@oh-my-pi/pi-catalog 18.2.7 → 18.2.9

package/CHANGELOG.md

@@ -2,6 +2,29 @@
 
 ## [Unreleased]
 
+### Changed
+
+- Updated default models for anthropic, amazon-bedrock, cloudflare-ai-gateway, kilo, litellm, opencode-zen, vercel-ai-gateway, and zenmux to claude-opus-5-5
+- Updated default model for commandcode to claude-sonnet-5
+
+## [18.2.9] - 2026-09-22
+
+### Fixed
+
+- Local OpenAI-compatible backends—including llama.cpp, LM Studio, vLLM, Ollama, and custom loopback or LAN hosts—now support maxTokens values above 64k without incorrectly applying the hosted OpenAI limit.
+- Fixed Union Alpha requests on OpenCode Go and Zen by using the correct Messages API.
+- Updated SuperGrok’s Grok 4.7 metadata to expose its documented 500K context window and low, medium, high, and xhigh thinking levels.
+- Fixed GPT-OSS tool interactions on Google Antigravity and Gemini CLI so function-call history is preserved correctly.
+- Fixed Devin model discovery for Enterprise credentials by falling back to the legacy Windsurf catalog when native discovery does not return the full model list.
+- Curated Xiaomi Token Plan (China) MiMo V2.6 metadata: context/output limits, reasoning, and image input ([#12841](https://github.com/can1357/oh-my-pi/pull/12841) by [@roboomp](https://github.com/roboomp)).
+
+## [18.2.8] - 2026-09-21
+
+### Added
+
+- Expanded OpenRouter provider support with embedding, reranking, video generation, text-to-speech, and speech-to-text capabilities, including five new speech-to-text models.
+- Added speech-to-text support to the OpenAI provider.
+
 ## [18.2.7] - 2026-09-21
 
 ### Added
@@ -181,9 +204,9 @@
 ### Fixed
 
 - Fixed OpenCode Go/Zen live model discovery (`GET /v1/models`) missing `x-opencode-session` and omp's `User-Agent`: discovery requests now attribute with the stable install id so the requests OpenCode flags as `Bun fetch` carry the required session header.
-	- Fixed GPT-6 Astra requests through GitHub Copilot failing with an unsupported endpoint error ([#10874](https://github.com/can1357/oh-my-pi/pull/10874) by [@xpcmdshell](https://github.com/xpcmdshell)).
-	- Fixed GPT-6 Astra showing as free with a 272K-token window in the OpenAI Codex catalog by applying its documented pricing; `/extended-context` enables the wire-advertised 872K-token maximum ([#10980](https://github.com/can1357/oh-my-pi/pull/10980) by [@H4vC](https://github.com/H4vC)).
-	- Made extended-context catalog rebuilds faster by resolving each model's maximum window once per process ([#11039](https://github.com/can1357/oh-my-pi/pull/11039) by [@H4vC](https://github.com/H4vC)).
+   - Fixed GPT-6 Astra requests through GitHub Copilot failing with an unsupported endpoint error ([#10874](https://github.com/can1357/oh-my-pi/pull/10874) by [@xpcmdshell](https://github.com/xpcmdshell)).
+   - Fixed GPT-6 Astra showing as free with a 272K-token window in the OpenAI Codex catalog by applying its documented pricing; `/extended-context` enables the wire-advertised 872K-token maximum ([#10980](https://github.com/can1357/oh-my-pi/pull/10980) by [@H4vC](https://github.com/H4vC)).
+   - Made extended-context catalog rebuilds faster by resolving each model's maximum window once per process ([#11039](https://github.com/can1357/oh-my-pi/pull/11039) by [@H4vC](https://github.com/H4vC)).
 
 ## [18.1.9] - 2026-09-04
 

package/README.md

@@ -4,18 +4,18 @@ Model catalog for [oh-my-pi](https://github.com/can1357/oh-my-pi): bundled model
 
 ## What's inside
 
-| Module                          | Purpose                                                                                                     |
-| ------------------------------- | ----------------------------------------------------------------------------------------------------------- |
-| `models.json` + `models`        | Bundled model database (pricing, context windows, modalities, thinking support)                             |
-| `provider-models`               | Provider catalog descriptors (`CATALOG_PROVIDERS`), per-provider model resolution rules                     |
-| `discovery`                     | Runtime model discovery for OpenAI-compatible endpoints, Gemini, Codex, Cursor, Antigravity, Ollama         |
-| `compat/rules`                  | Checked-in KDL policy tree: taxonomy (classes/families/revisions), class/provider cascade rules, runtime behavior vocabulary; compiled by `bun run gen:compat` into the committed `rules.json` |
+| Module                          | Purpose                                                                                                                                                                                                                                      |
+| ------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `models.json` + `models`        | Bundled model database (pricing, context windows, modalities, thinking support)                                                                                                                                                              |
+| `provider-models`               | Provider catalog descriptors (`CATALOG_PROVIDERS`), per-provider model resolution rules                                                                                                                                                      |
+| `discovery`                     | Runtime model discovery for OpenAI-compatible endpoints, Gemini, Codex, Cursor, Antigravity, Ollama                                                                                                                                          |
+| `compat/rules`                  | Checked-in KDL policy tree: taxonomy (classes/families/revisions), class/provider cascade rules, runtime behavior vocabulary; compiled by `bun run gen:compat` into the committed `rules.json`                                               |
 | `compat`                        | The rule engine: `classifyModel` (taxonomy), `resolveModelPolicy` (cascade), behavior accessors (`api-routes`, `model-limits`, `exclude-models`, `pricing-peer`), collapse, and OpenAI/Anthropic wire builders that consume resolved records |
-| `identity`                      | Mechanical id utilities: reference resolution against the bundled index, dialects, selection priority, tokenizer families                                |
-| `model-thinking`                | Runtime thinking helpers (`getSupportedEfforts`, effort clamping/mapping, wire-id routing) over resolved model records                                   |
-| `model-manager` / `model-cache` | Runtime model registry with discovery refresh and on-disk caching                                           |
-| `wire`                          | Wire-level helpers: Codex, Gemini headers, GitHub Copilot                                                   |
-| `effort`                        | Reasoning-effort level definitions                                                                          |
+| `identity`                      | Mechanical id utilities: reference resolution against the bundled index, dialects, selection priority, tokenizer families                                                                                                                    |
+| `model-thinking`                | Runtime thinking helpers (`getSupportedEfforts`, effort clamping/mapping, wire-id routing) over resolved model records                                                                                                                       |
+| `model-manager` / `model-cache` | Runtime model registry with discovery refresh and on-disk caching                                                                                                                                                                            |
+| `wire`                          | Wire-level helpers: Codex, Gemini headers, GitHub Copilot                                                                                                                                                                                    |
+| `effort`                        | Reasoning-effort level definitions                                                                                                                                                                                                           |
 
 Import from subpaths (`@oh-my-pi/pi-catalog/<module>`) or the root barrel.
 
@@ -42,13 +42,13 @@ Ships TypeScript source directly (no build step); requires Bun ≥ 1.3.14.
 
 The `models` subpath (also exported from the root) provides timestamp-aware pricing helpers:
 
-| API | Result |
-| --- | --- |
-| `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. |
-| `getTimeBasedPricingPeriod(cost, timestamp?)` | Returns `"peak"`, `"off-peak"`, or `undefined` without a schedule. |
-| `getNextTimeBasedPricingTransition(cost, timestamp?)` | Returns the next actual peak/off-peak change strictly after the timestamp, or `undefined` if none exists. |
+| API                                                               | Result                                                                                                    |
+| ----------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------- |
+| `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.                                                              |
+| `getTimeBasedPricingPeriod(cost, timestamp?)`                     | Returns `"peak"`, `"off-peak"`, or `undefined` without a schedule.                                        |
+| `getNextTimeBasedPricingTransition(cost, timestamp?)`             | Returns the next actual peak/off-peak change strictly after the timestamp, or `undefined` if none exists. |
 
 Timestamps are Unix milliseconds; omitted timestamps use the current time for scheduled pricing. Flat token prices are unaffected. Pricing selects the latest applicable effective rate card, then its long-context tier, then the peak/off-peak multiplier. A transition query concerns the recurring tariff, not dated rate-card changes.
 

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" | "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" | "local" | "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" | "stencil" | "synthetic" | "tavily" | "together" | "typesafe" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "web" | "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" | "local" | "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" | "singularityapi-dev" | "singularityapi-tech" | "stencil" | "synthetic" | "tavily" | "together" | "typesafe" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "web" | "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" | "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" | "stencil" | "synthetic" | "tavily" | "together" | "typesafe" | "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" | "singularityapi-dev" | "singularityapi-tech" | "stencil" | "synthetic" | "tavily" | "together" | "typesafe" | "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/compat/axes.d.ts

@@ -25,8 +25,8 @@ export interface AxisDef {
     shape: AxisShape;
     /** Wire axes only: records this key exists on. */
     records?: readonly CompatRecordName[];
-    /** Closed value vocabulary for string scalars / string arrays. */
-    values?: readonly string[];
+    /** Closed value vocabulary for scalars / arrays (strings, and booleans on flag axes). */
+    values?: readonly (string | boolean)[];
     /**
      * Object axes only: payload child names are literal wire JSON keys copied
      * verbatim (`extra-body`). Default object payloads author kebab-case names

package/dist/types/compat/catalog-policy.d.ts

@@ -0,0 +1,7 @@
+import type { Model } from "../types.js";
+/**
+ * The `catalog` axis bag authored in the KDL cascade rules for a model,
+ * memoized per distinct model. Frozen: the bag is shared across callers, so it
+ * must stay read-only — extract the value you need rather than handing it on.
+ */
+export declare function resolveCatalogPolicy(model: Model): Readonly<Record<string, unknown>>;

package/dist/types/compat/output-limits.d.ts

@@ -0,0 +1,3 @@
+import type { Model } from "../types.js";
+/** Whether discovery and transport policy allow preserving a caller's output cap. */
+export declare function supportsOutputTokenLimit(model: Model): boolean;

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

@@ -1,2 +1,2 @@
 /** Every chat-model provider with a catalog entry (`default-model` in its provider rules). */
-export type KnownProvider = "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" | "firepass" | "fireworks" | "github-copilot" | "gitlab-duo" | "gitlab-duo-agent" | "gmi-cloud" | "google" | "google-antigravity" | "google-gemini-cli" | "google-vertex" | "groq" | "huggingface" | "kilo" | "kimi-code" | "litellm" | "lm-studio" | "local" | "meta" | "minimax" | "minimax-code" | "minimax-code-cn" | "mistral" | "moonshot" | "muse-code" | "nanogpt" | "novita" | "nvidia" | "ollama" | "ollama-cloud" | "openai" | "openai-codex" | "opencode-go" | "opencode-zen" | "openrouter" | "qianfan" | "qwen-portal" | "sakana" | "siliconflow" | "siliconflow-cn" | "synthetic" | "together" | "typesafe" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "web" | "xai" | "xai-oauth" | "xiaomi" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "xiaomi-token-plan-sgp" | "yolo-auto" | "zai" | "zenmux" | "zhipu-coding-plan";
+export type KnownProvider = "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" | "firepass" | "fireworks" | "github-copilot" | "gitlab-duo" | "gitlab-duo-agent" | "gmi-cloud" | "google" | "google-antigravity" | "google-gemini-cli" | "google-vertex" | "groq" | "huggingface" | "kilo" | "kimi-code" | "litellm" | "lm-studio" | "local" | "meta" | "minimax" | "minimax-code" | "minimax-code-cn" | "mistral" | "moonshot" | "muse-code" | "nanogpt" | "novita" | "nvidia" | "ollama" | "ollama-cloud" | "openai" | "openai-codex" | "opencode-go" | "opencode-zen" | "openrouter" | "qianfan" | "qwen-portal" | "sakana" | "siliconflow" | "siliconflow-cn" | "singularityapi-dev" | "singularityapi-tech" | "synthetic" | "together" | "typesafe" | "umans" | "venice" | "vercel-ai-gateway" | "vllm" | "wafer-serverless" | "web" | "xai" | "xai-oauth" | "xiaomi" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "xiaomi-token-plan-sgp" | "yolo-auto" | "zai" | "zenmux" | "zhipu-coding-plan";

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

@@ -0,0 +1,5 @@
+import type { Model } from "../types.js";
+/** Whether the transport exposes native tools that an empty caller catalog cannot disable. */
+export declare function requiresNativeTools(model: Model): boolean;
+/** Whether disabling tools requires a history without prior tool calls or results. */
+export declare function requiresToolFreeHistoryForToolOptOut(model: Model): boolean;

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

@@ -5,7 +5,7 @@
  * `behavior.ts`, `resolve.ts`) exposes to consumers.
  */
 import type { Effort } from "../effort.js";
-import type { Api, ThinkingControlMode, TokenCost } from "../types.js";
+import type { Api, KindApiKind, ThinkingControlMode, TokenCost } from "../types.js";
 import type { RevisionOp } from "./revision.js";
 /** Class-membership matcher kinds, most to least specific. */
 export type MatcherKind = "exact" | "bounded" | "namespace" | "prefix" | "glob";
@@ -620,7 +620,7 @@ export interface CompiledProvider {
     /** Present only for providers enrolled in `generate-models.ts` discovery. */
     discovery?: CompiledProviderDiscovery;
     /** Non-chat model kinds mapped to their runtime transport APIs. */
-    kindApis?: Partial<Record<"image" | "tts" | "stt", Api>>;
+    kindApis?: Partial<Record<KindApiKind, Api>>;
     /** Authored bundled rows, when the provider cannot be discovered at generation time. */
     seed?: CompiledSeed;
 }

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

@@ -661,3 +661,29 @@ export interface CharmHyperModelManagerConfig {
  * alive by a stale bundled row.
  */
 export declare function charmHyperModelManagerOptions(config?: CharmHyperModelManagerConfig): ModelManagerOptions<"openai-completions">;
+export interface SingularityApiModelManagerConfig {
+    apiKey?: string;
+    baseUrl?: string;
+    fetch?: FetchImpl;
+}
+/**
+ * `singularityapi-dev` — SingularityAPI's pay-as-you-go universal gateway
+ * (`api.singularityapi.dev`): chat completions over a 300+ model catalog,
+ * plus image generation for the rows that advertise it.
+ * `GET /v1/models` publishes each row's per-endpoint capabilities — context
+ * window, max output tokens, and per-million pricing as 12-decimal strings —
+ * with `cache-control: no-store`, so discovery reads limits and tariffs
+ * straight off the wire and the endpoint list picks each row's transport.
+ * Rows without a reasoning vocabulary stay non-reasoning; reviewed KDL rules
+ * own the ladders the gateway leaves implicit (DeepSeek Flash/Pro, GPT-5.6
+ * flagships), because a model discovered as non-reasoning never sends a
+ * `reasoning_effort` and the gateway requires one alongside tools.
+ */
+export declare function singularityApiDevModelManagerOptions(config?: SingularityApiModelManagerConfig): ModelManagerOptions<Api>;
+/**
+ * `singularityapi-tech` — SingularityAPI's slot-reserved DeepSeek lanes
+ * (`api.singularityapi.tech`). `/v1/models` answers with bare `{id}` rows and
+ * no capability metadata, so rows keep the discovery defaults and the
+ * reviewed KDL rules own the wire shape, limits patch, and effort ladder.
+ */
+export declare function singularityApiTechModelManagerOptions(config?: SingularityApiModelManagerConfig): ModelManagerOptions<Api>;

package/dist/types/types.d.ts

@@ -5,13 +5,16 @@ export type { KnownProvider } from "./provider-models/descriptors.js";
 export type KnownApi = "openai-completions" | "openai-responses" | "openrouter" | "openai-codex-responses" | "azure-openai-responses" | "anthropic-messages" | "bedrock-converse-stream" | "google-generative-ai" | "google-gemini-cli" | "google-vertex" | "ollama-chat" | "cursor-agent" | "gitlab-duo-agent" | "devin-agent";
 export type Api = KnownApi | (string & {});
 /** Catalog kinds used to isolate role-specific runners from session chat models. */
-export declare const MODEL_KINDS: readonly ["chat", "tiny", "image", "tts", "stt", "search", "judge"];
+export declare const MODEL_KINDS: readonly ["chat", "tiny", "image", "tts", "stt", "search", "judge", "embedding", "rerank", "video"];
 /** Technical capability of a catalog model; absent model kinds mean chat. */
 export type ModelKind = (typeof MODEL_KINDS)[number];
+/** Kinds a provider maps to a runner transport through `kind-apis` in its KDL; discovery drops rows of these kinds when the provider declares no API. */
+export declare const KIND_API_KINDS: readonly ["image", "tts", "stt", "embedding", "rerank", "video"];
+export type KindApiKind = (typeof KIND_API_KINDS)[number];
 /** Grounding transport available to chat models selected by the web role. */
 export type WebSearchGrounding = "gemini" | "anthropic" | "codex" | "xai" | "openrouter";
 /** Non-chat runner protocols accepted by catalog seeds, outside the chat dispatch union. */
-export declare const RUNNER_APIS: readonly ["local-inference", "web-search", "typesafe", "openrouter-decisions", "openai-images", "openrouter-images", "xai-tts", "openai-speech"];
+export declare const RUNNER_APIS: readonly ["local-inference", "web-search", "typesafe", "openrouter-decisions", "openai-images", "openrouter-images", "xai-tts", "openai-speech", "openai-embeddings", "openrouter-rerank", "openrouter-video", "openai-transcriptions"];
 /** Resolve a model's kind while preserving chat semantics for existing catalog rows. */
 export declare function modelKind(model: Pick<Model, "kind">): ModelKind;
 /** Canonical thinking transport used by a model. */

package/dist/types/wire/singularityapi.d.ts

@@ -0,0 +1,29 @@
+/**
+ * SingularityAPI gateways, shared so a host migration — or a self-hosted
+ * proxy override — touches a single module.
+ *
+ * SingularityAPI ships two unrelated products behind one brand: the
+ * pay-as-you-go universal gateway (`singularityapi-dev`) and the
+ * slot-reserved DeepSeek lanes (`singularityapi-tech`). They accept disjoint
+ * keys, so each provider resolves its own canonical endpoint from here.
+ */
+export declare const SINGULARITYAPI_DEV_API_BASE_URL = "https://api.singularityapi.dev/v1";
+export declare const SINGULARITYAPI_TECH_API_BASE_URL = "https://api.singularityapi.tech/v1";
+/**
+ * Resolve a configured SingularityAPI base URL onto its gateway's `/v1`
+ * surface.
+ *
+ * Every consumer must agree on this, because they key different things off the
+ * result: discovery and inference target it, and the model-cache namespace is
+ * hashed from it. `ModelRegistry` hashes the raw configured value while the
+ * model-manager options hash a `/v1`-suffixed one, so a disagreement would
+ * split the namespace discovery writes from the one the registry reads and the
+ * authoritative roster would never come back.
+ *
+ * `canonical` names the product the caller belongs to: a blank or
+ * whitespace-only value means "not configured" and resolves to that canonical
+ * host, while anything else keeps its host and gains the `/v1` segment if it
+ * omits one. Passing the wrong product's canonical URL would silently point a
+ * provider at the other gateway, so there is no default.
+ */
+export declare function normalizeSingularityApiBaseUrl(baseUrl: string | undefined, canonical: string): string;

package/package.json

@@ -1,51 +1,25 @@
 {
-	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.2.7",
+	"version": "18.2.9",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
+	"keywords": [
+		"ai",
+		"catalog",
+		"discovery",
+		"llm",
+		"models"
+	],
 	"homepage": "https://omp.sh",
-	"author": "Stencil Labs, Inc.",
+	"bugs": {
+		"url": "https://github.com/can1357/oh-my-pi/issues"
+	},
 	"license": "MIT",
+	"author": "Stencil Labs, Inc.",
 	"repository": {
 		"type": "git",
 		"url": "git+https://github.com/can1357/oh-my-pi.git",
 		"directory": "packages/catalog"
 	},
-	"bugs": {
-		"url": "https://github.com/can1357/oh-my-pi/issues"
-	},
-	"keywords": [
-		"ai",
-		"llm",
-		"models",
-		"catalog",
-		"discovery"
-	],
-	"main": "./src/index.ts",
-	"types": "./dist/types/index.d.ts",
-	"scripts": {
-		"check": "oxlint . && oxfmt --check --no-error-on-unmatched-pattern 'src/**/*.{ts,tsx}' '{test,bench,examples,scripts}/**/*.ts' '*.ts' && bun run check:types",
-		"check:types": "tsgo -p tsconfig.json --noEmit",
-		"lint": "oxlint .",
-		"test": "bun test --parallel",
-		"fix": "oxlint --fix --fix-suggestions . && bun run fmt",
-		"fmt": "oxfmt --no-error-on-unmatched-pattern 'src/**/*.{ts,tsx}' '{test,bench,examples,scripts}/**/*.ts' '*.ts'",
-		"gen:models": "bun scripts/generate-models.ts",
-		"gen:compat": "bun scripts/compile-compat.ts",
-		"gen:proto": "bun scripts/generate-protocols.ts"
-	},
-	"dependencies": {
-		"@oh-my-pi/omptype": "18.2.7",
-		"@oh-my-pi/pi-utils": "18.2.7"
-	},
-	"devDependencies": {
-		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.2.7",
-		"@types/bun": "^1.3.14"
-	},
-	"engines": {
-		"bun": ">=1.3.14"
-	},
 	"files": [
 		"src",
 		"README.md",
@@ -54,6 +28,9 @@
 		"THIRD-PARTY-NOTICES.txt",
 		"dist/types"
 	],
+	"type": "module",
+	"main": "./src/index.ts",
+	"types": "./dist/types/index.d.ts",
 	"exports": {
 		".": {
 			"types": "./dist/types/index.d.ts",
@@ -100,5 +77,28 @@
 			"import": "./src/*.ts"
 		},
 		"./*.js": "./src/*.ts"
+	},
+	"scripts": {
+		"check": "oxlint . && oxfmt --check --no-error-on-unmatched-pattern 'src/**/*.{ts,tsx}' '{test,bench,examples,scripts}/**/*.ts' '*.ts' && bun run check:types",
+		"check:types": "tsgo -p tsconfig.json --noEmit",
+		"lint": "oxlint .",
+		"test": "bun test --parallel",
+		"fix": "oxlint --fix --fix-suggestions . && bun run fmt",
+		"fmt": "oxfmt --no-error-on-unmatched-pattern 'src/**/*.{ts,tsx}' '{test,bench,examples,scripts}/**/*.ts' '*.ts'",
+		"gen:models": "bun scripts/generate-models.ts",
+		"gen:compat": "bun scripts/compile-compat.ts",
+		"gen:proto": "bun scripts/generate-protocols.ts"
+	},
+	"dependencies": {
+		"@oh-my-pi/omptype": "18.2.9",
+		"@oh-my-pi/pi-utils": "18.2.9"
+	},
+	"devDependencies": {
+		"@bgotink/kdl": "0.4.0",
+		"@oh-my-pi/pi-ai": "18.2.9",
+		"@types/bun": "^1.3.14"
+	},
+	"engines": {
+		"bun": ">=1.3.14"
 	}
 }

package/src/compat/auth-ids.ts

@@ -67,6 +67,8 @@ export type AuthProviderId =
 	| "sakana"
 	| "siliconflow"
 	| "siliconflow-cn"
+	| "singularityapi-dev"
+	| "singularityapi-tech"
 	| "stencil"
 	| "synthetic"
 	| "tavily"
@@ -146,6 +148,8 @@ export type LoginProviderId =
 	| "sakana"
 	| "siliconflow"
 	| "siliconflow-cn"
+	| "singularityapi-dev"
+	| "singularityapi-tech"
 	| "stencil"
 	| "synthetic"
 	| "tavily"

package/src/compat/axes.ts

@@ -29,8 +29,8 @@ export interface AxisDef {
 	shape: AxisShape;
 	/** Wire axes only: records this key exists on. */
 	records?: readonly CompatRecordName[];
-	/** Closed value vocabulary for string scalars / string arrays. */
-	values?: readonly string[];
+	/** Closed value vocabulary for scalars / arrays (strings, and booleans on flag axes). */
+	values?: readonly (string | boolean)[];
 	/**
 	 * Object axes only: payload child names are literal wire JSON keys copied
 	 * verbatim (`extra-body`). Default object payloads author kebab-case names
@@ -284,6 +284,25 @@ export const AXES: Readonly<Record<string, AxisDef>> = {
 		shape: "scalar",
 		values: ["freeform", "function"],
 	},
+	"requires-native-tools": { key: "requiresNativeTools", set: "catalog", shape: "scalar", values: [true, false] },
+	"requires-tool-free-history-for-tool-opt-out": {
+		key: "requiresToolFreeHistoryForToolOptOut",
+		set: "catalog",
+		shape: "scalar",
+		values: [true, false],
+	},
+	"preserves-max-output-tokens": {
+		key: "preservesMaxOutputTokens",
+		set: "catalog",
+		shape: "scalar",
+		values: [true, false],
+	},
+	"omit-max-output-tokens": {
+		key: "omitMaxOutputTokens",
+		set: "catalog",
+		shape: "scalar",
+		values: [true, false],
+	},
 	"clamp-context-override": { key: "clampContextOverride", set: "catalog", shape: "scalar" },
 	"context-promotion-target": { key: "contextPromotionTarget", set: "catalog", shape: "scalar" },
 	"context-window-floor": { key: "contextWindowFloor", set: "catalog", shape: "scalar" },

package/src/compat/catalog-policy.ts

@@ -0,0 +1,28 @@
+import { toModelSpec } from "../provider-models/bundled-references";
+import type { Model } from "../types";
+import { resolveModelPolicy } from "./resolve";
+
+/**
+ * Rule-owned catalog-axis policy by model. Resolve once per process rather
+ * than walking the static policy cascade on every accessor call. The key
+ * mirrors the cascade target inputs — `providerType ?? provider`, api, model
+ * id, and the reasoning flag; identity-derived class/family/revision facts are
+ * a pure function of provider + id. Bounded: one entry per distinct model.
+ */
+const catalogPolicyCache = new Map<string, Readonly<Record<string, unknown>>>();
+const CATALOG_POLICY_CACHE_MAX = 8192;
+
+/**
+ * The `catalog` axis bag authored in the KDL cascade rules for a model,
+ * memoized per distinct model. Frozen: the bag is shared across callers, so it
+ * must stay read-only — extract the value you need rather than handing it on.
+ */
+export function resolveCatalogPolicy(model: Model): Readonly<Record<string, unknown>> {
+	const key = `${model.provider}\u0000${model.providerType ?? ""}\u0000${model.api}\u0000${model.id}\u0000${model.reasoning ? 1 : 0}`;
+	const cached = catalogPolicyCache.get(key);
+	if (cached !== undefined) return cached;
+	const policy = Object.freeze({ ...resolveModelPolicy(toModelSpec(model)).catalog });
+	if (catalogPolicyCache.size >= CATALOG_POLICY_CACHE_MAX) catalogPolicyCache.clear();
+	catalogPolicyCache.set(key, policy);
+	return policy;
+}

package/src/compat/output-limits.ts

@@ -0,0 +1,12 @@
+import type { Model } from "../types";
+import { resolveCatalogPolicy } from "./catalog-policy";
+
+/** Whether discovery and transport policy allow preserving a caller's output cap. */
+export function supportsOutputTokenLimit(model: Model): boolean {
+	const policy = resolveCatalogPolicy(model);
+	return (
+		model.omitMaxOutputTokens !== true &&
+		policy.omitMaxOutputTokens !== true &&
+		policy.preservesMaxOutputTokens !== false
+	);
+}

package/src/compat/provider-ids.ts

@@ -61,6 +61,8 @@ export type KnownProvider =
 	| "sakana"
 	| "siliconflow"
 	| "siliconflow-cn"
+	| "singularityapi-dev"
+	| "singularityapi-tech"
 	| "synthetic"
 	| "together"
 	| "typesafe"

package/src/compat/resolve.ts

@@ -562,7 +562,7 @@ function detectOpenAICompat(
 		dropThinkingWhenReasoningEffort: false,
 		nativeKimiK3Reasoning: false,
 		zaiReasoningEffortDialect: false,
-		clampOutputToModelMax: false,
+		clampOutputToModelMax: d.isLocalOpenAICompatBackend,
 		stripImageInput: false,
 		thinkingLoopGuard: undefined,
 		rejectRootObjectUnion: false,
@@ -765,7 +765,9 @@ function resolveOpenAIResponsesPolicy(
 		wireModelIdMode: isOpenRouter ? "openrouter" : "raw",
 		toolSchemaFlavor: facts.is("kimi") ? "moonshot-mfjs" : undefined,
 		alwaysSendMaxTokens: facts.is("kimi"),
-		clampOutputToModelMax: false,
+		clampOutputToModelMax:
+			PROXY_OPENAI_COMPAT_PROVIDERS[backendProvider] !== true &&
+			(LOCAL_OPENAI_COMPAT_PROVIDERS[backendProvider] === true || hasLocalLoopbackBaseUrl(baseUrl)),
 		supportsObfuscationOptOut: isOpenAIUrl || provider === "openai",
 		officialEndpoint: isOfficialOpenAIEndpoint(provider, baseUrl),
 		harmonyLeakMitigation: false,