@oh-my-pi/pi-catalog 18.1.17 → 18.1.18

package/CHANGELOG.md

@@ -2,10 +2,23 @@
 
 ## [Unreleased]
 
+## [18.1.18] - 2026-09-11
+
+### Added
+
+- `supports-server-compaction` compat axis (`compat.supportsServerCompaction`): whether a model line accepts Anthropic server-side compaction (`compact-2026-01-12`). Class rules enable it for Opus 4.6+, Sonnet 4.6+, and Fable/Mythos 5 on every Anthropic-messages host; the default is `false`.
+
+### Fixed
+
+- 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)).
+
 ## [18.1.17] - 2026-09-10
 
 ### Added
 
+- Added DeepSeek V4.1 Flash on OpenRouter with image input and low/high/max reasoning levels ([#11592](https://github.com/can1357/oh-my-pi/pull/11592) by [@mazzanfar](https://github.com/mazzanfar)).
 - Added DeepSeek cost estimates that follow published peak/off-peak rates.
 - Added dated, announced price changes to the catalog, so rates switch on their effective date (e.g. DeepSeek Pro moving to Flash rates).
 - Added Command Code as a built-in provider with API-key login, live model discovery, per-model pricing, native OpenAI/Anthropic-compatible routing, cache-aware token usage, and TTFT metrics ([#11391](https://github.com/can1357/oh-my-pi/pull/11391) by [@CherkaSSH](https://github.com/CherkaSSH)).

package/dist/types/types.d.ts

@@ -409,6 +409,20 @@ export interface OpenAICompat {
 export interface AnthropicCompat {
     /** Whether thinking requests may include `context_management` and its beta header. Default: true. */
     supportsContextManagement?: boolean;
+    /**
+     * Whether the model lineage supports Anthropic server-side compaction
+     * (`compact-2026-01-12`: the `compact_20260112` edit and replayed
+     * `compaction` blocks). Rule-owned per model line; the beta covers the
+     * adaptive-thinking generation onward and rejects older lines. Default: false.
+     */
+    supportsServerCompaction?: boolean;
+    /**
+     * Whether the model is served by the first-party Anthropic provider (its
+     * default route is the official API). Rule-owned on the provider; the
+     * compaction transport pairs it with a per-request effective-URL check
+     * because reroutes leave it stale-true. Default: false.
+     */
+    firstPartyProvider?: boolean;
     /**
      * Whether requests may carry `output_config.effort` (and its effort beta
      * header). Vertex AI rejects the field/header. Default: true.
@@ -756,6 +770,13 @@ export type ResolvedAnthropicCompat = Required<Omit<AnthropicCompat, "streamIdle
      * env headers, and cache-TTL shaping without per-request URL parsing.
      */
     officialEndpoint: boolean;
+    /**
+     * The model is served by the first-party Anthropic provider. Unlike
+     * `officialEndpoint` (URL-derived per deployment), this is provider
+     * identity for compaction gating, always paired with a per-request
+     * effective-URL check.
+     */
+    firstPartyProvider: boolean;
 };
 /**
  * Compatibility settings for the devin-agent (Codeium Cascade) API. Cascade
@@ -910,6 +931,8 @@ export interface Model<TApi extends Api = Api> {
     requiresGlyphTokenization?: boolean;
     /** Whether this model requires Cursor's tool-schema combiner projection. */
     requiresCursorToolSchemaProjection?: boolean;
+    /** Whether this model requires tool-result images hoisted into sibling user content blocks. */
+    requiresToolResultImageHoisting?: boolean;
     /**
      * Model id to send on the wire when it differs from `id`. Used by catalog
      * variants that present one upstream model under several local entries —
@@ -1083,7 +1106,7 @@ export interface Model<TApi extends Api = Api> {
  * vocabulary of `buildModel`. Identical to `Model` except `compat` carries the
  * sparse override shape and nothing is resolved yet.
  */
-export interface ModelSpec<TApi extends Api = Api> extends Omit<Model<TApi>, "compat" | "identity" | "compatConfig" | "requiresGlyphTokenization" | "requiresCursorToolSchemaProjection" | "supportsComputerUseConfig"> {
+export interface ModelSpec<TApi extends Api = Api> extends Omit<Model<TApi>, "compat" | "identity" | "compatConfig" | "requiresGlyphTokenization" | "requiresCursorToolSchemaProjection" | "requiresToolResultImageHoisting" | "supportsComputerUseConfig"> {
     /** Sparse compatibility overrides; resolved into `Model.compat` by `buildModel`. */
     compat?: CompatConfigOf<TApi>;
 }

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.1.17",
+	"version": "18.1.18",
 	"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": "18.1.17",
-		"@oh-my-pi/pi-utils": "18.1.17"
+		"@oh-my-pi/omptype": "18.1.18",
+		"@oh-my-pi/pi-utils": "18.1.18"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.1.17",
+		"@oh-my-pi/pi-ai": "18.1.18",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/build.ts

@@ -34,8 +34,8 @@ function isInputModalities(value: unknown): value is ("text" | "image")[] {
  * corrections (`cost-patch`, `limits-patch`, `long-context-cost`,
  * `context-window-floor`) overwrite upstream values; selection metadata
  * (`priority`, `apply-patch-tool-type`, `service-tier-cost`,
- * `requires-cursor-tool-schema-projection`) is rule-owned;
- * `context-promotion-target` fills only when the spec left it unset.
+ * `requires-cursor-tool-schema-projection`, `requires-tool-result-image-hoisting`)
+ * is rule-owned; `context-promotion-target` fills only when the spec left it unset.
  */
 function applyCatalogAssignments<TApi extends Api>(model: Model<TApi>, catalog: Record<string, unknown>): void {
 	const serviceTierCost = objectPayload(catalog.serviceTierCost);
@@ -63,6 +63,12 @@ function applyCatalogAssignments<TApi extends Api>(model: Model<TApi>, catalog:
 	} else {
 		delete model.requiresCursorToolSchemaProjection;
 	}
+	const requiresToolResultImageHoisting = catalog.requiresToolResultImageHoisting;
+	if (requiresToolResultImageHoisting === true) {
+		model.requiresToolResultImageHoisting = true;
+	} else {
+		delete model.requiresToolResultImageHoisting;
+	}
 	const contextPromotionTarget = catalog.contextPromotionTarget;
 	if (typeof contextPromotionTarget === "string" && model.contextPromotionTarget === undefined) {
 		model.contextPromotionTarget = contextPromotionTarget;

package/src/compat/axes.ts

@@ -192,6 +192,7 @@ export const AXES: Readonly<Record<string, AxisDef>> = {
 	"disable-adaptive-thinking": wire("disableAdaptiveThinking", ["anthropic"]),
 	"disable-strict-tools": wire("disableStrictTools", ["anthropic"]),
 	"escape-builtin-tool-names": wire("escapeBuiltinToolNames", ["anthropic"]),
+	"first-party-provider": wire("firstPartyProvider", ["anthropic"]),
 	"inject-claude-code-instruction": wire("injectClaudeCodeInstruction", ["anthropic"]),
 	"official-endpoint": wire("officialEndpoint", ["anthropic", "openai-responses"]),
 	"replay-unsigned-thinking": wire("replayUnsignedThinking", ["anthropic"]),
@@ -205,6 +206,7 @@ export const AXES: Readonly<Record<string, AxisDef>> = {
 	"supports-mid-conversation-system": wire("supportsMidConversationSystem", ["anthropic"]),
 	"supports-mid-conversation-tool-changes": wire("supportsMidConversationToolChanges", ["anthropic"]),
 	"supports-per-message-effort": wire("supportsPerMessageEffort", ["anthropic"]),
+	"supports-server-compaction": wire("supportsServerCompaction", ["anthropic"]),
 	"supports-thinking-binding-controls": wire("supportsThinkingBindingControls", ["anthropic"]),
 	"supports-turn-scoped-system": wire("supportsTurnScopedSystem", ["anthropic"]),
 
@@ -285,6 +287,11 @@ export const AXES: Readonly<Record<string, AxisDef>> = {
 		set: "catalog",
 		shape: "scalar",
 	},
+	"requires-tool-result-image-hoisting": {
+		key: "requiresToolResultImageHoisting",
+		set: "catalog",
+		shape: "scalar",
+	},
 	priority: { key: "priority", set: "catalog", shape: "scalar" },
 	"service-tier-cost": { key: "serviceTierCost", set: "catalog", shape: "object" },
 	"time-based-cost": { key: "timeBased", set: "catalog", shape: "object" },

package/src/compat/resolve.ts

@@ -843,6 +843,8 @@ function resolveAnthropicPolicy(
 		officialEndpoint: official,
 		signingEndpoint,
 		supportsContextManagement: true,
+		supportsServerCompaction: false,
+		firstPartyProvider: false,
 		supportsOutputEffort: true,
 		disableStrictTools: isAzure,
 		disableAdaptiveThinking: false,

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

@@ -7,6 +7,30 @@ class "anthropic" {
 			thinking-prefix-binding #true
 		}
 	}
+	// Server-side compaction (compact-2026-01-12) is a lineage capability: the
+	// documented model list is the adaptive-thinking generation onward (Opus
+	// 4.6+, Sonnet 4.6+, Fable/Mythos 5+). Whether a host delivers the beta is
+	// decided at request time from the effective endpoint, not here.
+	family "fable" {
+		revision ">=5" {
+			supports-server-compaction #true
+		}
+	}
+	family "mythos" {
+		revision ">=5" {
+			supports-server-compaction #true
+		}
+	}
+	family "opus" {
+		revision ">=4.6" {
+			supports-server-compaction #true
+		}
+	}
+	family "sonnet" {
+		revision ">=4.6" {
+			supports-server-compaction #true
+		}
+	}
 	on "anthropic" "cloudflare-ai-gateway" "google-vertex" {
 		family "fable" {
 			revision ">=5" {

package/src/compat/rules/providers/amazon-bedrock.kdl

@@ -15,6 +15,8 @@ provider "amazon-bedrock" {
 	}
 	class "openai" {
 		thinking-mode "effort"
+		// OpenAI models reject Converse image blocks nested inside toolResult content.
+		requires-tool-result-image-hoisting #true
 	}
 	class "deepseek" {
 		requires-reasoning-content-for-all-assistant-turns #true
@@ -28,6 +30,11 @@ provider "amazon-bedrock" {
 	class "unknown" {
 		thinking-efforts "minimal" "low" "medium" "high"
 		thinking-mode "budget"
+		// Opaque application-inference-profile ARNs carry no vendor signal, so an
+		// OpenAI-backed profile classifies here and would otherwise nest images in
+		// toolResult content and hit the same rejection. Hoisting is accepted by
+		// every Converse model measured, so it is the safe default for unknowns.
+		requires-tool-result-image-hoisting #true
 	}
 	// residue: taxonomy ranks and exact globs do not isolate these models.
 	models "moonshot.kimi-k2-thinking" {

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

@@ -1,6 +1,14 @@
 // Provider-wire compat for "anthropic"; regenerated from the frozen census using deployment contract selectors.
 
 provider "anthropic" {
+	// First-party provider: models served through this provider reach the
+	// official API unless rerouted. A dedicated axis rather than
+	// `official-endpoint`, which stays URL-derived: a custom `baseUrl` on this
+	// provider must keep `officialEndpoint: false` (SDK `X-Api-Key`
+	// suppression and friends read that flag). The compaction transport
+	// combines this fact with a per-request effective-URL check, so it never
+	// authorizes a block on its own.
+	first-party-provider #true
 	class "anthropic" {
 		revision ">=3.7 <4.6" {
 			thinking-efforts "minimal" "low" "medium" "high" "xhigh"

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

@@ -16,6 +16,7 @@ provider "deepseek" {
 	// bare alias carries no family, and the retired `v4-flash`/`-vision-exp` ids
 	// are still accepted and billed at the Flash card.
 	models "deepseek-flash" "deepseek-v4-flash" "deepseek-v4-flash-vision-exp" {
+		clamp-output-to-model-max #true
 		time-based-cost {
 			off-peak-multiplier 0.5
 			peak-windows {

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

@@ -1,6 +1,16 @@
 // Provider-wire compat for "opencode-go"; regenerated from the frozen census using deployment contract selectors.
 
 provider "opencode-go" {
+	// The Flash lanes accept images despite their IDs lacking a vision suffix;
+	// verified against the gateway (unguessable pixel-rendered text read back
+	// verbatim). Mirrors the OpenRouter carve-out for deepseek-v4.1-flash.
+	// The modality is declared here as well: live gateway discovery seeds
+	// `input: ["text"]`, and the wire guard requires the declared modality, so
+	// clearing `strip-image-input` alone would leave the lane text-only.
+	models "deepseek-flash" "deepseek-v4.1-flash" {
+		input-modalities "text" "image"
+		strip-image-input #false
+	}
 	// Replaces the OpenCode reasoning-model whenThinking fallback object.
 	when-thinking {
 		requires-reasoning-content-for-tool-calls #true

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

@@ -1,6 +1,10 @@
 // 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.