@oh-my-pi/pi-catalog 18.1.18 → 18.1.20

package/CHANGELOG.md

@@ -2,6 +2,24 @@
 
 ## [Unreleased]
 
+## [18.1.20] - 2026-09-13
+
+### Fixed
+
+- Fixed DeepSeek V4.1 Flash dropping image attachments on hosts without their own carve-out, so the natively multimodal model is no longer treated as text-only ([#11992](https://github.com/can1357/oh-my-pi/pull/11992) by [@infernix](https://github.com/infernix)).
+- LiteLLM model groups keep the image input their deployment declares, instead of having attachments replaced with the "model does not support vision" placeholder ([#11982](https://github.com/can1357/oh-my-pi/issues/11982), [#11985](https://github.com/can1357/oh-my-pi/pull/11985) by [@lz37](https://github.com/lz37)).
+- Fixed OpenCode Zen/Go Muse Spark models failing every tool-call turn with a 400 "reasoning encrypted_content was not issued to this caller" error: the gateways proxy the Responses lane to Meta but can't round-trip encrypted reasoning, so those SKUs no longer request or replay it ([#11928](https://github.com/can1357/oh-my-pi/issues/11928)).
+
+## [18.1.19] - 2026-09-12
+
+### Added
+
+- Added Charm Hyper as a built-in provider with API-key login, live model discovery, and per-model pricing, effort ladders, and limits read straight from its catalog ([#11656](https://github.com/can1357/oh-my-pi/pull/11656) by [@oldschoola](https://github.com/oldschoola)).
+
+### Fixed
+
+- `anthropic/claude-fable-5-1` cache reads now cost Anthropic's published $0.25/MTok instead of $1.00, so session cost and usage reports no longer overstate cache-read spend by 4x ([#11862](https://github.com/can1357/oh-my-pi/pull/11862) by [@camjac251](https://github.com/camjac251)).
+
 ## [18.1.18] - 2026-09-11
 
 ### Added
@@ -13,6 +31,7 @@
 - OpenCode Go's DeepSeek Flash lanes (`deepseek-flash`, `deepseek-v4.1-flash`) now declare image input. The gateway serves them with vision despite the IDs carrying no vision suffix, so the class-wide `strip-image-input` rule was dropping attachments the endpoint reads; the modality is declared too, since live discovery seeds these lanes text-only ([#11774](https://github.com/can1357/oh-my-pi/pull/11774) by [@STRML](https://github.com/STRML)).
 - Amazon Bedrock OpenAI models, plus unclassified profiles such as opaque application-inference-profile ARNs, now carry the compatibility policy required to preserve image-bearing tool results ([#11681](https://github.com/can1357/oh-my-pi/issues/11681)).
 - DeepSeek V4.1 Flash requests now honor the documented 384K output maximum instead of being capped at 64K ([#11769](https://github.com/can1357/oh-my-pi/issues/11769)).
+- Fixed the first-party `deepseek-flash` alias missing the V4.1 Flash wire contract: it now sends `max_tokens` with `reasoning_content` and replays reasoning and assistant content on tool calls with no tool choice ([#11799](https://github.com/can1357/oh-my-pi/pull/11799) by [@brit](https://github.com/brit)).
 
 ## [18.1.17] - 2026-09-10
 

package/dist/types/compat/auth-ids.d.ts

@@ -1,4 +1,4 @@
 /** Every provider with an `auth/<id>.kdl` policy. */
-export type AuthProviderId = "abliteration" | "aiand" | "aimlapi" | "alibaba-coding-plan" | "alibaba-token-plan" | "amazon-bedrock" | "anthropic" | "azure" | "baseten" | "bedrock-mantle" | "cerebras" | "cline-pass" | "cloudflare-ai-gateway" | "commandcode" | "coreweave" | "cursor" | "deepinfra" | "deepseek" | "devin" | "exa" | "firepass" | "fireworks" | "github-copilot" | "gitlab-duo" | "gitlab-duo-agent" | "gmi-cloud" | "google" | "google-antigravity" | "google-gemini-cli" | "google-vertex" | "groq" | "huggingface" | "kagi" | "kilo" | "kimi-code" | "litellm" | "llama.cpp" | "lm-studio" | "meta" | "minimax" | "minimax-code" | "minimax-code-cn" | "mistral" | "moonshot" | "muse-code" | "nanogpt" | "novita" | "nvidia" | "ollama" | "ollama-cloud" | "openai" | "openai-codex" | "openai-codex-device" | "opencode-go" | "opencode-zen" | "openrouter" | "parallel" | "perplexity" | "qianfan" | "qwen-portal" | "sakana" | "siliconflow" | "siliconflow-cn" | "synthetic" | "tavily" | "together" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "xai" | "xai-oauth" | "xiaomi" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "xiaomi-token-plan-sgp" | "yolo-auto" | "zai" | "zai-coding-plan" | "zenmux" | "zhipu-coding-plan";
+export type AuthProviderId = "abliteration" | "aiand" | "aimlapi" | "alibaba-coding-plan" | "alibaba-token-plan" | "amazon-bedrock" | "anthropic" | "azure" | "baseten" | "bedrock-mantle" | "cerebras" | "charm-hyper" | "cline-pass" | "cloudflare-ai-gateway" | "commandcode" | "coreweave" | "cursor" | "deepinfra" | "deepseek" | "devin" | "exa" | "firepass" | "fireworks" | "github-copilot" | "gitlab-duo" | "gitlab-duo-agent" | "gmi-cloud" | "google" | "google-antigravity" | "google-gemini-cli" | "google-vertex" | "groq" | "huggingface" | "kagi" | "kilo" | "kimi-code" | "litellm" | "llama.cpp" | "lm-studio" | "meta" | "minimax" | "minimax-code" | "minimax-code-cn" | "mistral" | "moonshot" | "muse-code" | "nanogpt" | "novita" | "nvidia" | "ollama" | "ollama-cloud" | "openai" | "openai-codex" | "openai-codex-device" | "opencode-go" | "opencode-zen" | "openrouter" | "parallel" | "perplexity" | "qianfan" | "qwen-portal" | "sakana" | "siliconflow" | "siliconflow-cn" | "synthetic" | "tavily" | "together" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "xai" | "xai-oauth" | "xiaomi" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "xiaomi-token-plan-sgp" | "yolo-auto" | "zai" | "zai-coding-plan" | "zenmux" | "zhipu-coding-plan";
 /** Providers whose policy declares a `login` flow (the `/login` roster). */
-export type LoginProviderId = "abliteration" | "aiand" | "alibaba-coding-plan" | "alibaba-token-plan" | "anthropic" | "baseten" | "cerebras" | "cline-pass" | "cloudflare-ai-gateway" | "commandcode" | "coreweave" | "cursor" | "deepinfra" | "deepseek" | "devin" | "exa" | "firepass" | "fireworks" | "github-copilot" | "gitlab-duo" | "gitlab-duo-agent" | "gmi-cloud" | "google-antigravity" | "google-gemini-cli" | "huggingface" | "kagi" | "kilo" | "kimi-code" | "litellm" | "llama.cpp" | "lm-studio" | "meta" | "minimax-code" | "minimax-code-cn" | "moonshot" | "muse-code" | "nanogpt" | "novita" | "nvidia" | "ollama" | "ollama-cloud" | "openai-codex" | "openai-codex-device" | "opencode-go" | "opencode-zen" | "openrouter" | "parallel" | "perplexity" | "qianfan" | "qwen-portal" | "sakana" | "siliconflow" | "siliconflow-cn" | "synthetic" | "tavily" | "together" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "xai" | "xai-oauth" | "xiaomi" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "xiaomi-token-plan-sgp" | "yolo-auto" | "zai" | "zai-coding-plan" | "zenmux" | "zhipu-coding-plan";
+export type LoginProviderId = "abliteration" | "aiand" | "alibaba-coding-plan" | "alibaba-token-plan" | "anthropic" | "baseten" | "cerebras" | "charm-hyper" | "cline-pass" | "cloudflare-ai-gateway" | "commandcode" | "coreweave" | "cursor" | "deepinfra" | "deepseek" | "devin" | "exa" | "firepass" | "fireworks" | "github-copilot" | "gitlab-duo" | "gitlab-duo-agent" | "gmi-cloud" | "google-antigravity" | "google-gemini-cli" | "huggingface" | "kagi" | "kilo" | "kimi-code" | "litellm" | "llama.cpp" | "lm-studio" | "meta" | "minimax-code" | "minimax-code-cn" | "moonshot" | "muse-code" | "nanogpt" | "novita" | "nvidia" | "ollama" | "ollama-cloud" | "openai-codex" | "openai-codex-device" | "opencode-go" | "opencode-zen" | "openrouter" | "parallel" | "perplexity" | "qianfan" | "qwen-portal" | "sakana" | "siliconflow" | "siliconflow-cn" | "synthetic" | "tavily" | "together" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "xai" | "xai-oauth" | "xiaomi" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "xiaomi-token-plan-sgp" | "yolo-auto" | "zai" | "zai-coding-plan" | "zenmux" | "zhipu-coding-plan";

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

@@ -88,6 +88,14 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly catalogDiscovery: {
         readonly label: "Cerebras";
     };
+}, {
+    readonly id: "charm-hyper";
+    readonly defaultModel: "glm-5.3";
+    readonly envVars: readonly ["CHARM_HYPER_API_KEY", "HYPER_API_KEY"];
+    readonly createModelManagerOptions: (config: ModelManagerConfig) => import("../index.js").ModelManagerOptions<"openai-completions", unknown>;
+    readonly allowUnauthenticated: true;
+    readonly dynamicModelsAuthoritative: true;
+    readonly skipCrossProviderReferenceFills: true;
 }, {
     readonly id: "cloudflare-ai-gateway";
     readonly defaultModel: "anthropic/claude-opus-4-8";

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

@@ -745,3 +745,22 @@ export interface CommandCodeModelManagerConfig {
  * or context window.
  */
 export declare function commandCodeModelManagerOptions(config?: CommandCodeModelManagerConfig): ModelManagerOptions<Api>;
+/**
+ * Configuration for the Charm Hyper model manager.
+ *
+ * `baseUrl` overrides the gateway root for tests and self-hosted proxies; a
+ * value that omits the `/v1` surface gains one, so a host-only override
+ * behaves like every sibling provider's.
+ */
+export interface CharmHyperModelManagerConfig {
+    apiKey?: string;
+    baseUrl?: string;
+    fetch?: FetchImpl;
+}
+/**
+ * Charm Hyper's gateway catalog. `/v1/models` is public and carries the live
+ * tariff, so discovery runs with or without a key and the snapshot is
+ * authoritative: a model the gateway stops serving is pruned rather than kept
+ * alive by a stale bundled row.
+ */
+export declare function charmHyperModelManagerOptions(config?: CharmHyperModelManagerConfig): ModelManagerOptions<"openai-completions">;

package/dist/types/wire/charm-hyper.d.ts

@@ -0,0 +1,21 @@
+/**
+ * Charm Hyper gateway endpoint, shared so a host migration — or a self-hosted
+ * proxy override — touches a single module.
+ */
+export declare const CHARM_HYPER_API_BASE_URL = "https://hyper.charm.land/v1";
+/**
+ * Resolve a configured Charm Hyper base URL onto the gateway's `/v1` surface.
+ *
+ * Every consumer must agree on this, because they key different things off the
+ * result: inference and discovery target it, `/usage` sends a bearer token to
+ * it, and the model cache namespace is hashed from it. Three separate copies
+ * previously disagreed for a blank value — the model manager treated it as
+ * absent and used the canonical host, while the others produced a bare `/v1` —
+ * so a whitespace-only override silently split inference, balance checks and
+ * caching across different endpoints.
+ *
+ * A blank or whitespace-only value therefore means "not configured" and
+ * resolves to the canonical host; anything else keeps its host and gains the
+ * `/v1` segment if it omits one.
+ */
+export declare function normalizeCharmHyperBaseUrl(baseUrl?: string): string;

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.1.18",
+	"version": "18.1.20",
 	"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": "18.1.18",
-		"@oh-my-pi/pi-utils": "18.1.18"
+		"@oh-my-pi/omptype": "18.1.20",
+		"@oh-my-pi/pi-utils": "18.1.20"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.1.18",
+		"@oh-my-pi/pi-ai": "18.1.20",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/auth-ids.ts

@@ -13,6 +13,7 @@ export type AuthProviderId =
 	| "baseten"
 	| "bedrock-mantle"
 	| "cerebras"
+	| "charm-hyper"
 	| "cline-pass"
 	| "cloudflare-ai-gateway"
 	| "commandcode"
@@ -94,6 +95,7 @@ export type LoginProviderId =
 	| "anthropic"
 	| "baseten"
 	| "cerebras"
+	| "charm-hyper"
 	| "cline-pass"
 	| "cloudflare-ai-gateway"
 	| "commandcode"

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

@@ -5,7 +5,7 @@ login-order "openai-codex" "anthropic" "zai" "zai-coding-plan" "kimi-code" "open
 	"cursor" "devin" "google-antigravity" "google-gemini-cli" "openai-codex-device" "xai" "xai-oauth" \
 	"gitlab-duo" "gitlab-duo-agent" "alibaba-coding-plan" "alibaba-token-plan" "aiand" "abliteration" "zhipu-coding-plan" \
 	"umans" "qwen-portal" "sakana" "minimax-code" "minimax-code-cn" "xiaomi" "xiaomi-token-plan-sgp" \
-	"xiaomi-token-plan-ams" "xiaomi-token-plan-cn" "firepass" "cline-pass" "commandcode" "deepseek" "muse-code" "meta" "moonshot" \
+	"xiaomi-token-plan-ams" "xiaomi-token-plan-cn" "firepass" "cline-pass" "commandcode" "charm-hyper" "deepseek" "muse-code" "meta" "moonshot" \
 	"cerebras" "baseten" "fireworks" "together" "nvidia" "novita" "deepinfra" "huggingface" "perplexity" \
 	"qianfan" "venice" "siliconflow" "siliconflow-cn" "synthetic" "nanogpt" "wafer-serverless" "coreweave" \
 	"vercel-ai-gateway" "cloudflare-ai-gateway" "litellm" "kilo" "zenmux" "opencode-zen" "opencode-go" \

package/src/compat/rules/auth/charm-hyper.kdl

@@ -0,0 +1,19 @@
+// Validation targets `/v1/credits`, not `/v1/models`: the models endpoint is
+// public and answers 200 with no credentials or a bogus key (verified
+// 2026-09-11), so a models-endpoint probe against it could never reject a bad
+// key. `/v1/credits` is an authenticated, non-billable GET that returns the
+// account balance on success and 401 `{"error":"authentication failed"}`
+// otherwise — the same trick cline-pass uses to validate against its own
+// `/users/me`, and cheaper than a chat-completions probe that would bill the
+// key on every login.
+auth "charm-hyper" {
+	name "Charm Hyper"
+	env "CHARM_HYPER_API_KEY" "HYPER_API_KEY"
+	login "api-key" {
+		auth-url "https://hyper.charm.land/"
+		instructions "Create or copy an API key from the Charm Hyper dashboard"
+		prompt "Paste your Charm Hyper API key" placeholder="sk-hyper-..."
+		normalize "strip-bearer"
+		validate "models-endpoint" label="Charm Hyper" url="https://hyper.charm.land/v1/credits"
+	}
+}

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

@@ -169,6 +169,23 @@ class "anthropic" {
 				cache-read 1.0
 				cache-write 12.5
 			}
+			// Fable 5.1 cut cache reads to $0.25/MTok, 0.025x input, a 75% rate
+			// unique to the 5.1 generation; Fable 5 stays at the 0.1x family
+			// default. https://www.anthropic.com/claude/fable
+			//
+			// `cost-patch` is an object axis, so the winning rule supplies the
+			// whole payload rather than merging field-wise over the family
+			// patch above. Restate input/output/cache-write, or a
+			// discovery-sourced 5.1 row (Anthropic's /v1/models omits pricing)
+			// would keep only the cache-read correction.
+			revision "=5.1" {
+				cost-patch {
+					input 10.0
+					output 50.0
+					cache-read 0.25
+					cache-write 12.5
+				}
+			}
 			limits-patch {
 				context-window 1000000
 				max-tokens 128000

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

@@ -73,6 +73,17 @@ class "deepseek" {
 	models token="vision" {
 		strip-image-input #false
 	}
+	// DeepSeek V4.1 Flash is natively multimodal (image-text-to-text), but its
+	// id carries no `vision` token, so the class-wide strip above dropped
+	// attachments the model reads on every deployment without its own carve-out.
+	// The glob also spans `-vision`/`-ocr` ids, which already set the same value
+	// below, so the explicit priority resolves that equal-rank tie instead of
+	// throwing AmbiguousOverlapError. Only the strip is disabled; the declared
+	// modality stays host-derived, so endpoints whose OpenAI-compatible surface
+	// rejects `image_url` keep their text-only contract.
+	models "*v4.1-flash*" priority=10 {
+		strip-image-input #false
+	}
 	on "ollama-cloud" "nvidia" "deepseek" "fireworks" "nanogpt" "opencode-go" "openrouter" {
 		stream-markup-healing-pattern "dsml"
 	}

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

@@ -28,8 +28,10 @@ class "minimax" {
 	}
 	// MiniMax-M3: 512K is the standard pricing-tier boundary leaked into
 	// upstream metadata, not the model ceiling (documented: 1M in / 128K out).
-	on "minimax" "minimax-cn" "minimax-code" "minimax-code-cn" {
-		models "MiniMax-M3" {
+	// Exact selectors are case-sensitive, so both served spellings are listed:
+	// MiniMax hosts publish `MiniMax-M3`, Charm Hyper publishes `minimax-m3`.
+	on "charm-hyper" "minimax" "minimax-cn" "minimax-code" "minimax-code-cn" {
+		models "MiniMax-M3" "minimax-m3" {
 			limits-patch {
 				context-window 1000000
 				max-tokens 128000

package/src/compat/rules/providers/charm-hyper.kdl

@@ -0,0 +1,82 @@
+// Charm Hyper gateway deployment contract.
+//
+// Capability data is NOT restated here: `GET /v1/models` publishes each row's
+// context window, output cap, vision flag, effort vocabulary and per-million
+// tariff, and `charmHyperModelManagerOptions` maps all of it verbatim —
+// including `reasoningDisableMode`, which is derived per model from whether
+// the advertised vocabulary contains `none`. Only facts the gateway does not
+// publish, or publishes wrongly, belong in this file.
+provider "charm-hyper" {
+	// OpenAI-dialect gateway in front of open-weight and third-party models.
+	// `developer` and `store` are accepted (HTTP 200) but the gateway accepts
+	// every unknown field it is handed, so acceptance is not evidence the
+	// semantics are honored; both stay off and system prompts ride the
+	// `system` role, matching every sibling gateway. The wire takes
+	// `max_tokens` (verified 2026-09-11).
+	supports-developer-role #false
+	supports-store #false
+	max-tokens-field "max_tokens"
+	thinking-mode "effort"
+	// Reasoning arrives as `reasoning_content` on both the streamed delta and
+	// the final message (verified 2026-09-11).
+	reasoning-content-field "reasoning_content"
+	// ---------------------------------------------------------------------------
+	// Blockless rows are left non-reasoning on purpose.
+	//
+	// Four ids serve a thinking model while publishing no `reasoning` block:
+	// glm-5, glm-5.1, kimi-k2-thinking and minimax-m2.7 each returned
+	// populated `reasoning_content` and non-zero `reasoning_tokens` on a
+	// multi-step prompt (2026-09-11). The other blockless rows
+	// (gemma-4-26b-a4b-it, kimi-k2.5, kimi-k2.7-code, llama-3.3-70b-instruct,
+	// llama-4-maverick-17b-128e-instruct-fp8,
+	// qwen3-coder-480b-a35b-instruct-int4-mixed-ar, qwen3-next-80b-a3b-instruct)
+	// returned none.
+	//
+	// No rule restores them, because none can be written honestly. Those four
+	// expose no effort dial, and `ThinkingConfig.efforts` is documented as
+	// never empty — a reasoning model without a controllable surface carries
+	// `thinking: undefined`, which no KDL axis can express (the only upgrade
+	// path is an exact `thinking-efforts` ladder). Synthesizing a ladder would
+	// add silent no-op rungs: the gateway answers 200 to efforts it never
+	// advertised and simply ignores them, and a repeat measurement of effort
+	// sensitivity on these ids came back non-monotonic, i.e. sampling noise
+	// rather than a hidden dial.
+	//
+	// The cost is narrow and was checked: streamed `reasoning_content` is
+	// parsed and displayed with no `model.reasoning` gate, the effort
+	// parameter is moot without a dial, and the developer-role and
+	// zai-replay paths are already disabled for this provider.
+	// ---------------------------------------------------------------------------
+	// Residue: misreported output caps. The gateway derives
+	// `max_output_tokens` as a flat fraction of the context window rather
+	// than reporting a real ceiling — 10% on 9 rows, the dominant
+	// convention — and the values are advisory, not enforced: glm-5.1
+	// publishes 3276 yet produced 14066 tokens, and minimax-m2.7 publishes
+	// 6553 yet produced 8418, both stopping on `finish_reason: "stop"`
+	// rather than `length`. An over-stated cap is never rejected either
+	// (`max_tokens: 200000` returns 200), so the risk is one-sided:
+	// over-stating costs only the context share 10% already reserves
+	// everywhere else, while under-stating truncates real output.
+	//
+	// Both corrections take a same-window peer's published value rather than
+	// inventing a ceiling, and both clear the measured floor:
+	//   glm-5.1      (202750) is 1.6%; sibling glm-5 (202752) publishes 20275.
+	//   minimax-m2.7 (262100) is 2.5%, the lone outlier in its window class;
+	//                kimi-k2.5 (262144), kimi-k2.6 (262000) and
+	//                qwen3-next-80b (262144) all publish 26214.
+	//
+	// deepseek-v4.1-flash is the only other 2.5% row and is deliberately left
+	// alone: its 1M window has no peer group that agrees on a fraction, and a
+	// long generation stopped naturally at 8504, well inside the published
+	// 26214. Measured 2026-09-11.
+	models "glm-5.1" {
+		limits-patch {
+			max-tokens 20275
+		}
+	}
+	models "minimax-m2.7" {
+		limits-patch {
+			max-tokens 26214
+		}
+	}
+}

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

@@ -82,11 +82,22 @@ provider "deepseek" {
 		}
 	}
 	// residue: taxonomy ranks and exact globs do not isolate these models.
-	models "deepseek-v4-flash" "deepseek-v4.1-flash-expires-on-0910" {
+	// The bare `deepseek-flash` alias serves DeepSeek-V4.1-Flash, so it shares
+	// the V4.1 wire contract: max_tokens + reasoning_content, mandatory replay
+	// (synthetic placeholders rejected upstream), no tool choice, and the
+	// low/high/max effort ladder the bare alias cannot inherit by family.
+	models "deepseek-flash" "deepseek-v4-flash" "deepseek-v4.1-flash-expires-on-0910" {
+		// Stale source capability data: discovery reports the bare alias as
+		// non-reasoning, so opt into the cascade upgrade for the exact ladder.
+		thinking-upgrade-neutral #true
 		max-tokens-field "max_tokens"
 		reasoning-content-field "reasoning_content"
 		requires-assistant-content-for-tool-calls #true
 		requires-reasoning-content-for-tool-calls #true
+		allows-synthetic-reasoning-content-for-tool-calls #false
 		supports-tool-choice #false
+		// The bare alias carries no taxonomy family, so it cannot inherit the
+		// flash ladder; the exact rule owns it instead.
+		thinking-efforts "low" "high" "max"
 	}
 }

package/src/compat/rules/providers/opencode-go.kdl

@@ -45,6 +45,18 @@ provider "opencode-go" {
 	class "kimi" {
 		thinking-mode "effort"
 	}
+	// The Go gateway proxies Muse Spark's Responses lane to Meta but cannot
+	// round-trip encrypted reasoning: the upstream issues `encrypted_content`
+	// bound to the gateway's own caller, so replaying it on a later step 400s
+	// with "reasoning `encrypted_content` was not issued to this caller"
+	// (#11928). Stop requesting it and drop native reasoning items from replay
+	// so multi-step tool-call turns stop failing.
+	class "meta" {
+		family "muse-spark" {
+			include-encrypted-reasoning #false
+			filter-reasoning-history #true
+		}
+	}
 	class "mimo" {
 		family "v2" {
 			thinking-mode "effort"

package/src/compat/rules/providers/opencode-zen.kdl

@@ -26,6 +26,18 @@ provider "opencode-zen" {
 	class "kimi" {
 		thinking-mode "effort"
 	}
+	// The Zen gateway proxies Muse Spark's Responses lane to Meta but cannot
+	// round-trip encrypted reasoning: the upstream issues `encrypted_content`
+	// bound to the gateway's own caller, so replaying it on a later step 400s
+	// with "reasoning `encrypted_content` was not issued to this caller"
+	// (#11928). Stop requesting it and drop native reasoning items from replay
+	// so multi-step tool-call turns stop failing.
+	class "meta" {
+		family "muse-spark" {
+			include-encrypted-reasoning #false
+			filter-reasoning-history #true
+		}
+	}
 	class "mimo" {
 		family "v2" {
 			thinking-mode "effort"

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

@@ -1,10 +1,6 @@
 // Provider-wire compat for "openrouter"; regenerated from the frozen census using deployment contract selectors.
 
 provider "openrouter" {
-	// V4.1 Flash accepts images despite its ID lacking a vision suffix.
-	models "deepseek/deepseek-v4.1-flash" {
-		strip-image-input #false
-	}
 	// Replaces the OpenRouter provider wire-model-id dispatch branch.
 	wire-model-id-mode "openrouter"
 	// Replaces the strict-mode provider whitelist entry.