@oh-my-pi/pi-catalog 18.8.8 → 18.9.0

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

@@ -111,6 +111,12 @@ provider "factory-droid" {
 			cost input=0 output=0 cache-read=0 cache-write=0
 			limits context=180000 max-tokens=32000
 		}
+		model "claude-haiku-5-5" name="Haiku 5.5" {
+			reasoning #true
+			input "text" "image"
+			cost input=0 output=0 cache-read=0 cache-write=0
+			limits context=872000 max-tokens=128000
+		}
 		model "claude-haiku-4-5-20251001" name="Haiku 4.5" {
 			reasoning #true
 			input "text" "image"
@@ -273,6 +279,12 @@ provider "factory-droid" {
 			cost input=0 output=0 cache-read=0 cache-write=0
 			limits context=1007232 max-tokens=32768
 		}
+		model "mistral-large-4" name="Mistral Large 4" {
+			reasoning #true
+			input "text" "image"
+			cost input=0 output=0 cache-read=0 cache-write=0
+			limits context=524288 max-tokens=64000
+		}
 		model "mistral-medium-3.5" name="Mistral Medium 3.5" {
 			reasoning #true
 			input "text" "image"
@@ -400,7 +412,7 @@ provider "factory-droid" {
 	models "claude-fable-5.1" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic"
-		region-upstreams-eu
+		region-upstreams-eu "vertex_anthropic"
 		credit-rates { input 4; output 5; cache-read 0.025 }
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
 		thinking-requires-effort #false
@@ -410,7 +422,7 @@ provider "factory-droid" {
 	models "claude-fable-5" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic"
-		region-upstreams-eu
+		region-upstreams-eu "vertex_anthropic"
 		credit-rates { input 4; output 5 }
 		list-price-from "anthropic"
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
@@ -439,7 +451,7 @@ provider "factory-droid" {
 	models "claude-opus-5" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic" "azure_anthropic" "snowflake"
-		region-upstreams-eu "bedrock_anthropic"
+		region-upstreams-eu "vertex_anthropic" "bedrock_anthropic"
 		credit-rates { input 2 }
 		list-price-from "anthropic"
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
@@ -458,7 +470,7 @@ provider "factory-droid" {
 	models "claude-opus-4-8" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic" "azure_anthropic" "snowflake"
-		region-upstreams-eu "bedrock_anthropic"
+		region-upstreams-eu "vertex_anthropic" "bedrock_anthropic"
 		credit-rates { input 2 }
 		list-price-from "anthropic"
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
@@ -477,7 +489,7 @@ provider "factory-droid" {
 	models "claude-opus-4-7" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic"
-		region-upstreams-eu "bedrock_anthropic"
+		region-upstreams-eu "vertex_anthropic" "bedrock_anthropic"
 		credit-rates { input 2 }
 		list-price-from "anthropic"
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
@@ -506,9 +518,10 @@ provider "factory-droid" {
 	models "claude-sonnet-5-5" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic" "azure_anthropic"
-		credit-rates { input 0.8; output 5 }
+		credit-rates { input 0.8; output 5; cache-read 0.05 }
+		list-price-from "anthropic"
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
-		thinking-requires-effort #true
+		thinking-requires-effort #false
 		thinking-default-level "high"
 		entitlement { feature-flag "claude_sonnet_5_5" }
 	}
@@ -540,6 +553,16 @@ provider "factory-droid" {
 		thinking-requires-effort #false
 		default-reasoning-off #true
 	}
+	models "claude-haiku-5-5" {
+		routing-family "anthropic"
+		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic" "azure_anthropic"
+		credit-rates { input 0.04; output 5 }
+		list-price-from "anthropic"
+		thinking-efforts "low" "medium" "high" "xhigh" "max"
+		thinking-requires-effort #false
+		thinking-default-level "medium"
+		entitlement { feature-flag "claude_haiku_5_5" }
+	}
 	models "claude-haiku-4-5-20251001" {
 		routing-family "anthropic"
 		upstream-rotation "anthropic" "vertex_anthropic" "bedrock_anthropic"
@@ -562,7 +585,7 @@ provider "factory-droid" {
 	}
 	models "gpt-6.1-sol" {
 		routing-family "openai"
-		upstream-rotation "openai"
+		upstream-rotation "openai" "bedrock_openai" "databricks"
 		region-upstreams-eu "openai"
 		credit-rates { input 0.8; output 5; cache-read 0.05 }
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
@@ -589,7 +612,6 @@ provider "factory-droid" {
 		thinking-efforts "low" "medium" "high" "xhigh" "max"
 		thinking-requires-effort #false
 		thinking-default-level "medium"
-		entitlement { feature-flag "gpt_6_luna" }
 	}
 	models "gpt-5.6-sol" {
 		routing-family "openai"
@@ -749,6 +771,7 @@ provider "factory-droid" {
 	models "gemini-3.8-flash" {
 		routing-family "google"
 		upstream-rotation "google"
+		region-upstreams-eu "google"
 		credit-rates { input 0.6; output 5 }
 		list-price-from "google"
 		thinking-efforts "low" "medium" "high"
@@ -758,6 +781,7 @@ provider "factory-droid" {
 	models "gemini-3.7-flash" {
 		routing-family "google"
 		upstream-rotation "google"
+		region-upstreams-eu "google"
 		credit-rates { input 0.6; output 5 }
 		list-price-from "google"
 		thinking-efforts "low" "medium" "high"
@@ -767,6 +791,7 @@ provider "factory-droid" {
 	models "gemini-3.6-flash" {
 		routing-family "google"
 		upstream-rotation "google"
+		region-upstreams-eu "google"
 		credit-rates { input 0.6; output 5 }
 		list-price-from "google"
 		thinking-efforts "low" "medium" "high"
@@ -776,6 +801,7 @@ provider "factory-droid" {
 	models "gemini-3.5-flash" {
 		routing-family "google"
 		upstream-rotation "google"
+		region-upstreams-eu "google"
 		credit-rates { input 0.6 }
 		list-price-from "google"
 		thinking-efforts "minimal" "low" "medium" "high"
@@ -801,6 +827,19 @@ provider "factory-droid" {
 		thinking-requires-effort #false
 		thinking-default-level "high"
 	}
+	// Factory bills the Standard list-price ceiling, above the bundled Mistral
+	// row, so no list price is borrowed.
+	models "mistral-large-4" {
+		routing-family "factory"
+		upstream-rotation "mistral"
+		region-upstreams-us
+		region-upstreams-eu "mistral"
+		credit-rates { input 0.544; output 3.08; cache-read 0.103 }
+		thinking-efforts "high"
+		thinking-requires-effort #false
+		thinking-default-level "high"
+		entitlement { feature-flag "mistral_large_4" }
+	}
 	models "mistral-medium-3.5" {
 		routing-family "factory"
 		upstream-rotation "mistral"
@@ -889,7 +928,6 @@ provider "factory-droid" {
 		thinking-efforts "low" "high" "max"
 		thinking-requires-effort #false
 		thinking-default-level "high"
-		entitlement { feature-flag "deepseek_v4_1_flash" }
 	}
 	models "deepseek-v4-flash-0731" {
 		routing-family "factory"
@@ -999,7 +1037,8 @@ provider "factory-droid" {
 		// Native declares every Responses tool `strict: false`; the field must reach the wire.
 		supports-strict-mode #true
 	}
-	models "glm-*" "kimi-*" "deepseek-*" "nemotron-*" "inkling" "mistral-medium-3.5" "qwen3.8-max" "minimax-m3" {
+	models "glm-*" "kimi-*" "deepseek-*" "nemotron-*" "inkling" "mistral-medium-3.5" "mistral-large-4" "qwen3.8-max" \
+		"minimax-m3" {
 		clamp-output-to-model-max #true
 		completions-reasoning-mode "none"
 		completions-reasoning-history "omit"
@@ -1039,7 +1078,7 @@ provider "factory-droid" {
 	}
 	// The replay codec follows the selected route, including EU GLM routes.
 	on-upstream "mistral" {
-		models "glm-*" "mistral-medium-3.5" {
+		models "glm-*" "mistral-medium-3.5" "mistral-large-4" {
 			completions-reasoning-mode "effort"
 			completions-reasoning-history "omit"
 			mistral-reasoning-content-parts #true
@@ -1058,6 +1097,7 @@ provider "factory-droid" {
 	// Messages contract on every Factory Anthropic route: non-thinking-led
 	// history drops thinking, disabled thinking omits the field, and the effort
 	// beta rides only where the route requires it alongside output effort.
+	// Where disabled thinking is sent, replayed history drops its thinking.
 	on-api "anthropic-messages" {
 		strip-thinking-history #true
 		disabled-thinking "omit"
@@ -1073,6 +1113,12 @@ provider "factory-droid" {
 			disabled-thinking "disabled"
 		}
 	}
+	// Adaptive models the registry marks `supportsDisabledThinking` send the
+	// explicit disabled form at Off on every upstream.
+	models "claude-sonnet-4-6" "claude-sonnet-5" "claude-opus-4-6" "claude-opus-4-7" "claude-opus-4-8" \
+		"claude-opus-4-8-fast" "claude-opus-5" "claude-opus-5-fast" "claude-haiku-5-5" {
+		disabled-thinking "disabled"
+	}
 	models "claude-*" "atlas-07-21" "aster-07-15" {
 		anthropic-thinking "adaptive-summarized"
 	}
@@ -1091,9 +1137,15 @@ provider "factory-droid" {
 		anthropic-thinking "budget-effort"
 	}
 	// Always-thinking models keep adaptive thinking on disabled turns, on every upstream.
-	models "claude-fable-5.1" "claude-opus-5-5" "claude-opus-5-5-fast" "claude-sonnet-5-5" {
+	models "claude-fable-5.1" "claude-opus-5-5" "claude-opus-5-5-fast" {
 		disabled-thinking "adaptive"
 	}
+	// Sonnet 5.5 rejects `disabled`: Off is `between_tools`, pinned to high
+	// effort because `between_tools` rejects xhigh and max.
+	models "claude-sonnet-5-5" {
+		disabled-thinking "between-tools"
+		between-tools-effort "high"
+	}
 	// Fast tiers are distinct SKUs served with `speed: "fast"`.
 	models "claude-opus-5-5-fast" "claude-opus-5-fast" "claude-opus-4-8-fast" {
 		fast-mode #true
@@ -1113,6 +1165,13 @@ provider "factory-droid" {
 	on-upstream "bedrock_anthropic" {
 		anthropic-tool-streaming-beta #true
 	}
+	// Azure Foundry enforces the preserved-thinking prefix check on Factory's
+	// account, so prefix-bound models (`thinking-prefix-binding`) ask it to drop
+	// stale signed blocks instead of failing. The other Anthropic upstreams
+	// predate default enforcement; the field would opt them into it.
+	on-upstream "azure_anthropic" {
+		supports-thinking-binding-controls #true
+	}
 	// xAI and Google sessions lock their upstream.
 	on-upstream "xai" "google" {
 		routing-session-lock #true
@@ -1132,6 +1191,11 @@ provider "factory-droid" {
 			reasoning-disable-mode "none-effort"
 		}
 	}
+	// GPT models whose ladder has an Off/None rung send the literal `none`
+	// tier on every upstream.
+	models "gpt-5.2" "gpt-5.6-sol" "gpt-5.6-sol-fast" "gpt-5.6-terra" "gpt-5.6-luna" "gpt-6-sol" "gpt-6-luna" {
+		reasoning-disable-mode "none-effort"
+	}
 	models "gpt-*" {
 		omit-max-output-tokens #true
 		responses-verbosity "low"

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

@@ -1,7 +1,10 @@
-// Catalog entry for "litellm"; no deployment-contract compat rules are needed.
+// Provider-wire compat for "litellm".
 
 provider "litellm" {
 	default-model "claude-opus-5-5"
 	env "LITELLM_API_KEY"
 	discovery label="LiteLLM" allow-unauthenticated=#true
+
+	// The proxy groups request logs into sessions by this header.
+	session-header "x-litellm-session-id"
 }

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

@@ -24,7 +24,11 @@ provider "mistral" {
 		}
 	}
 
-	thinking-mode "effort"
+	// Mistral's chat-completions endpoint derives cache affinity from the
+	// top-level `prompt_cache_key` (https://docs.mistral.ai/studio/conversations/advanced/prompt-caching);
+	// without it, requests only hit replica-local implicit cache. Same shape as
+	// the kimi-code rule (#6049).
+	supports-prompt-cache-key #true
 	class "unknown" {
 		thinking-efforts "minimal" "low" "medium" "high" "xhigh"
 	}

package/src/compat/rules/providers/openai-codex.kdl

@@ -77,16 +77,6 @@ provider "openai-codex" {
 			cache-read 1.0
 			cache-write 0
 		}
-		service-tier-cost {
-			flex 0.5
-			priority 2.5
-			// Ultrafast is Astra-only. This table uses OpenAI's included-usage
-			// multipliers for Codex: Ultrafast 8x and Fast 2.5x Standard (purchased
-			// credits / Enterprise pay-as-you-go bill 6x and 2x instead). See
-			// https://learn.chatgpt.com/docs/agent-configuration/speed and
-			// https://learn.chatgpt.com/docs/pricing.
-			ultrafast 8
-		}
 		limits-patch {
 			context-window 272000
 		}
@@ -117,6 +107,18 @@ provider "openai-codex" {
 		}
 		max-context-window 922000
 	}
+	// Codex Ultrafast ships on GPT-6 Astra and GPT-6.1 Sol. This table uses
+	// OpenAI's included-usage multipliers for Codex: Ultrafast 8x and Fast 2.5x
+	// Standard (purchased credits / Enterprise pay-as-you-go bill 6x and 2x
+	// instead). See https://learn.chatgpt.com/docs/agent-configuration/speed
+	// and https://learn.chatgpt.com/docs/pricing.
+	models "gpt-6-astra" "gpt-6-astra-wm" "gpt-6.1-sol" "gpt-6.1-sol-wm" {
+		service-tier-cost {
+			flex 0.5
+			priority 2.5
+			ultrafast 8
+		}
+	}
 	models "gpt-6-luna" "gpt-6-luna-wm" {
 		cost-patch {
 			input 0.1

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

@@ -176,8 +176,10 @@ provider "openai" {
 			cache-read 2.0
 			cache-write 25.0
 		}
-		// Ultrafast is Astra-only: $60/$6/$75/$300 per 1M short-context
-		// (6x the standard card), and the same 6x holds on the long-context tier.
+	}
+	// API Ultrafast is 6x the standard card on both context tiers: GPT-6 Astra
+	// $60/$6/$75/$300 and GPT-6.1 Sol $12/$0.60/$15/$60 per 1M short-context.
+	models "gpt-6-astra*" "gpt-6.1-sol*" {
 		service-tier-cost {
 			ultrafast 6
 		}

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

@@ -116,4 +116,24 @@ provider "opencode-zen" {
 		thinking-efforts "minimal" "low" "medium" "high"
 		thinking-mode "effort"
 	}
+	// Zen serves Jev over System One, not chat completions.
+	seed api="typesafe" bundle="always" base-url="https://opencode.ai/zen" {
+		model "jev-1.13" name="Jev 1.13" {
+			reasoning #false
+			input "text"
+			cost input=0.042 output=0 cache-read=0 cache-write=0
+			limits
+			supports-tools #false
+		}
+		model "jev-1.13-free" name="Jev 1.13 Free" {
+			reasoning #false
+			input "text"
+			cost input=0 output=0 cache-read=0 cache-write=0
+			limits
+			supports-tools #false
+		}
+	}
+	models "jev-1.13" "jev-1.13-free" {
+		kind "judge"
+	}
 }

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

@@ -159,6 +159,12 @@ provider "openrouter" {
 		family "o-series" {
 			thinking-efforts "minimal" "low" "medium" "high"
 		}
+		// OpenRouter's Responses API does not document `configuration_update`
+		// input items (and serves `-pro` rows that would need pro mode, where the
+		// item is unsupported): GPT-6 effort changes stay on the request level.
+		revision ">=6 <7" {
+			supports-configuration-update #false
+		}
 	}
 	class "stepfun" {
 		family "step" {

package/src/compat/rules/providers/zhipu-coding-plan.kdl

@@ -5,6 +5,16 @@ provider "zhipu-coding-plan" {
 	env "ZHIPU_API_KEY"
 	dynamic-models-authoritative #true
 	discovery label="Zhipu Coding Plan"
+	// The coding-plan upstream snapshot omits FlashX although the served SKU
+	// has the same limits and mandatory-thinking ladder as GLM-5.3-Flash.
+	seed api="openai-completions" base-url="https://open.bigmodel.cn/api/coding/paas/v4" bundle="always" {
+		model "glm-5.3-flashx" name="GLM-5.3-FlashX" {
+			reasoning #true
+			input "text" "image"
+			cost input=0 output=0 cache-read=0 cache-write=0
+			limits context=1000000 max-tokens=131072
+		}
+	}
 
 	// Replaces the Zhipu provider thinking-format branch.
 	thinking-format "zai"

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

@@ -92,13 +92,13 @@ behavior {
 	// Factory Droid bills each model from one subscription pool: Droid Core
 	// (flat rate) or Standard Credits. Unknown ids have no inferred entitlement.
 	quota-tiers provider="factory-droid" {
-		tier "core" "inkling" "mistral-medium-3.5" "glm-5.3-flash" "glm-5.3" "glm-5.2" "glm-5.2-fast" \
+		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"
 		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" \
-			"claude-sonnet-4-6" "claude-sonnet-4-5-20250929" "claude-haiku-4-5-20251001" "gpt-6-astra" \
+			"claude-sonnet-4-6" "claude-sonnet-4-5-20250929" "claude-haiku-5-5" "claude-haiku-4-5-20251001" "gpt-6-astra" \
 			"gpt-6.1-sol" "gpt-6-sol" "gpt-6-luna" "gpt-5.6-sol" "gpt-5.6-sol-fast" "gpt-5.6-terra" \
 			"gpt-5.6-luna" "gpt-5.5" "gpt-5.5-fast" "gpt-5.5-pro" "gpt-5.4" "gpt-5.4-fast" "gpt-5.4-mini" \
 			"gpt-5.4-mini-fast" "gpt-5.3-codex" "gpt-5.3-codex-fast" "gpt-5.2" "garnet-07-15" \
@@ -279,7 +279,8 @@ behavior {
 			exact="claude-opus-4-8" exact="claude-opus-4-8-fast" exact="claude-opus-4-7" \
 			exact="claude-opus-4-6" exact="claude-opus-4-5-20251101" exact="claude-sonnet-5-5" \
 			exact="claude-sonnet-5" exact="claude-sonnet-4-6" exact="claude-sonnet-4-5-20250929" \
-			exact="claude-haiku-4-5-20251001" exact="atlas-07-21" exact="aster-07-15" exact="minimax-m2.7"
+			exact="claude-haiku-5-5" exact="claude-haiku-4-5-20251001" exact="atlas-07-21" exact="aster-07-15" \
+			exact="minimax-m2.7"
 		route "openai-responses" exact="gpt-6-astra" exact="gpt-6.1-sol" exact="gpt-6-sol" exact="gpt-6-luna" \
 			exact="gpt-5.6-sol" exact="gpt-5.6-sol-fast" exact="gpt-5.6-terra" exact="gpt-5.6-luna" \
 			exact="gpt-5.5" exact="gpt-5.5-fast" exact="gpt-5.5-pro" exact="gpt-5.4" exact="gpt-5.4-fast" \
@@ -288,7 +289,7 @@ behavior {
 		route "google-generate" exact="garnet-07-15" exact="gemini-3.1-pro-preview" exact="gemini-3.8-flash" \
 			exact="gemini-3.7-flash" exact="gemini-3.6-flash" exact="gemini-3.5-flash" \
 			exact="gemini-3-flash-preview"
-		route "openai-completions" exact="inkling" exact="mistral-medium-3.5" exact="glm-5.3-flash" \
+		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"