request time.
+ * Fire Pass exposes subscription models through dedicated router endpoints
+ * at `accounts/fireworks/routers/<id>` rather than the `models/` namespace.
+ * We keep a friendly public id (e.g. `glm-5.2-fast`, `kimi-k3-fast`) in the catalog
+ * and translate to the wire form (`accounts/fireworks/routers/glm-5p2-fast`) at request time.
  */
 export declare function toFirepassPublicModelId(modelId: string): string;
 export declare function toFirepassWireModelId(modelId: string): string;

package/dist/types/provider-models/descriptors.d.ts

@@ -147,7 +147,7 @@ export declare const CATALOG_PROVIDERS: readonly [{
     };
 }, {
     readonly id: "firepass";
-    readonly defaultModel: "kimi-k2.6-turbo";
+    readonly defaultModel: "glm-5.2-fast";
     readonly envVars: readonly ["FIREPASS_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("../index.js").ModelManagerOptions<"openai-completions", unknown>;
 }, {

package/dist/types/provider-models/openai-compat.d.ts

@@ -332,14 +332,15 @@ export interface FireworksModelManagerConfig {
     fetch?: FetchImpl;
 }
 export declare function fireworksModelManagerOptions(config?: FireworksModelManagerConfig): ModelManagerOptions<"openai-completions">;
+export declare const FIREPASS_STATIC_MODELS: readonly ModelSpec<"openai-completions">[];
 export interface FirepassModelManagerConfig {
     apiKey?: string;
     baseUrl?: string;
     fetch?: FetchImpl;
 }
 /**
- * Fire Pass is a Fireworks subscription product that exposes a single router
- * model (Kimi K2.6 Turbo) under `accounts/fireworks/routers/kimi-k2p6-turbo`.
+ * Fire Pass is a Fireworks subscription product that exposes router models
+ * (GLM 5.2 Fast, Kimi K3 Fast) under `accounts/fireworks/routers/<id>`.
  * The dedicated `fpk_…` keys do not authorize `/v1/models`, so this manager
  * never performs dynamic discovery — the bundled catalog entry is canonical.
  * See https://docs.fireworks.ai/firepass.

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.1.14",
+	"version": "18.1.16",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Stencil Labs, Inc.",
@@ -35,12 +35,12 @@
 		"gen:proto": "bun scripts/generate-protocols.ts"
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.1.14",
-		"@oh-my-pi/pi-utils": "18.1.14"
+		"@oh-my-pi/omptype": "18.1.16",
+		"@oh-my-pi/pi-utils": "18.1.16"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.1.14",
+		"@oh-my-pi/pi-ai": "18.1.16",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/cascade.ts

@@ -3,7 +3,7 @@
  * for one structured model target from the compiled rule tree.
  *
  * Faithful port of the o2 reference resolver (`cascade.rs`): rules are
- * conjunctions over `(class, provider, family, revision, models)`; per axis
+ * conjunctions over `(class, provider, api, family, revision, models)`; per axis
  * the matching rule with the greatest `(model-selector exactness,
  * constrained-dimension count, priority)` tuple wins, and an equal-tuple
  * same-axis contest throws {@link AmbiguousOverlapError}. Declaration and
@@ -74,6 +74,7 @@ function buildRuleIndex(cascade: CompiledCascade): IndexedRule[] {
 		const dimensions =
 			Number(compiled.class !== undefined) +
 			Number(compiled.providers !== undefined) +
+			Number(compiled.apis !== undefined) +
 			Number(compiled.family !== undefined) +
 			Number(compiled.revision !== undefined) +
 			Number(compiled.models !== undefined);
@@ -117,6 +118,7 @@ function rankRule(
 	const { compiled } = rule;
 	if (compiled.class !== undefined && compiled.class !== target.class) return undefined;
 	if (compiled.providers !== undefined && !compiled.providers.includes(target.provider)) return undefined;
+	if (compiled.apis !== undefined && !compiled.apis.includes(target.api)) return undefined;
 	if (compiled.family !== undefined && compiled.family !== target.family) return undefined;
 	if (rule.revision !== undefined && (!revision || !revisionSatisfies(revision, rule.revision))) return undefined;
 	let exactness = 0;

package/src/compat/delegation.ts

@@ -27,6 +27,7 @@ export function resolveDelegationBias(model: Model): DelegationBias {
 	const { identity } = model;
 	const bias = resolveCascade({
 		provider: model.provider,
+		api: model.api,
 		class: identity.class,
 		model: model.id,
 		reasoning: Boolean(model.reasoning),

package/src/compat/resolve.ts

@@ -1173,6 +1173,7 @@ function fillExplicitThinking<TApi extends Api>(
 function buildResolveTarget<TApi extends Api>(spec: ModelSpec<TApi>, identity: ModelIdentity): ResolveTarget {
 	const target: ResolveTarget = {
 		provider: spec.provider,
+		api: spec.api,
 		class: identity.class,
 		model: spec.id,
 		reasoning: Boolean(spec.reasoning),

package/src/compat/rules/README.md

@@ -5,7 +5,7 @@ This tree is the checked-in source of model identity and compatibility policy. A
 There are three ownership strata:
 
 - `taxonomy/*.kdl` defines identity: class membership, product families, revision extraction, reviewed exact corrections, and suffix collapse.
-- `classes/*.kdl` defines model-lineage truths: behavior inherent to a model line, optionally scoped to the providers where the census established it.
+- `classes/*.kdl` defines model-lineage truths: behavior inherent to a model line, optionally scoped to the providers or request adapters where the census established it.
 - `providers/*.kdl` defines deployment contracts: behavior imposed by a host, plus documented per-model residue that taxonomy cannot express exactly.
 - `runtime/behavior.kdl` defines heuristics used before or outside exact model lookup: responses routing, API routes, quota tiers, plan requirements, model limits, roster exclusions, hosted defaults, pricing peers.
 - `auth/<provider>.kdl` defines the provider's auth contract: display name, env-var fallback, credential storage/format, and the declarative login / refresh flow that `@oh-my-pi/pi-ai`'s registry engines interpret (see [Auth grammar](#auth-grammar)).
@@ -151,10 +151,15 @@ discovery {
 
 ## Cascade grammar
 
-A cascade document starts with `class` or `provider`. Every selector adds a conjunct to the current rule. Axis directives may appear directly in any permitted scope, and nested selector blocks may appear alongside them.
+A cascade document starts with `class` or `provider`. Every selector adds a conjunct to the current rule. `on` scopes by deployment provider; `on-api` scopes by request adapter, including custom provider names. Axis directives may appear directly in any permitted scope, and nested selector blocks may appear alongside them.
 
 ```kdl
 class "gemini" {
+    on-api "google-generative-ai" {
+        revision ">=3" {
+            requires-skip-thought-signature #true
+        }
+    }
     on "google" "google-vertex" "openrouter" {
         family "flash" {
             revision ">=2.5 <3.8" {
@@ -175,14 +180,15 @@ provider "openrouter" {
 
 | Selector | Form | Matching semantics |
 | --- | --- | --- |
-| `class` | `class "id" { ... }` | Exact class ID. At document root it may contain `on`, `family`, `revision`, and `models`. Under `provider` it may contain `family`, `revision`, and `models`. |
+| `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` and `models`. |
 | `on` | `on "provider-a" "provider-b" { ... }` | One or more provider IDs, combined as OR. It is allowed only under a root `class`, and may contain `family`, `revision`, and `models`. |
+| `on-api` | `on-api "adapter-a" "adapter-b" { ... }` | One or more request adapter IDs, combined as OR. It is allowed only under a root `class`, and 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 cannot contain another selector. `token="name"` matches an ASCII-case-insensitive token bounded by non-alphanumerics. |
 
-Class, provider/`on`, and family selector 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.
+Class, provider/`on`, `on-api`, and family selector 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.
 
 A `models` string without `*` is an exact, case-sensitive match against the provider-relative model identifier. A string containing `*` is an anchored, ASCII-case-insensitive wildcard match. Prefer taxonomy ranks; retain exact/glob lists only when they isolate the census member set exactly, and keep a `// residue:` comment explaining why ranks do not.
 
@@ -212,7 +218,7 @@ Rules resolve independently per axis. A matching rule is ranked by:
 The tuple is compared lexicographically, greatest first:
 
 - model exactness is `2` when any matching `models` selector is exact, `1` when the best matching selector is a glob or token, and `0` when the rule has no `models` selector;
-- dimension count is the number of present dimensions among class, provider/`on`, family, revision, and models;
+- dimension count is the number of present dimensions among class, provider/`on`, API/`on-api`, family, revision, and models;
 - priority is the local block's `priority`, defaulting to `0`.
 
 The highest-ranked matching assignment wins for that axis. Two distinct rules that tie on all three components and assign the same axis are an ambiguity error even if their values are equal. File and declaration order never resolve the tie; add an explicit priority only after confirming the overlap is intentional.

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

@@ -1,14 +1,13 @@
 auth "firepass" {
-	name "Fire Pass (Fireworks Kimi K2.6 Turbo subscription)"
+	name "Fire Pass (Fireworks subscription)"
 	login "api-key" {
 		// Fire Pass is a Fireworks subscription product whose dedicated `fpk_…` API
-		// keys are scoped to the `accounts/fireworks/routers/kimi-k2p6-turbo` router
-		// (Kimi K2.6 Turbo). The key does NOT authorize `/v1/models`, so validation
-		// pings the chat completions endpoint with the router id directly.
+		// keys are scoped to router endpoints. The key does NOT authorize `/v1/models`,
+		// so validation pings the chat completions endpoint with the router id directly.
 		// See https://docs.fireworks.ai/firepass.
 		auth-url "https://app.fireworks.ai/settings/users/api-keys"
 		instructions "Create a dedicated Fire Pass API key in the Fireworks dashboard"
 		prompt "Paste your Fire Pass API key" placeholder="fpk_..."
-		validate "chat-completions" label="Fire Pass" base-url="https://api.fireworks.ai/inference/v1" model="accounts/fireworks/routers/kimi-k2p6-turbo"
+		validate "chat-completions" label="Fire Pass" base-url="https://api.fireworks.ai/inference/v1" model="accounts/fireworks/routers/glm-5p2-fast"
 	}
 }

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

@@ -88,9 +88,11 @@ class "gemini" {
 		// Replaces the Gemini 3+ mandatory-reasoning fallback.
 		thinking-requires-effort #true
 	}
-	// Public Gemini requires the bypass sentinel on every unsigned call. The
-	// Cloud Code Assist first-call variant is a host contract in providers/.
-	on "google" "opencode-zen" {
+	// Public Gemini requires the bypass sentinel on every unsigned call and
+	// supports function-part ids. Scope by adapter so custom providers using
+	// the same wire protocol inherit both contracts. The Cloud Code Assist
+	// first-call variant is a host contract in providers/.
+	on-api "google-generative-ai" {
 		revision ">=3" {
 			requires-skip-thought-signature #true
 			supports-function-part-id #true

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

@@ -7,9 +7,13 @@ provider "firepass" {
 	}
 	// Replaces the Fire Pass wire-model-id dispatch baseline.
 	wire-model-id-mode "firepass"
-	// residue: taxonomy ranks and exact globs do not isolate these models.
-	models "kimi-k2.6-turbo" {
-		thinking-efforts "low" "medium" "high" "xhigh" "max"
+	class "glm" priority=1 {
 		thinking-mode "effort"
+		thinking-efforts "minimal" "low" "medium" "high" "max"
+	}
+
+	class "kimi" priority=1 {
+		thinking-mode "effort"
+		thinking-efforts "low" "high" "max"
 	}
 }

package/src/compat/rules.json

@@ -4822,11 +4822,10 @@
 				}
 			},
 			{
-				"source": "classes/gemini.kdl:94",
+				"source": "classes/gemini.kdl:96",
 				"class": "gemini",
-				"providers": [
-					"google",
-					"opencode-zen"
+				"apis": [
+					"google-generative-ai"
 				],
 				"revision": [
 					{
@@ -4840,7 +4839,7 @@
 				}
 			},
 			{
-				"source": "classes/gemini.kdl:100",
+				"source": "classes/gemini.kdl:102",
 				"class": "gemini",
 				"providers": [
 					"google-antigravity",
@@ -7997,25 +7996,37 @@
 				}
 			},
 			{
-				"source": "providers/firepass.kdl:11",
+				"source": "providers/firepass.kdl:10",
+				"class": "glm",
 				"providers": [
 					"firepass"
 				],
-				"models": [
-					{
-						"kind": "exact",
-						"value": "kimi-k2.6-turbo"
-					}
-				],
+				"priority": 1,
 				"thinking": {
+					"mode": "effort",
 					"efforts": [
+						"minimal",
 						"low",
 						"medium",
 						"high",
-						"xhigh",
 						"max"
-					],
-					"mode": "effort"
+					]
+				}
+			},
+			{
+				"source": "providers/firepass.kdl:15",
+				"class": "kimi",
+				"providers": [
+					"firepass"
+				],
+				"priority": 1,
+				"thinking": {
+					"mode": "effort",
+					"efforts": [
+						"low",
+						"high",
+						"max"
+					]
 				}
 			},
 			{
@@ -16406,7 +16417,7 @@
 			{
 				"id": "firepass",
 				"apiKeyFormat": "bearer",
-				"name": "Fire Pass (Fireworks Kimi K2.6 Turbo subscription)",
+				"name": "Fire Pass (Fireworks subscription)",
 				"login": {
 					"kind": "api-key",
 					"authUrl": "https://app.fireworks.ai/settings/users/api-keys",
@@ -16417,7 +16428,7 @@
 						"kind": "chat-completions",
 						"label": "Fire Pass",
 						"baseUrl": "https://api.fireworks.ai/inference/v1",
-						"model": "accounts/fireworks/routers/kimi-k2p6-turbo"
+						"model": "accounts/fireworks/routers/glm-5p2-fast"
 					}
 				}
 			},

package/src/compat/types.ts

@@ -198,6 +198,8 @@ export interface CompiledRule {
 	source: string;
 	class?: string;
 	providers?: string[];
+	/** Request adapter identifiers matched by an `on-api` selector. */
+	apis?: string[];
 	family?: string;
 	revision?: CompiledRevisionTerm[];
 	models?: CompiledSelector[];
@@ -568,6 +570,8 @@ export interface ModelIdentity {
 export interface ResolveTarget {
 	/** Deployment provider hosting the model. */
 	provider: string;
+	/** Request adapter used to serialize the model. */
+	api: string;
 	/** Centrally classified vendor lineage. */
 	class: string;
 	/** Classified product family within the class, when known. */

package/src/fireworks-model-id.ts

@@ -14,10 +14,10 @@ export function toFireworksWireModelId(modelId: string): string {
 }
 
 /**
- * Fire Pass exposes its Kimi K2.6 Turbo subscription through a dedicated router
- * endpoint at `accounts/fireworks/routers/<id>` rather than the `models/` namespace.
- * We keep a friendly public id (e.g. `kimi-k2.6-turbo`) in the catalog and translate
- * to the wire form (`accounts/fireworks/routers/kimi-k2p6-turbo`) at request time.
+ * Fire Pass exposes subscription models through dedicated router endpoints
+ * at `accounts/fireworks/routers/<id>` rather than the `models/` namespace.
+ * We keep a friendly public id (e.g. `glm-5.2-fast`, `kimi-k3-fast`) in the catalog
+ * and translate to the wire form (`accounts/fireworks/routers/glm-5p2-fast`) at request time.
  */
 export function toFirepassPublicModelId(modelId: string): string {
 	const stripped = modelId.startsWith(FIREPASS_WIRE_PREFIX) ? modelId.slice(FIREPASS_WIRE_PREFIX.length) : modelId;

package/src/models.json

@@ -70726,41 +70726,138 @@
 		}
 	},
 	"firepass": {
-		"kimi-k2.6-turbo": {
-			"id": "kimi-k2.6-turbo",
-			"name": "Kimi K2.6 Turbo (Fire Pass)",
+		"glm-5.2-fast": {
+			"id": "glm-5.2-fast",
+			"name": "GLM 5.2 Fast (Fire Pass)",
 			"api": "openai-completions",
 			"provider": "firepass",
 			"baseUrl": "https://api.fireworks.ai/inference/v1",
 			"reasoning": true,
 			"input": [
-				"text",
-				"image"
+				"text"
 			],
 			"cost": {
-				"input": 0,
-				"output": 0,
-				"cacheRead": 0,
+				"input": 2.1,
+				"output": 6.6,
+				"cacheRead": 0.21,
 				"cacheWrite": 0
 			},
-			"contextWindow": 262144,
-			"maxTokens": 32768,
+			"contextWindow": 1048576,
+			"maxTokens": 131072,
+			"identity": {
+				"class": "glm",
+				"revision": "5.2.0"
+			},
+			"requiresGlyphTokenization": false,
+			"tokenizer": "glm5",
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
+					"minimal",
 					"low",
 					"medium",
 					"high",
-					"xhigh",
 					"max"
-				]
+				],
+				"effortMap": {
+					"minimal": "none"
+				}
 			},
-			"tokenizer": "kimi-k2",
-			"requiresGlyphTokenization": false,
+			"supportsComputerUse": false,
+			"compat": {
+				"supportsStore": false,
+				"supportsDeveloperRole": false,
+				"supportsMultipleSystemMessages": true,
+				"supportsReasoningEffort": true,
+				"supportsReasoningParams": true,
+				"supportsSamplingParams": true,
+				"supportsPenaltyAndStopParams": true,
+				"reasoningEffortMap": {},
+				"supportsUsageInStreaming": true,
+				"alwaysSendMaxTokens": false,
+				"disableReasoningOnForcedToolChoice": false,
+				"disableReasoningOnToolChoice": false,
+				"supportsToolChoice": true,
+				"supportsForcedToolChoice": true,
+				"supportsNamedToolChoice": true,
+				"maxTokensField": "max_tokens",
+				"requiresToolResultName": false,
+				"requiresAssistantAfterToolResult": false,
+				"requiresThinkingAsText": false,
+				"requiresMistralToolIds": false,
+				"thinkingFormat": "openai",
+				"reasoningDisableMode": "lowest-effort",
+				"omitReasoningEffort": false,
+				"includeEncryptedReasoning": true,
+				"filterReasoningHistory": false,
+				"reasoningContentField": "reasoning_content",
+				"requiresReasoningContentForToolCalls": false,
+				"requiresReasoningContentForAllAssistantTurns": false,
+				"allowsSyntheticReasoningContentForToolCalls": true,
+				"replayReasoningContent": false,
+				"qwenPreserveThinking": false,
+				"qwenTemplateReasoningEffort": false,
+				"requiresAssistantContentForToolCalls": false,
+				"supportsPromptCacheBreakpoints": false,
+				"isOpenRouterHost": false,
+				"wireModelIdMode": "firepass",
+				"isVercelGatewayHost": false,
+				"supportsStrictMode": false,
+				"toolStrictMode": "mixed",
+				"stripDeepseekSpecialTokens": false,
+				"streamMarkupHealingPattern": "thinking",
+				"reasoningDeltasMayBeCumulative": false,
+				"emptyLengthFinishIsContextError": false,
+				"usesOpenAIToolCallIdLimit": false,
+				"dropThinkingWhenReasoningEffort": false,
+				"nativeKimiK3Reasoning": false,
+				"zaiReasoningEffortDialect": false,
+				"clampOutputToModelMax": false,
+				"stripImageInput": false,
+				"rejectRootObjectUnion": false,
+				"retryWithoutStrictOnGrammarError": false,
+				"supportsPromptCacheKey": false
+			}
+		},
+		"kimi-k3-fast": {
+			"id": "kimi-k3-fast",
+			"name": "Kimi K3 Fast (Fire Pass)",
+			"api": "openai-completions",
+			"provider": "firepass",
+			"baseUrl": "https://api.fireworks.ai/inference/v1",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 4.5,
+				"output": 22.5,
+				"cacheRead": 0.45,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 131072,
 			"identity": {
 				"class": "kimi",
-				"family": "k2.6"
+				"family": "k3"
 			},
+			"requiresGlyphTokenization": false,
+			"tokenizer": "kimi-k2",
+			"thinking": {
+				"mode": "effort",
+				"efforts": [
+					"low",
+					"high",
+					"max"
+				],
+				"defaultLevel": "max",
+				"effortMap": {
+					"max": "max"
+				},
+				"requiresEffort": true
+			},
+			"supportsComputerUse": false,
 			"compat": {
 				"supportsStore": false,
 				"supportsDeveloperRole": false,
@@ -70769,7 +70866,12 @@
 				"supportsReasoningParams": true,
 				"supportsSamplingParams": true,
 				"supportsPenaltyAndStopParams": true,
-				"reasoningEffortMap": {},
+				"reasoningEffortMap": {
+					"minimal": "low",
+					"medium": "high",
+					"xhigh": "max",
+					"max": "max"
+				},
 				"supportsUsageInStreaming": true,
 				"alwaysSendMaxTokens": true,
 				"disableReasoningOnForcedToolChoice": true,
@@ -70802,9 +70904,8 @@
 				"supportsStrictMode": false,
 				"toolStrictMode": "mixed",
 				"toolSchemaFlavor": "moonshot-mfjs",
-				"streamIdleTimeoutMs": 300000,
 				"stripDeepseekSpecialTokens": false,
-				"streamMarkupHealingPattern": "kimi",
+				"streamMarkupHealingPattern": "thinking",
 				"reasoningDeltasMayBeCumulative": false,
 				"emptyLengthFinishIsContextError": false,
 				"usesOpenAIToolCallIdLimit": false,
@@ -70816,8 +70917,7 @@
 				"rejectRootObjectUnion": false,
 				"retryWithoutStrictOnGrammarError": false,
 				"supportsPromptCacheKey": false
-			},
-			"supportsComputerUse": false
+			}
 		}
 	},
 	"fireworks": {
@@ -390874,4 +390974,4 @@
 			}
 		}
 	}
-}
+}

package/src/provider-models/descriptors.ts

@@ -196,7 +196,7 @@ export const CATALOG_PROVIDERS = [
 	},
 	{
 		id: "firepass",
-		defaultModel: "kimi-k2.6-turbo",
+		defaultModel: "glm-5.2-fast",
 		envVars: ["FIREPASS_API_KEY"],
 		createModelManagerOptions: (config: ModelManagerConfig) => firepassModelManagerOptions(config),
 	},

package/src/provider-models/openai-compat.ts

@@ -2432,9 +2432,40 @@ export function fireworksModelManagerOptions(
 }
 
 // ---------------------------------------------------------------------------
-// 7.6 Fire Pass (Fireworks Kimi K2.6 Turbo subscription)
+// 7.6 Fire Pass (Fireworks subscription)
 // ---------------------------------------------------------------------------
 
+// Pricing and limits are the published Fire Pass router tariff
+// (https://docs.fireworks.ai/firepass), not upstream-discoverable: dedicated
+// `fpk_…` keys never authorize `/v1/models`, so this seed is the authoritative
+// source. cacheRead is 0.1x input; cacheWrite stays 0 (not billed separately).
+export const FIREPASS_STATIC_MODELS: readonly ModelSpec<"openai-completions">[] = [
+	{
+		id: "glm-5.2-fast",
+		name: "GLM 5.2 Fast (Fire Pass)",
+		api: "openai-completions",
+		provider: "firepass",
+		baseUrl: "https://api.fireworks.ai/inference/v1",
+		reasoning: true,
+		input: ["text"],
+		cost: { input: 2.1, output: 6.6, cacheRead: 0.21, cacheWrite: 0 },
+		contextWindow: 1_048_576,
+		maxTokens: 131_072,
+	},
+	{
+		id: "kimi-k3-fast",
+		name: "Kimi K3 Fast (Fire Pass)",
+		api: "openai-completions",
+		provider: "firepass",
+		baseUrl: "https://api.fireworks.ai/inference/v1",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: { input: 4.5, output: 22.5, cacheRead: 0.45, cacheWrite: 0 },
+		contextWindow: 1_048_576,
+		maxTokens: 131_072,
+	},
+];
+
 export interface FirepassModelManagerConfig {
 	apiKey?: string;
 	baseUrl?: string;
@@ -2442,8 +2473,8 @@ export interface FirepassModelManagerConfig {
 }
 
 /**
- * Fire Pass is a Fireworks subscription product that exposes a single router
- * model (Kimi K2.6 Turbo) under `accounts/fireworks/routers/kimi-k2p6-turbo`.
+ * Fire Pass is a Fireworks subscription product that exposes router models
+ * (GLM 5.2 Fast, Kimi K3 Fast) under `accounts/fireworks/routers/<id>`.
  * The dedicated `fpk_…` keys do not authorize `/v1/models`, so this manager
  * never performs dynamic discovery — the bundled catalog entry is canonical.
  * See https://docs.fireworks.ai/firepass.