refactor(catalog)!: split model catalog from pi-ai

Move bundled models, model cache/manager, thinking metadata, effort helpers, provider descriptors/discovery, wire constants, and model identity utilities into the new @oh-my-pi/pi-catalog package. Update pi-ai to keep provider runtime/auth concerns, move catalog provider metadata into CATALOG_PROVIDERS, and migrate coding-agent, agent, stats, docs, and tests to import catalog values from pi-catalog. Split coding-agent model registry helpers into discovery, roles, and models config modules while preserving registry orchestration. BREAKING CHANGE: @oh-my-pi/pi-ai no longer exports catalog subpaths such as /models, /model-cache, /model-manager, /model-thinking, /effort, /provider-models*, discovery helpers, and provider wire constants; use the matching @oh-my-pi/pi-catalog subpaths instead.
2026-06-10 04:05:19 +02:00
parent b3166b3dd2
commit 1b9d9d0851
320 changed files with 4155 additions and 3194 deletions
@@ -11,6 +11,7 @@ This repo contains multiple packages, but **`packages/coding-agent/`** is the pr
 | Package                 | Description                                          |
 | ----------------------- | ---------------------------------------------------- |
 | `packages/ai`           | Multi-provider LLM client with streaming support     |
+| `packages/catalog`      | Model catalog: bundled models.json, provider descriptors, model identity/classification |
 | `packages/agent`        | Agent runtime with tool calling and state management |
 | `packages/coding-agent` | Main CLI application (primary focus)                 |
 | `packages/tui`          | Terminal UI library with differential rendering      |
@@ -19,6 +20,8 @@ This repo contains multiple packages, but **`packages/coding-agent/`** is the pr
 | `packages/utils`        | Shared utilities (logger, streams, temp files)       |
 | `crates/pi-natives`     | Rust crate for performance-critical text/grep ops    |

+**Catalog import convention**: code in this repo imports catalog *values* (bundled models, model-thinking helpers, identity, descriptors, model manager/cache) from `@oh-my-pi/pi-catalog/<module>` — never via `@oh-my-pi/pi-ai`. The pi-ai barrel re-exports only the model/effort *types* its own signatures use (`Model`, `Api`, `ThinkingConfig`, `Effort`, …); type-only imports of those from `@oh-my-pi/pi-ai` are fine.
+
 ## Code Quality

 - No `any` unless absolutely necessary.
@@ -147,15 +150,15 @@ Manual reader loops only when the protocol requires it (SSE, streaming JSON-RPC)

 ## Generated Files

-**NEVER edit `packages/ai/src/models.json` directly.** It is generated from upstream sources (models.dev, provider catalog discovery, OpenCode docs) by `packages/ai/scripts/generate-models.ts` and the descriptors/resolvers in `packages/ai/src/provider-models/`. Hand-edits get overwritten on the next regen.
+**NEVER edit `packages/catalog/src/models.json` directly.** It is generated from upstream sources (models.dev, provider catalog discovery, OpenCode docs) by `packages/catalog/scripts/generate-models.ts` and the descriptors/resolvers in `packages/catalog/src/provider-models/`. Hand-edits get overwritten on the next regen.

 To change an entry, fix the source:
- **Resolution rules / per-id overrides** → relevant resolver in `packages/ai/src/provider-models/openai-compat.ts` (e.g. `createOpenCodeApiResolution`'s id-override map).
- **Provider descriptors** (filtering, transforms, defaults, headers, compat overrides) → `packages/ai/src/provider-models/descriptors.ts` or the provider-specific descriptor.
- **Generator-level fixups** (premium multipliers, codex pricing fallback, fallback models, post-processing) → `packages/ai/scripts/generate-models.ts`.
- **Thinking metadata / generated policies** → `packages/ai/src/model-thinking.ts` (`applyGeneratedModelPolicies`).
+- **Resolution rules / per-id overrides** → relevant resolver in `packages/catalog/src/provider-models/openai-compat.ts` (e.g. `createOpenCodeApiResolution`'s id-override map).
+- **Provider catalog entries** (default model, discovery factory/flags) → the `CATALOG_PROVIDERS` table in `packages/catalog/src/provider-models/descriptors.ts`.
+- **Generator-level fixups** (premium multipliers, codex pricing fallback, fallback models, post-processing) → `packages/catalog/scripts/generate-models.ts`.
+- **Thinking metadata / generated policies** → `packages/catalog/src/model-thinking.ts` (`applyGeneratedModelPolicies`); model-id classification (family/version parsing) lives in `packages/catalog/src/identity/classify.ts`.

-Regenerate with `bun --cwd=packages/ai run generate-models` and commit `models.json` alongside the source change. Add a regression test against the **resolver/descriptor**, not the bundled JSON, so it survives upstream metadata shifts.
+Regenerate with `bun --cwd=packages/catalog run generate-models` and commit `models.json` alongside the source change. Add a regression test against the **resolver/descriptor**, not the bundled JSON, so it survives upstream metadata shifts.

 ## Logging

@@ -62,7 +62,7 @@
 			"!**/test-sessions.ts",
 			"!**/template.generated.ts",
 			"!**/docs-index.generated.ts",
-			"!**/gen/agent_pb.ts",
+			"!**/agent_pb.ts",
 			"!.worktrees/**/*",
 			"!.wt/**/*"
 		]
@@ -18,6 +18,7 @@
      "version": "15.10.10",
      "dependencies": {
        "@oh-my-pi/pi-ai": "catalog:",
+        "@oh-my-pi/pi-catalog": "catalog:",
        "@oh-my-pi/pi-natives": "catalog:",
        "@oh-my-pi/pi-utils": "catalog:",
        "@opentelemetry/api": "catalog:",
@@ -33,6 +34,7 @@
      "version": "15.10.10",
      "dependencies": {
        "@bufbuild/protobuf": "catalog:",
+        "@oh-my-pi/pi-catalog": "catalog:",
        "@oh-my-pi/pi-utils": "catalog:",
        "openai": "catalog:",
        "partial-json": "catalog:",
@@ -42,11 +44,24 @@
        "@types/bun": "catalog:",
      },
    },
+    "packages/catalog": {
+      "name": "@oh-my-pi/pi-catalog",
+      "version": "15.10.10",
+      "dependencies": {
+        "@bufbuild/protobuf": "catalog:",
+        "@oh-my-pi/pi-utils": "catalog:",
+        "zod": "catalog:",
+      },
+      "devDependencies": {
+        "@oh-my-pi/pi-ai": "catalog:",
+        "@types/bun": "catalog:",
+      },
+    },
    "packages/coding-agent": {
      "name": "@oh-my-pi/pi-coding-agent",
      "version": "15.10.10",
      "bin": {
-        "omp": "src/cli.ts",
+        "omp": "dist/cli.js",
      },
      "dependencies": {
        "@agentclientprotocol/sdk": "catalog:",
@@ -56,6 +71,7 @@
        "@oh-my-pi/omp-stats": "catalog:",
        "@oh-my-pi/pi-agent-core": "catalog:",
        "@oh-my-pi/pi-ai": "catalog:",
+        "@oh-my-pi/pi-catalog": "catalog:",
        "@oh-my-pi/pi-mnemopi": "catalog:",
        "@oh-my-pi/pi-natives": "catalog:",
        "@oh-my-pi/pi-tui": "catalog:",
@@ -132,6 +148,7 @@
      },
      "dependencies": {
        "@oh-my-pi/pi-ai": "catalog:",
+        "@oh-my-pi/pi-catalog": "catalog:",
        "@oh-my-pi/pi-utils": "catalog:",
        "@tailwindcss/node": "catalog:",
        "chart.js": "catalog:",
@@ -252,6 +269,7 @@
    "@oh-my-pi/omp-stats": "15.10.10",
    "@oh-my-pi/pi-agent-core": "15.10.10",
    "@oh-my-pi/pi-ai": "15.10.10",
+    "@oh-my-pi/pi-catalog": "15.10.10",
    "@oh-my-pi/pi-coding-agent": "15.10.10",
    "@oh-my-pi/pi-mnemopi": "15.10.10",
    "@oh-my-pi/pi-natives": "15.10.10",
@@ -641,6 +659,8 @@

    "@oh-my-pi/pi-ai": ["@oh-my-pi/pi-ai@workspace:packages/ai"],

+    "@oh-my-pi/pi-catalog": ["@oh-my-pi/pi-catalog@workspace:packages/catalog"],
+
    "@oh-my-pi/pi-coding-agent": ["@oh-my-pi/pi-coding-agent@workspace:packages/coding-agent"],

    "@oh-my-pi/pi-mnemopi": ["@oh-my-pi/pi-mnemopi@workspace:packages/mnemopi"],
@@ -1,26 +1,42 @@
 # Adding a provider

-Providers in `packages/ai` are described by a single declarative
-`ProviderDefinition` and collected in one registry. Every scattered structure —
-the `KnownProvider` / `OAuthProvider` type unions, `PROVIDER_DESCRIPTORS`,
-`DEFAULT_MODEL_PER_PROVIDER`, the `serviceProviderMap` env-key fallbacks, the
-`/login` provider list, the `refreshOAuthToken` / `AuthStorage.login` dispatch,
-and the coding-agent callback maps — is **derived** from that registry.
+A provider is described in two halves:
+
+- **Catalog half** (`packages/catalog`): one entry in the `CATALOG_PROVIDERS`
+  table (`packages/catalog/src/provider-models/descriptors.ts`) carrying the
+  `id`, `defaultModel`, runtime model-discovery factory, and catalog-generation
+  wiring. `KnownProvider`, `PROVIDER_DESCRIPTORS`, and
+  `DEFAULT_MODEL_PER_PROVIDER` are derived from this table.
+- **Auth half** (`packages/ai`): one declarative `ProviderDefinition` in the
+  registry carrying env-key fallbacks and login/refresh flows. The
+  `OAuthProvider` union, the env-key map, the `/login` provider list, the
+  `refreshOAuthToken` / `AuthStorage.login` dispatch, and the coding-agent
+  callback maps are derived from the registry.

 **Scope.** This is for a provider that reuses an existing wire API
 (`openai-completions`, `anthropic-messages`, `google-generative-ai`, …) — the
 common case for gateways and API-key providers, since stream dispatch keys on
 `model.api`, not `model.provider`. Adding a *new wire protocol* (a new
 `KnownApi`) is a separate task that also touches `stream.ts` dispatch,
-`api-registry.ts`, and `types.ts`.
+`api-registry.ts`, and the catalog `types.ts`.

 ## Shape

-For the common case, a provider is still **one new def file + one registry line**:
+For the common case, a provider is **one catalog entry + one def file + one registry line**:

-1. **Create `packages/ai/src/registry/<id>.ts`** exporting one
-   `export const <camelId>Provider = { … } as const satisfies ProviderDefinition;`.
-2. **Add it to the `ALL` array** in `packages/ai/src/registry/registry.ts`
+1. **Add an entry to `CATALOG_PROVIDERS`** in
+   `packages/catalog/src/provider-models/descriptors.ts` with the `id`,
+   `defaultModel`, the plain API-key env var(s) as `envVars`, and (usually) a
+   `createModelManagerOptions` factory. For a
+   simple OpenAI-compatible gateway, build the factory in
+   `packages/catalog/src/provider-models/openai-compat.ts` or inline with the
+   exported `createSimpleOpenAICompletionsOptions(providerId, baseUrl, config)`.
+2. **Create `packages/ai/src/registry/<id>.ts`** exporting one
+   `export const <camelId>Provider = { … } as const satisfies ProviderDefinition;`
+   with the auth fields (`login`, …). Plain env-var names live in the catalog
+   entry's `envVars`; set `envKeys` only for computed resolvers (Foundry/ADC/
+   Bedrock-style probes).
+3. **Add it to the `ALL` array** in `packages/ai/src/registry/registry.ts`
   (one import + one array entry). `ALL` order is the `/login` list order for
   loginable providers.

@@ -34,26 +50,33 @@ For a **non-trivial provider-local OAuth flow**, put the implementation in
 file. The shared OAuth flow infrastructure it builds on lives in the same
 `registry/oauth/` directory.

-Either way, descriptors, default-model map, env-key map, login list, and refresh
-dispatch all update automatically, and the `KnownProvider` / `OAuthProvider`
-unions gain the new id by derivation.
+Descriptors, the default-model map, env-key map, login list, and refresh
+dispatch all update automatically; the `KnownProvider` union gains the new id
+from the catalog table and `OAuthProvider` from the registry.

-## `ProviderDefinition` fields
+## Field reference

-See `packages/ai/src/registry/types.ts` for the authoritative,
-JSDoc-annotated interface. Presence of a field opts the provider into a derived
-structure:
+**Catalog table entry** (`ProviderCatalogEntry`, see
+`packages/catalog/src/provider-models/descriptor-types.ts` for JSDoc):
+
+| Field | Effect |
+|---|---|
+| `id` | Required. Member of `KnownProvider`. |
+| `defaultModel` | Required. Preferred model when no explicit selection is made. |
+| `envVars` | Env var name(s), in order, for the runtime API-key fallback (`getEnvApiKey`). |
+| `createModelManagerOptions` | Runtime model-discovery factory. Present (and not `specialModelManager`) ⇒ appears in `PROVIDER_DESCRIPTORS`. |
+| `allowUnauthenticated` | Runtime creates a model manager even without a key. |
+| `dynamicModelsAuthoritative` | Successful discovery replaces bundled models. |
+| `catalogDiscovery` | `{ label, envVars?, oauthProvider?, allowUnauthenticated? }` for offline catalog generation (`generate-models.ts`). `envVars` here overrides the entry-level list when generation uses different credentials (e.g. `cursor`). |
+| `specialModelManager` | Bespoke runtime factory (`google-antigravity` / `google-gemini-cli` / `openai-codex`); excluded from `PROVIDER_DESCRIPTORS`. |
+
+**Registry definition** (`ProviderDefinition`, see
+`packages/ai/src/registry/types.ts`):

 | Field | Effect |
 |---|---|
 | `id`, `name` | Required. `name` shows in the `/login` list. |
-| `defaultModel` | Present ⇒ member of `KnownProvider` (a chat-model provider). |
-| `createModelManagerOptions` | Runtime model-discovery factory. Present (and not `specialModelManager`) ⇒ appears in `PROVIDER_DESCRIPTORS`. |
-| `allowUnauthenticated` | Runtime creates a model manager even without a key. |
-| `dynamicModelsAuthoritative` | Successful discovery replaces bundled models. |
-| `catalogDiscovery` | `{ label, envVars, oauthProvider?, allowUnauthenticated? }` for offline catalog generation (`generate-models.ts`). |
-| `specialModelManager` | Bespoke runtime factory (`google-antigravity` / `google-gemini-cli` / `openai-codex`); excluded from `PROVIDER_DESCRIPTORS`. |
-| `envKeys` | Env-var fallback for `getEnvApiKey`: a var name string or a `() => string \| undefined` resolver. |
+| `envKeys` | Computed env fallback for `getEnvApiKey`, overriding the catalog entry's `envVars`: a var name string or a `() => string \| undefined` resolver. Omit when `envVars` covers it. |
 | `login` | Interactive login. Present ⇒ member of `OAuthProvider`, shown in `/login`, dispatchable via `AuthStorage.login`. Returns an api-key `string` or `OAuthCredentials`. |
 | `refreshToken` | OAuth refresher; omit for static-token providers (the dispatch returns credentials unchanged). |
 | `storeCredentialsAs` | Store credentials under a different provider id (e.g. `openai-codex-device` ⇒ `openai-codex`). |
@@ -24,6 +24,7 @@
      "@oh-my-pi/omp-stats": "15.10.10",
      "@oh-my-pi/pi-agent-core": "15.10.10",
      "@oh-my-pi/pi-ai": "15.10.10",
+      "@oh-my-pi/pi-catalog": "15.10.10",
      "@oh-my-pi/pi-coding-agent": "15.10.10",
      "@oh-my-pi/pi-mnemopi": "15.10.10",
      "@oh-my-pi/pi-natives": "15.10.10",
@@ -152,7 +153,7 @@
    "publish": "bun run prepublishOnly && npm publish -ws --access public",
    "publish:dry": "bun run prepublishOnly && npm publish -ws --access public --dry-run",
    "release": "bun scripts/release.ts",
-    "generate-models": "bun --cwd=packages/ai run generate-models",
+    "generate-models": "bun --cwd=packages/catalog run generate-models",
    "generate-docs-index": "bun --cwd=packages/coding-agent run generate-docs-index",
    "generate-template": "bun --cwd=packages/coding-agent run generate-template",
    "check-spoofed-versions": "bun scripts/check-spoofed-versions.ts"
@@ -5,6 +5,7 @@
 ### Changed

 - Editorial pass over the compaction prompts: fixed garbled grammar and missing articles, RFC-keyed prohibitions, deduped restated instructions; parsed markers (`<read-files>`/`<modified-files>`/`<previous-summary>`) and all output-format headings left byte-identical
+- Catalog imports moved to the new `@oh-my-pi/pi-catalog` package: subpath imports (`calculateCost`, Codex wire constants) plus catalog values previously taken from the `@oh-my-pi/pi-ai` root (`getBundledModel`, `clampThinkingLevelForModel`), which pi-ai no longer re-exports; type-only `Model`/`Api`/`Effort` imports from pi-ai are unchanged

 ## [15.10.8] - 2026-06-09
 ### Added
@@ -36,6 +36,7 @@
 	},
 	"dependencies": {
 		"@oh-my-pi/pi-ai": "catalog:",
+		"@oh-my-pi/pi-catalog": "catalog:",
 		"@oh-my-pi/pi-natives": "catalog:",
 		"@oh-my-pi/pi-utils": "catalog:",
 		"@opentelemetry/api": "catalog:"
@@ -9,7 +9,6 @@ import {
 	type CursorExecHandlers,
 	type CursorToolResultHandler,
 	type Effort,
-	getBundledModel,
 	type ImageContent,
 	type Message,
 	type Model,
@@ -22,6 +21,7 @@ import {
 	type ToolChoice,
 	type ToolResultMessage,
 } from "@oh-my-pi/pi-ai";
+import { getBundledModel } from "@oh-my-pi/pi-catalog/models";
 import { abortReasonText, agentLoop, agentLoopContinue } from "./agent-loop";
 import type { AppendOnlyContextManager } from "./append-only-context";
 import type { HarmonyAuditEvent } from "./harmony-leak";
@@ -7,7 +7,6 @@

 import {
 	type AssistantMessage,
-	clampThinkingLevelForModel,
 	Effort,
 	type FetchImpl,
 	type Message,
@@ -15,6 +14,7 @@ import {
 	type Model,
 	type Usage,
 } from "@oh-my-pi/pi-ai";
+import { clampThinkingLevelForModel } from "@oh-my-pi/pi-catalog/model-thinking";
 import { countTokens } from "@oh-my-pi/pi-natives";
 import { logger, prompt } from "@oh-my-pi/pi-utils";
 import { type AgentTelemetry, instrumentedCompleteSimple } from "../telemetry";
@@ -12,12 +12,6 @@
 *   with `{ summary, shortSummary? }`.
 */

-import {
-	CODEX_BASE_URL,
-	getCodexAccountId,
-	OPENAI_HEADER_VALUES,
-	OPENAI_HEADERS,
-} from "@oh-my-pi/pi-ai/providers/openai-codex/constants";
 import { parseTextSignature } from "@oh-my-pi/pi-ai/providers/openai-responses-shared";
 import { transformMessages } from "@oh-my-pi/pi-ai/providers/transform-messages";
 import type { AssistantMessage, FetchImpl, Message, Model } from "@oh-my-pi/pi-ai/types";
@@ -26,6 +20,12 @@ import {
 	getOpenAIResponsesHistoryPayload,
 	normalizeResponsesToolCallId,
 } from "@oh-my-pi/pi-ai/utils";
+import {
+	CODEX_BASE_URL,
+	getCodexAccountId,
+	OPENAI_HEADER_VALUES,
+	OPENAI_HEADERS,
+} from "@oh-my-pi/pi-catalog/wire/codex";
 import { logger } from "@oh-my-pi/pi-utils";

 // ============================================================================
@@ -13,8 +13,8 @@ import {
 	type StopReason,
 	type ToolCall,
 } from "@oh-my-pi/pi-ai";
-import { calculateCost } from "@oh-my-pi/pi-ai/models";
 import { parseStreamingJson } from "@oh-my-pi/pi-ai/utils/json-parse";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
 import { readSseJson } from "@oh-my-pi/pi-utils";

 // Event stream adapter for proxy SSE events
@@ -9,7 +9,7 @@ import {
 } from "@oh-my-pi/pi-agent-core/compaction";
 import type { AssistantMessage, Model } from "@oh-my-pi/pi-ai";
 import * as ai from "@oh-my-pi/pi-ai";
-import { getBundledModel } from "@oh-my-pi/pi-ai/models";
+import { getBundledModel } from "@oh-my-pi/pi-catalog/models";

 // Pins the fix for the "raw 401 surfaced as Compaction failed:" bug.
 //
@@ -10,7 +10,7 @@ import {
 import { ThinkingLevel } from "@oh-my-pi/pi-agent-core/thinking";
 import type { AssistantMessage, Model } from "@oh-my-pi/pi-ai";
 import * as ai from "@oh-my-pi/pi-ai";
-import { getBundledModel } from "@oh-my-pi/pi-ai/models";
+import { getBundledModel } from "@oh-my-pi/pi-catalog/models";

 // Pins fix #1 of the compaction effort-override bug. Before this fix,
 // `generateHandoff` (and the three other compaction summarizers) hardcoded
@@ -4,7 +4,7 @@ import { AUTO_HANDOFF_THRESHOLD_FOCUS, generateHandoff, renderHandoffPrompt } fr
 import type { AssistantMessage, Model, ToolCall } from "@oh-my-pi/pi-ai";
 import * as ai from "@oh-my-pi/pi-ai";
 import { Effort } from "@oh-my-pi/pi-ai";
-import { getBundledModel } from "@oh-my-pi/pi-ai/models";
+import { getBundledModel } from "@oh-my-pi/pi-catalog/models";

 function createAssistantMessage(content: AssistantMessage["content"]): AssistantMessage {
 	return {
@@ -9,7 +9,7 @@ import {
 	signalListLabel,
 } from "@oh-my-pi/pi-agent-core/harmony-leak";
 import type { AssistantMessage, Model, ToolCall } from "@oh-my-pi/pi-ai";
-import { getBundledModel } from "@oh-my-pi/pi-ai";
+import { getBundledModel } from "@oh-my-pi/pi-catalog/models";
 import corpus from "./fixtures/harmony-leak-corpus.json" with { type: "json" };
 import { createAssistantMessage } from "./helpers";

@@ -2,11 +2,18 @@

 ## [Unreleased]

+### Breaking Changes
+
+- The model catalog moved to the new `@oh-my-pi/pi-catalog` package. Deep subpath exports `@oh-my-pi/pi-ai/models.json`, `/models`, `/model-cache`, `/model-manager`, `/model-thinking`, `/effort`, `/provider-models*`, `/utils/discovery*`, `/providers/openai-codex/constants`, `/providers/google-gemini-headers`, and `/providers/openai-completions-compat` are gone — import the `@oh-my-pi/pi-catalog` equivalents (`/models.json`, `/models`, `/model-cache`, `/model-manager`, `/model-thinking`, `/effort`, `/provider-models*`, `/discovery*`, `/wire/codex`, `/wire/gemini-headers`, `/compat/openai`). The pi-ai root barrel re-exports only the model/effort *types* its own signatures use (`Model`, `Api`, `ThinkingConfig`, `Effort`, `Usage`, compat interfaces) — catalog *values* (`getBundledModel(s)`, `calculateCost`, `modelsAreEqual`, `clampThinkingLevelForModel`, `DEFAULT_MODEL_PER_PROVIDER`, …) must be imported from `@oh-my-pi/pi-catalog`.
+- `ProviderDefinition` is now auth-only: `defaultModel`, `createModelManagerOptions`, `catalogDiscovery`, `dynamicModelsAuthoritative`, `allowUnauthenticated`, and `specialModelManager` moved to pi-catalog's `CATALOG_PROVIDERS` table, and `KnownProviderId` was replaced by pi-catalog's `KnownProvider` (registry completeness is enforced by a compile-time check against that union). The pure GitHub Copilot key/endpoint helpers moved from `registry/oauth/github-copilot` to `@oh-my-pi/pi-catalog/wire/github-copilot`.
+
 ### Changed

 - Reduced idle-watchdog churn on the token hot path: the abort promise/listener is created once per stream instead of per yielded item, the deadline uses a persistent re-armed timer instead of a `setTimeout` create/destroy pair per delta, and the persistent race promises are re-minted every 1024 items so per-race reaction records cannot accumulate for the stream's whole life.
 - Memoized Anthropic many-image downscaling by content-block identity, so long sessions with stable message objects no longer re-decode and re-encode every oversized image on each request and retry.
 - Tool-argument validation errors now truncate embedded argument strings at 256 chars per field — a failed `write`-class call no longer echoes hundreds of KB of payload back to the model as the error message.
+- Auth storage no longer issues per-boot no-op writes: the schema-version row is only rewritten when the recorded version actually changes, and the credential identity-key backfill skips rows whose derived identity is null — reopening a current-schema database now performs zero write transactions
+- Plain provider env-var names moved to the catalog table: registry defs dropped their 48 `envKeys` literals (including the pure `$pickenv` pickers for `huggingface`/`qwen-portal`/`xai-oauth`), `getEnvApiKey` now derives those fallbacks from `CATALOG_PROVIDERS[].envVars`, and `envKeys` remains only for computed resolvers (Anthropic Foundry, Vertex ADC, Bedrock credential chains) and non-catalog providers (`kagi`, `tavily`, `parallel`, `perplexity`)

 ### Fixed

@@ -34,11 +34,11 @@
 		"lint": "biome lint .",
 		"test": "bun test --parallel",
 		"fix": "biome check --write --unsafe .",
-		"fmt": "biome format --write .",
-		"generate-models": "bun scripts/generate-models.ts"
+		"fmt": "biome format --write ."
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "catalog:",
+		"@oh-my-pi/pi-catalog": "catalog:",
 		"@oh-my-pi/pi-utils": "catalog:",
 		"openai": "catalog:",
 		"partial-json": "catalog:",
@@ -80,26 +80,10 @@
 			"types": "./src/auth-gateway/*.ts",
 			"import": "./src/auth-gateway/*.ts"
 		},
-		"./models.json": {
-			"types": "./src/models.json.d.ts",
-			"import": "./src/models.json"
-		},
-		"./provider-models": {
-			"types": "./src/provider-models/index.ts",
-			"import": "./src/provider-models/index.ts"
-		},
-		"./provider-models/*": {
-			"types": "./src/provider-models/*.ts",
-			"import": "./src/provider-models/*.ts"
-		},
 		"./providers/*": {
 			"types": "./src/providers/*.ts",
 			"import": "./src/providers/*.ts"
 		},
-		"./providers/cursor/gen/*": {
-			"types": "./src/providers/cursor/gen/*.ts",
-			"import": "./src/providers/cursor/gen/*.ts"
-		},
 		"./providers/openai-codex/*": {
 			"types": "./src/providers/openai-codex/*.ts",
 			"import": "./src/providers/openai-codex/*.ts"
@@ -112,14 +96,6 @@
 			"types": "./src/utils/*.ts",
 			"import": "./src/utils/*.ts"
 		},
-		"./utils/discovery": {
-			"types": "./src/utils/discovery/index.ts",
-			"import": "./src/utils/discovery/index.ts"
-		},
-		"./utils/discovery/*": {
-			"types": "./src/utils/discovery/*.ts",
-			"import": "./src/utils/discovery/*.ts"
-		},
 		"./oauth": {
 			"types": "./src/registry/oauth/index.ts",
 			"import": "./src/registry/oauth/index.ts"
@@ -74,7 +74,7 @@ const PASSTHROUGH_HEADER_NAMES: Record<string, true> = {
 	"openai-organization": true,
 	"openai-project": true,
 	"openai-beta": true,
-	// Codex / ChatGPT-OAuth backend headers (see openai-codex/constants.ts).
+	// Codex / ChatGPT-OAuth backend headers (see @oh-my-pi/pi-catalog/wire/codex).
 	// `session_id` and `conversation_id` thread the upstream session so prompt
 	// caching and per-conversation rate limiting work; `chatgpt-account-id` and
 	// `originator` identify the calling account and client surface.
@@ -17,10 +17,11 @@
 *   POST /v1/messages                      → Anthropic messages in/out
 *   POST /v1/responses                     → OpenAI Responses in/out
 */
+
+import { Effort } from "@oh-my-pi/pi-catalog/effort";
 import { extractRetryHint, logger } from "@oh-my-pi/pi-utils";
 import type { ApiKeyResolver } from "../auth-retry";
 import type { AuthStorage } from "../auth-storage";
-import { Effort } from "../effort";
 import * as anthropicMessages from "../providers/anthropic-messages-server";
 import * as openaiChat from "../providers/openai-chat-server";
 import * as openaiResponses from "../providers/openai-responses-server";
@@ -1,4 +1,4 @@
-import type { Effort } from "../effort";
+import type { Effort } from "@oh-my-pi/pi-catalog/effort";
 import type {
 	AssistantMessage,
 	AssistantMessageEventStream,
@@ -5,13 +5,7 @@ export { type AuthGatewayBootOptions, type ModelResolver, startAuthGateway } fro
 export * from "./auth-gateway/types";
 export * from "./auth-retry";
 export * from "./auth-storage";
-export * from "./effort";
-export * from "./model-cache";
-export * from "./model-manager";
-export * from "./model-thinking";
-export * from "./models";
 export * from "./provider-details";
-export * from "./provider-models";
 export * from "./providers/anthropic";
 export * from "./providers/anthropic-client";
 export * from "./providers/azure-openai-responses";
@@ -19,7 +13,6 @@ export type * from "./providers/cursor";
 export * from "./providers/gitlab-duo";
 export type * from "./providers/google";
 export type * from "./providers/google-gemini-cli";
-export * from "./providers/google-gemini-headers";
 export type * from "./providers/google-vertex";
 export * from "./providers/kimi";
 export * from "./providers/mock";
@@ -42,7 +35,6 @@ export * from "./usage/minimax-code";
 export * from "./usage/openai-codex";
 export * from "./usage/zai";
 export * from "./utils/anthropic-auth";
-export * from "./utils/discovery";
 export * from "./utils/event-stream";
 export * from "./utils/overflow";
 export * from "./utils/retry";
@@ -1,43 +0,0 @@
-/**
- * Provider descriptors and the default-model map, derived from the single-source
- * provider registry (`../registry`).
- *
- * The descriptor/catalog types and guards now live in the registry; they are
- * re-exported here for back-compat with `generate-models.ts` and existing
- * `@oh-my-pi/pi-ai/provider-models` consumers.
- */
-import { PROVIDER_REGISTRY } from "../registry";
-import type { ProviderDescriptor } from "../registry/types";
-import type { KnownProvider } from "../types";
-
-export * from "../registry/types";
-
-/**
- * Runtime model-discovery descriptors: every registry provider that exposes a
- * standard model-manager factory. Special-managed providers
- * (`google-antigravity`/`google-gemini-cli`/`openai-codex`) are built bespoke in
- * the coding-agent runtime and are excluded here.
- */
-export const PROVIDER_DESCRIPTORS: readonly ProviderDescriptor[] = PROVIDER_REGISTRY.flatMap(provider => {
-	const { createModelManagerOptions } = provider;
-	if (!createModelManagerOptions || provider.specialModelManager) {
-		return [];
-	}
-	return [
-		{
-			providerId: provider.id,
-			defaultModel: provider.defaultModel ?? "",
-			createModelManagerOptions,
-			allowUnauthenticated: provider.allowUnauthenticated,
-			dynamicModelsAuthoritative: provider.dynamicModelsAuthoritative,
-			catalogDiscovery: provider.catalogDiscovery,
-		},
-	];
-});
-
-/** Default model IDs for all known providers, derived from the registry. */
-export const DEFAULT_MODEL_PER_PROVIDER: Record<KnownProvider, string> = Object.fromEntries(
-	PROVIDER_REGISTRY.filter(provider => provider.defaultModel != null).map(
-		provider => [provider.id, provider.defaultModel] as [string, string],
-	),
-) as Record<KnownProvider, string>;
@@ -7,10 +7,10 @@
 * Bun's native `HTTPS_PROXY` support.
 */

+import type { Effort } from "@oh-my-pi/pi-catalog/effort";
+import { mapEffortToAnthropicAdaptiveEffort, requireSupportedEffort } from "@oh-my-pi/pi-catalog/model-thinking";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
 import { $env, $flag, extractHttpStatusFromError, fetchWithRetry } from "@oh-my-pi/pi-utils";
-import type { Effort } from "../effort";
-import { mapEffortToAnthropicAdaptiveEffort, requireSupportedEffort } from "../model-thinking";
-import { calculateCost } from "../models";
 import type {
 	Api,
 	AssistantMessage,
@@ -2,6 +2,15 @@ import * as nodeCrypto from "node:crypto";
 import * as fs from "node:fs";
 import { scheduler } from "node:timers/promises";
 import * as tls from "node:tls";
+import {
+	hasOpus47ApiRestrictions,
+	isAnthropicFableOrMythosModel,
+	mapEffortToAnthropicAdaptiveEffort,
+	supportsMidConversationSystemMessages,
+} from "@oh-my-pi/pi-catalog/model-thinking";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
+import { isAnthropicOAuthToken } from "@oh-my-pi/pi-catalog/utils";
+import { parseGitHubCopilotApiKey } from "@oh-my-pi/pi-catalog/wire/github-copilot";
 import {
 	$env,
 	extractHttpStatusFromError,
@@ -12,15 +21,7 @@ import {
 	logger,
 	readSseEvents,
 } from "@oh-my-pi/pi-utils";
-import {
-	hasOpus47ApiRestrictions,
-	isAnthropicFableOrMythosModel,
-	mapEffortToAnthropicAdaptiveEffort,
-	supportsMidConversationSystemMessages,
-} from "../model-thinking";
-import { calculateCost } from "../models";
 import { isUsageLimitError } from "../rate-limit-utils";
-import { parseGitHubCopilotApiKey } from "../registry/oauth/github-copilot";
 import { getEnvApiKey, OUTPUT_FALLBACK_BUFFER } from "../stream";
 import type {
 	Api,
@@ -47,13 +48,7 @@ import type {
 	Usage,
 } from "../types";
 import { resolveServiceTier } from "../types";
-import {
-	isAnthropicOAuthToken,
-	isRecord,
-	normalizeSystemPrompts,
-	normalizeToolCallId,
-	resolveCacheRetention,
-} from "../utils";
+import { isRecord, normalizeSystemPrompts, normalizeToolCallId, resolveCacheRetention } from "../utils";
 import { createAbortSourceTracker } from "../utils/abort";
 import { AssistantMessageEventStream } from "../utils/event-stream";
 import { isFoundryEnabled } from "../utils/foundry";
@@ -3,35 +3,7 @@ import * as fs from "node:fs/promises";
 import http2 from "node:http2";
 import { create, fromBinary, fromJson, type JsonValue, toBinary, toJson } from "@bufbuild/protobuf";
 import { ValueSchema } from "@bufbuild/protobuf/wkt";
-import { $env, extractHttpStatusFromError, sanitizeText } from "@oh-my-pi/pi-utils";
-import { calculateCost } from "../models";
-import type {
-	Api,
-	AssistantMessage,
-	Context,
-	CursorExecHandlerResult,
-	CursorExecHandlers,
-	CursorMcpCall,
-	CursorShellStreamCallbacks,
-	CursorToolResultHandler,
-	ImageContent,
-	Message,
-	Model,
-	StreamFunction,
-	StreamOptions,
-	TextContent,
-	ThinkingContent,
-	Tool,
-	ToolCall,
-	ToolResultMessage,
-} from "../types";
-import { normalizeSystemPrompts } from "../utils";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { createRequestDebugSession, isRequestDebugEnabled, type RequestDebugResponseLog } from "../utils/request-debug";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { toolWireSchema } from "../utils/schema/wire";
-import type { McpToolDefinition } from "./cursor/gen/agent_pb";
+import type { McpToolDefinition } from "@oh-my-pi/pi-catalog/discovery/cursor-gen/agent_pb";
 import {
 	AgentClientMessageSchema,
 	AgentConversationTurnStructureSchema,
@@ -128,7 +100,35 @@ import {
 	WriteShellStdinErrorSchema,
 	WriteShellStdinResultSchema,
 	WriteSuccessSchema,
-} from "./cursor/gen/agent_pb";
+} from "@oh-my-pi/pi-catalog/discovery/cursor-gen/agent_pb";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
+import { $env, extractHttpStatusFromError, sanitizeText } from "@oh-my-pi/pi-utils";
+import type {
+	Api,
+	AssistantMessage,
+	Context,
+	CursorExecHandlerResult,
+	CursorExecHandlers,
+	CursorMcpCall,
+	CursorShellStreamCallbacks,
+	CursorToolResultHandler,
+	ImageContent,
+	Message,
+	Model,
+	StreamFunction,
+	StreamOptions,
+	TextContent,
+	ThinkingContent,
+	Tool,
+	ToolCall,
+	ToolResultMessage,
+} from "../types";
+import { normalizeSystemPrompts } from "../utils";
+import { AssistantMessageEventStream } from "../utils/event-stream";
+import { parseStreamingJson } from "../utils/json-parse";
+import { createRequestDebugSession, isRequestDebugEnabled, type RequestDebugResponseLog } from "../utils/request-debug";
+import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
+import { toolWireSchema } from "../utils/schema/wire";

 export const CURSOR_API_URL = "https://api2.cursor.sh";
 export const CURSOR_CLIENT_VERSION = "cli-2026.01.09-231024f";
@@ -1,4 +1,4 @@
-import { getGitHubCopilotBaseUrl, parseGitHubCopilotApiKey } from "../registry/oauth/github-copilot";
+import { getGitHubCopilotBaseUrl, parseGitHubCopilotApiKey } from "@oh-my-pi/pi-catalog/wire/github-copilot";
 import type { Message } from "../types";
 /**
 * Infer whether the current request to Copilot is user-initiated or agent-initiated.
@@ -5,8 +5,13 @@
 */
 import { createHash, randomBytes, randomUUID } from "node:crypto";
 import { scheduler } from "node:timers/promises";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
+import {
+	ANTIGRAVITY_SYSTEM_INSTRUCTION,
+	getAntigravityUserAgent,
+	getGeminiCliHeaders,
+} from "@oh-my-pi/pi-catalog/wire/gemini-headers";
 import { extractHttpStatusFromError, fetchWithRetry, readSseJson } from "@oh-my-pi/pi-utils";
-import { calculateCost } from "../models";
 import type {
 	Api,
 	AssistantMessage,
@@ -24,7 +29,6 @@ import { appendRawHttpRequestDumpFor400, type RawHttpRequestDump, withHttpStatus
 // Refresh is the sole responsibility of AuthStorage (broker-aware, single-flighted);
 // the stream provider trusts the access token threaded through `options.apiKey`.
 import { normalizeSchemaForCCA } from "../utils/schema";
-import { ANTIGRAVITY_SYSTEM_INSTRUCTION, getAntigravityUserAgent, getGeminiCliHeaders } from "./google-gemini-headers";
 import type { Content, FunctionCallingConfigMode, ThinkingConfig } from "./google-shared";
 import {
 	convertMessages,
@@ -80,7 +84,7 @@ export {
 	getAntigravityUserAgent,
 	getGeminiCliHeaders,
 	getGeminiCliUserAgent,
-} from "./google-gemini-headers";
+} from "@oh-my-pi/pi-catalog/wire/gemini-headers";

 // Retry configuration
 const MAX_RETRIES = 3;
@@ -2,8 +2,8 @@
 * Shared utilities for Google Generative AI and Google Cloud Code Assist providers.
 */

+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
 import { extractHttpStatusFromError, readSseJson } from "@oh-my-pi/pi-utils";
-import { calculateCost } from "../models";
 import type {
 	Api,
 	AssistantMessage,
@@ -1,5 +1,12 @@
 import * as os from "node:os";
 import { scheduler } from "node:timers/promises";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
+import {
+	CODEX_BASE_URL,
+	getCodexAccountId,
+	OPENAI_HEADER_VALUES,
+	OPENAI_HEADERS,
+} from "@oh-my-pi/pi-catalog/wire/codex";
 import {
 	$env,
 	$flag,
@@ -20,7 +27,6 @@ import type {
 	ResponseReasoningItem,
 } from "openai/resources/responses/responses";
 import packageJson from "../../package.json" with { type: "json" };
-import { calculateCost } from "../models";
 import { getEnvApiKey } from "../stream";
 import {
 	type Api,
@@ -58,7 +64,6 @@ import { createRequestDebugSession, isRequestDebugEnabled, type RequestDebugResp
 import { adaptSchemaForStrict, NO_STRICT, sanitizeSchemaForOpenAIResponses, toolWireSchema } from "../utils/schema";
 import { notifyRawSseEvent } from "../utils/sse-debug";
 import { compactGrammarDefinition } from "./grammar";
-import { CODEX_BASE_URL, getCodexAccountId, OPENAI_HEADER_VALUES, OPENAI_HEADERS } from "./openai-codex/constants";
 import {
 	type CodexRequestOptions,
 	type InputItem,
@@ -1,5 +1,5 @@
-import type { Effort } from "../../effort";
-import { requireSupportedEffort } from "../../model-thinking";
+import type { Effort } from "@oh-my-pi/pi-catalog/effort";
+import { requireSupportedEffort } from "@oh-my-pi/pi-catalog/model-thinking";
 import type { Api, Model } from "../../types";

 export interface ReasoningConfig {
@@ -1,4 +1,4 @@
-import { toNumber } from "../../utils";
+import { toNumber } from "@oh-my-pi/pi-catalog/utils";

 export type CodexRateLimit = {
 	used_percent?: number;
@@ -1,3 +1,9 @@
+import { detectOpenAICompat, type ResolvedOpenAICompat, resolveOpenAICompat } from "@oh-my-pi/pi-catalog/compat/openai";
+import type { Effort } from "@oh-my-pi/pi-catalog/effort";
+import { toFirepassWireModelId, toFireworksWireModelId } from "@oh-my-pi/pi-catalog/fireworks-model-id";
+import { getSupportedEfforts } from "@oh-my-pi/pi-catalog/model-thinking";
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
+import { parseGitHubCopilotApiKey } from "@oh-my-pi/pi-catalog/wire/github-copilot";
 import { $env, extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
 import OpenAI, { APIConnectionTimeoutError as OpenAIConnectionTimeoutError } from "openai";
 import type {
@@ -10,10 +16,6 @@ import type {
 	ChatCompletionToolMessageParam,
 } from "openai/resources/chat/completions";
 import packageJson from "../../package.json" with { type: "json" };
-import type { Effort } from "../effort";
-import { getSupportedEfforts } from "../model-thinking";
-import { calculateCost } from "../models";
-import { parseGitHubCopilotApiKey } from "../registry/oauth/github-copilot";
 import { getKimiCommonHeaders } from "../registry/oauth/kimi";
 import { getEnvApiKey } from "../stream";
 import {
@@ -43,7 +45,6 @@ import {
 import { normalizeSystemPrompts } from "../utils";
 import { createAbortSourceTracker } from "../utils/abort";
 import { AssistantMessageEventStream } from "../utils/event-stream";
-import { toFirepassWireModelId, toFireworksWireModelId } from "../utils/fireworks-model-id";
 import {
 	type CapturedHttpErrorResponse,
 	finalizeErrorMessage,
@@ -73,7 +74,6 @@ import {
 	hasCopilotVisionInput,
 	resolveGitHubCopilotBaseUrl,
 } from "./github-copilot-headers";
-import { detectOpenAICompat, type ResolvedOpenAICompat, resolveOpenAICompat } from "./openai-completions-compat";
 import { createInitialResponsesAssistantMessage } from "./openai-responses-shared";
 import { transformMessages } from "./transform-messages";
 import {
@@ -1,3 +1,4 @@
+import { calculateCost } from "@oh-my-pi/pi-catalog/models";
 import { logger, structuredCloneJSON } from "@oh-my-pi/pi-utils";
 import type OpenAI from "openai";
 import type {
@@ -11,7 +12,6 @@ import type {
 	ResponseOutputMessage,
 	ResponseReasoningItem,
 } from "openai/resources/responses/responses";
-import { calculateCost } from "../models";
 import {
 	type Api,
 	type AssistantMessage,
@@ -1,3 +1,4 @@
+import { parseGitHubCopilotApiKey } from "@oh-my-pi/pi-catalog/wire/github-copilot";
 import { $env, extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
 import OpenAI, { APIConnectionTimeoutError as OpenAIConnectionTimeoutError } from "openai";
 import type {
@@ -6,7 +7,6 @@ import type {
 	ResponseInput,
 	ResponseStreamEvent,
 } from "openai/resources/responses/responses";
-import { parseGitHubCopilotApiKey } from "../registry/oauth/github-copilot";
 import { getEnvApiKey } from "../stream";
 import type {
 	AssistantMessage,
@@ -1,12 +1,6 @@
-import { aimlApiModelManagerOptions } from "../provider-models/openai-compat";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const aimlApiProvider = {
 	id: "aimlapi",
 	name: "AIML API",
-	defaultModel: "gpt-4o",
-	createModelManagerOptions: (config: ModelManagerConfig) => aimlApiModelManagerOptions(config),
-	dynamicModelsAuthoritative: true,
-	catalogDiscovery: { label: "AIML API", envVars: ["AIMLAPI_API_KEY"] },
-	envKeys: "AIMLAPI_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { alibabaCodingPlanModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://modelstudio.console.alibabacloud.com/";
 const API_BASE_URL = "https://coding-intl.dashscope.aliyuncs.com/v1";
@@ -46,9 +45,5 @@ export async function loginAlibabaCodingPlan(options: OAuthController): Promise<
 export const alibabaCodingPlanProvider = {
 	id: "alibaba-coding-plan",
 	name: "Alibaba Coding Plan",
-	defaultModel: "qwen3.5-plus",
-	createModelManagerOptions: (config: ModelManagerConfig) => alibabaCodingPlanModelManagerOptions(config),
-	catalogDiscovery: { label: "Alibaba Coding Plan", envVars: ["ALIBABA_CODING_PLAN_API_KEY"] },
-	envKeys: "ALIBABA_CODING_PLAN_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginAlibabaCodingPlan(cb),
 } as const satisfies ProviderDefinition;
@@ -4,7 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const amazonBedrockProvider = {
 	id: "amazon-bedrock",
 	name: "Amazon Bedrock",
-	defaultModel: "us.anthropic.claude-opus-4-6-v1",
 	// Amazon Bedrock accepts bearer tokens, IAM keys, profiles, ECS/IRSA credential chains.
 	envKeys: () => {
 		const hasEcsCredentials =
@@ -1,14 +1,11 @@
 import { $pickenv } from "@oh-my-pi/pi-utils";
-import { anthropicModelManagerOptions } from "../provider-models/openai-compat";
 import { isFoundryEnabled } from "../utils/foundry";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const anthropicProvider = {
 	id: "anthropic",
 	name: "Anthropic (Claude Pro/Max)",
-	defaultModel: "claude-opus-4-6",
-	createModelManagerOptions: (config: ModelManagerConfig) => anthropicModelManagerOptions(config),
 	// Foundry mode optionally switches Anthropic auth to enterprise gateway credentials.
 	envKeys: () =>
 		isFoundryEnabled()
@@ -1,7 +1,6 @@
-import { cerebrasModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginCerebras = createApiKeyLogin({
 	providerLabel: "Cerebras",
@@ -20,9 +19,5 @@ export const loginCerebras = createApiKeyLogin({
 export const cerebrasProvider = {
 	id: "cerebras",
 	name: "Cerebras",
-	defaultModel: "zai-glm-4.6",
-	createModelManagerOptions: (config: ModelManagerConfig) => cerebrasModelManagerOptions(config),
-	catalogDiscovery: { label: "Cerebras", envVars: ["CEREBRAS_API_KEY"] },
-	envKeys: "CEREBRAS_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginCerebras(cb),
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { cloudflareAiGatewayModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://developers.cloudflare.com/ai-gateway/configuration/authentication/";

@@ -41,9 +40,5 @@ export async function loginCloudflareAiGateway(options: OAuthController): Promis
 export const cloudflareAiGatewayProvider = {
 	id: "cloudflare-ai-gateway",
 	name: "Cloudflare AI Gateway",
-	defaultModel: "claude-sonnet-4-5",
-	createModelManagerOptions: (config: ModelManagerConfig) => cloudflareAiGatewayModelManagerOptions(config),
-	catalogDiscovery: { label: "Cloudflare AI Gateway", envVars: ["CLOUDFLARE_AI_GATEWAY_API_KEY"] },
-	envKeys: "CLOUDFLARE_AI_GATEWAY_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginCloudflareAiGateway(cb),
 } as const satisfies ProviderDefinition;
@@ -1,14 +1,9 @@
-import { cursorModelManagerOptions } from "../provider-models/special";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const cursorProvider = {
 	id: "cursor",
 	name: "Cursor (Claude, GPT, etc.)",
-	defaultModel: "claude-sonnet-4-6",
-	createModelManagerOptions: (config: ModelManagerConfig) => cursorModelManagerOptions(config),
-	catalogDiscovery: { label: "Cursor", envVars: ["CURSOR_API_KEY"], oauthProvider: "cursor" },
-	envKeys: "CURSOR_ACCESS_TOKEN",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginCursor } = await import("./oauth/cursor");
@@ -1,7 +1,6 @@
-import { deepseekModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthController, OAuthLoginCallbacks, OAuthPrompt } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const innerLogin = createApiKeyLogin({
 	providerLabel: "DeepSeek",
@@ -42,9 +41,5 @@ export const loginDeepSeek = async (options: OAuthController): Promise<string> =
 export const deepseekProvider = {
 	id: "deepseek",
 	name: "DeepSeek",
-	defaultModel: "deepseek-v4-pro",
-	createModelManagerOptions: (config: ModelManagerConfig) => deepseekModelManagerOptions(config),
-	catalogDiscovery: { label: "DeepSeek", envVars: ["DEEPSEEK_API_KEY"] },
-	envKeys: "DEEPSEEK_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginDeepSeek(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { firepassModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 /**
 * Fire Pass login flow.
@@ -29,8 +28,5 @@ export const loginFirepass = createApiKeyLogin({
 export const firepassProvider = {
 	id: "firepass",
 	name: "Fire Pass (Fireworks Kimi K2.6 Turbo subscription)",
-	defaultModel: "kimi-k2.6-turbo",
-	createModelManagerOptions: (config: ModelManagerConfig) => firepassModelManagerOptions(config),
-	envKeys: "FIREPASS_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginFirepass(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { fireworksModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginFireworks = createApiKeyLogin({
 	providerLabel: "Fireworks",
@@ -19,9 +18,5 @@ export const loginFireworks = createApiKeyLogin({
 export const fireworksProvider = {
 	id: "fireworks",
 	name: "Fireworks",
-	defaultModel: "kimi-k2.6",
-	createModelManagerOptions: (config: ModelManagerConfig) => fireworksModelManagerOptions(config),
-	catalogDiscovery: { label: "Fireworks", envVars: ["FIREWORKS_API_KEY"] },
-	envKeys: "FIREWORKS_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginFireworks(cb),
 } as const satisfies ProviderDefinition;
@@ -1,13 +1,9 @@
-import { githubCopilotModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const githubCopilotProvider = {
 	id: "github-copilot",
 	name: "GitHub Copilot",
-	defaultModel: "gpt-4o",
-	createModelManagerOptions: (config: ModelManagerConfig) => githubCopilotModelManagerOptions(config),
-	envKeys: "COPILOT_GITHUB_TOKEN",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginGitHubCopilot } = await import("./oauth/github-copilot");
@@ -4,8 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const gitlabDuoProvider = {
 	id: "gitlab-duo",
 	name: "GitLab Duo",
-	defaultModel: "duo-chat-sonnet-4-5",
-	envKeys: "GITLAB_TOKEN",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginGitLabDuo } = await import("./oauth/gitlab-duo");
@@ -4,8 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const googleAntigravityProvider = {
 	id: "google-antigravity",
 	name: "Antigravity (Gemini 3, Claude, GPT-OSS)",
-	defaultModel: "gemini-3-pro-high",
-	specialModelManager: true,
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginAntigravity } = await import("./oauth/google-antigravity");
@@ -4,8 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const googleGeminiCliProvider = {
 	id: "google-gemini-cli",
 	name: "Google Cloud Code Assist (Gemini CLI)",
-	defaultModel: "gemini-2.5-pro",
-	specialModelManager: true,
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginGeminiCli } = await import("./oauth/google-gemini-cli");
@@ -2,8 +2,7 @@ import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
 import { $env } from "@oh-my-pi/pi-utils";
-import { googleVertexModelManagerOptions } from "../provider-models/google";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 let cachedVertexAdcCredentialsExists: boolean | null = null;

@@ -24,9 +23,6 @@ function hasVertexAdcCredentials(): boolean {
 export const googleVertexProvider = {
 	id: "google-vertex",
 	name: "Google Vertex AI",
-	defaultModel: "gemini-3-pro-preview",
-	createModelManagerOptions: (config: ModelManagerConfig) => googleVertexModelManagerOptions(config),
-	allowUnauthenticated: true,
 	// Vertex AI supports either GOOGLE_CLOUD_API_KEY or Application Default Credentials.
 	envKeys: () => {
 		if ($env.GOOGLE_CLOUD_API_KEY) {
@@ -1,10 +1,6 @@
-import { googleModelManagerOptions } from "../provider-models/google";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const googleProvider = {
 	id: "google",
 	name: "Google Gemini",
-	defaultModel: "gemini-2.5-pro",
-	createModelManagerOptions: (config: ModelManagerConfig) => googleModelManagerOptions(config),
-	envKeys: "GEMINI_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,10 +1,6 @@
-import { groqModelManagerOptions } from "../provider-models/openai-compat";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const groqProvider = {
 	id: "groq",
 	name: "Groq",
-	defaultModel: "openai/gpt-oss-120b",
-	createModelManagerOptions: (config: ModelManagerConfig) => groqModelManagerOptions(config),
-	envKeys: "GROQ_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,8 +1,6 @@
-import { $pickenv } from "@oh-my-pi/pi-utils";
-import { huggingfaceModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL =
 	"https://huggingface.co/settings/tokens/new?ownUserPermissions=inference.serverless.write&tokenType=fineGrained";
@@ -49,9 +47,5 @@ export async function loginHuggingface(options: OAuthController): Promise<string
 export const huggingfaceProvider = {
 	id: "huggingface",
 	name: "Hugging Face Inference",
-	defaultModel: "deepseek-ai/DeepSeek-R1",
-	createModelManagerOptions: (config: ModelManagerConfig) => huggingfaceModelManagerOptions(config),
-	catalogDiscovery: { label: "Hugging Face", envVars: ["HUGGINGFACE_HUB_TOKEN", "HF_TOKEN"] },
-	envKeys: () => $pickenv("HUGGINGFACE_HUB_TOKEN", "HF_TOKEN"),
 	login: (cb: OAuthLoginCallbacks) => loginHuggingface(cb),
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { kiloModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController, OAuthCredentials } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const KILO_DEVICE_AUTH_BASE_URL = "https://api.kilo.ai/api/device-auth";
 const POLL_INTERVAL_MS = 5000;
@@ -89,9 +88,5 @@ export async function loginKilo(callbacks: OAuthController): Promise<OAuthCreden
 export const kiloProvider = {
 	id: "kilo",
 	name: "Kilo Gateway",
-	defaultModel: "anthropic/claude-sonnet-4.5",
-	createModelManagerOptions: (config: ModelManagerConfig) => kiloModelManagerOptions(config),
-	catalogDiscovery: { label: "Kilo Gateway", envVars: ["KILO_API_KEY"], allowUnauthenticated: true },
-	envKeys: "KILO_API_KEY",
 	login: loginKilo,
 } as const satisfies ProviderDefinition;
@@ -1,13 +1,9 @@
-import { kimiCodeModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const kimiCodeProvider = {
 	id: "kimi-code",
 	name: "Kimi Code",
-	defaultModel: "kimi-k2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) => kimiCodeModelManagerOptions(config),
-	catalogDiscovery: { label: "Kimi Code", envVars: ["KIMI_API_KEY"] },
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginKimi } = await import("./oauth/kimi");
@@ -1,6 +1,5 @@
-import { litellmModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://docs.litellm.ai/docs/proxy/deploy";

@@ -40,9 +39,5 @@ export async function loginLiteLLM(options: OAuthController): Promise<string> {
 export const litellmProvider = {
 	id: "litellm",
 	name: "LiteLLM",
-	defaultModel: "claude-opus-4-6",
-	createModelManagerOptions: (config: ModelManagerConfig) => litellmModelManagerOptions(config),
-	catalogDiscovery: { label: "LiteLLM", envVars: ["LITELLM_API_KEY"], allowUnauthenticated: true },
-	envKeys: "LITELLM_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginLiteLLM(cb),
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { lmStudioModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const PROVIDER_ID = "lm-studio";
 export const DEFAULT_LOCAL_TOKEN = "lm-studio-local";
@@ -27,9 +26,5 @@ export async function loginLmStudio(options: OAuthController): Promise<string> {
 export const lmStudioProvider = {
 	id: "lm-studio",
 	name: "LM Studio (Local OpenAI-compatible)",
-	defaultModel: "llama-3-8b",
-	createModelManagerOptions: (config: ModelManagerConfig) => lmStudioModelManagerOptions(config),
-	allowUnauthenticated: true,
-	envKeys: "LM_STUDIO_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginLmStudio(cb),
 } as const satisfies ProviderDefinition;
@@ -4,8 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const minimaxCodeCnProvider = {
 	id: "minimax-code-cn",
 	name: "MiniMax Coding Plan (China)",
-	defaultModel: "MiniMax-M2.5",
-	envKeys: "MINIMAX_CODE_CN_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginMiniMaxCodeCn } = await import("./oauth/minimax-code");
@@ -4,8 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const minimaxCodeProvider = {
 	id: "minimax-code",
 	name: "MiniMax Coding Plan (International)",
-	defaultModel: "MiniMax-M2.5",
-	envKeys: "MINIMAX_CODE_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginMiniMaxCode } = await import("./oauth/minimax-code");
@@ -3,6 +3,4 @@ import type { ProviderDefinition } from "./types";
 export const minimaxProvider = {
 	id: "minimax",
 	name: "MiniMax",
-	defaultModel: "MiniMax-M2.5",
-	envKeys: "MINIMAX_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,10 +1,6 @@
-import { mistralModelManagerOptions } from "../provider-models/openai-compat";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const mistralProvider = {
 	id: "mistral",
 	name: "Mistral",
-	defaultModel: "devstral-medium-latest",
-	createModelManagerOptions: (config: ModelManagerConfig) => mistralModelManagerOptions(config),
-	envKeys: "MISTRAL_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { moonshotModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginMoonshot = createApiKeyLogin({
 	providerLabel: "Moonshot",
@@ -19,9 +18,5 @@ export const loginMoonshot = createApiKeyLogin({
 export const moonshotProvider = {
 	id: "moonshot",
 	name: "Moonshot (Kimi API)",
-	defaultModel: "kimi-k2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) => moonshotModelManagerOptions(config),
-	catalogDiscovery: { label: "Moonshot", envVars: ["MOONSHOT_API_KEY"] },
-	envKeys: "MOONSHOT_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginMoonshot(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { nanoGptModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginNanoGPT = createApiKeyLogin({
 	providerLabel: "NanoGPT",
@@ -19,9 +18,5 @@ export const loginNanoGPT = createApiKeyLogin({
 export const nanogptProvider = {
 	id: "nanogpt",
 	name: "NanoGPT",
-	defaultModel: "openai/gpt-5.4",
-	createModelManagerOptions: (config: ModelManagerConfig) => nanoGptModelManagerOptions(config),
-	catalogDiscovery: { label: "NanoGPT", envVars: ["NANO_GPT_API_KEY"] },
-	envKeys: "NANO_GPT_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginNanoGPT(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { nvidiaModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://org.ngc.nvidia.com/setup/personal-keys";
 const API_BASE_URL = "https://integrate.api.nvidia.com/v1";
@@ -57,9 +56,5 @@ export async function loginNvidia(options: OAuthController): Promise<string> {
 export const nvidiaProvider = {
 	id: "nvidia",
 	name: "NVIDIA",
-	defaultModel: "nvidia/llama-3.1-nemotron-70b-instruct",
-	createModelManagerOptions: (config: ModelManagerConfig) => nvidiaModelManagerOptions(config),
-	catalogDiscovery: { label: "NVIDIA", envVars: ["NVIDIA_API_KEY"] },
-	envKeys: "NVIDIA_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginNvidia(cb),
 } as const satisfies ProviderDefinition;
@@ -2,18 +2,19 @@
 * GitHub Copilot OAuth flow (opencode OAuth app)
 */
 import { scheduler } from "node:timers/promises";
-import { getBundledModels } from "../../models";
+import { getBundledModels } from "@oh-my-pi/pi-catalog/models";
+import {
+	getGitHubCopilotBaseUrl,
+	isPublicGitHubHost,
+	normalizeDomain,
+	normalizeGitHubCopilotEnterpriseDomain,
+	OPENCODE_HEADERS,
+} from "@oh-my-pi/pi-catalog/wire/github-copilot";
 import type { FetchImpl } from "../../types";
 import type { OAuthCredentials } from "./types";

 const CLIENT_ID = "Ov23li8tweQw6odWQebz";

-export const COPILOT_USER_AGENT = "opencode/1.3.15" as const;
-
-export const OPENCODE_HEADERS = {
-	"User-Agent": COPILOT_USER_AGENT,
-} as const;
-
 const INITIAL_POLL_INTERVAL_MULTIPLIER = 1.2;
 const SLOW_DOWN_POLL_INTERVAL_MULTIPLIER = 1.4;

@@ -46,58 +47,6 @@ type DeviceTokenErrorResponse = {
 	interval?: number;
 };

-type GitHubCopilotApiKeyPayload = {
-	token?: unknown;
-	enterpriseUrl?: unknown;
-};
-
-export type ParsedGitHubCopilotApiKey = {
-	accessToken: string;
-	enterpriseUrl?: string;
-};
-
-const PUBLIC_GITHUB_HOSTS = new Set(["api.github.com", "github.com", "www.github.com"]);
-
-function isPublicGitHubHost(host: string): boolean {
-	return PUBLIC_GITHUB_HOSTS.has(host.trim().toLowerCase());
-}
-
-export function normalizeGitHubCopilotEnterpriseDomain(input: string | undefined): string | undefined {
-	const trimmed = input?.trim();
-	if (!trimmed) return undefined;
-	const normalized = normalizeDomain(trimmed) ?? trimmed.toLowerCase();
-	if (!normalized || isPublicGitHubHost(normalized)) return undefined;
-	return normalized;
-}
-
-export function parseGitHubCopilotApiKey(apiKeyRaw: string): ParsedGitHubCopilotApiKey {
-	try {
-		const parsed = JSON.parse(apiKeyRaw) as GitHubCopilotApiKeyPayload;
-		if (typeof parsed.token === "string") {
-			return {
-				accessToken: parsed.token,
-				enterpriseUrl:
-					typeof parsed.enterpriseUrl === "string"
-						? normalizeGitHubCopilotEnterpriseDomain(parsed.enterpriseUrl)
-						: undefined,
-			};
-		}
-	} catch {}
-
-	return { accessToken: apiKeyRaw };
-}
-
-export function normalizeDomain(input: string): string | null {
-	const trimmed = input.trim();
-	if (!trimmed) return null;
-	try {
-		const url = trimmed.includes("://") ? new URL(trimmed) : new URL(`https://${trimmed}`);
-		return url.hostname;
-	} catch {
-		return null;
-	}
-}
-
 function getUrls(domain: string): {
 	deviceCodeUrl: string;
 	accessTokenUrl: string;
@@ -108,15 +57,6 @@ function getUrls(domain: string): {
 	};
 }

-export function getGitHubCopilotBaseUrl(enterpriseDomain?: string): string {
-	const normalizedEnterpriseDomain = normalizeGitHubCopilotEnterpriseDomain(enterpriseDomain);
-	if (!normalizedEnterpriseDomain) return "https://api.githubcopilot.com";
-	const host = normalizedEnterpriseDomain.startsWith("copilot-api.")
-		? normalizedEnterpriseDomain
-		: `copilot-api.${normalizedEnterpriseDomain}`;
-	return `https://${host}`;
-}
-
 async function fetchJson(url: string, init: RequestInit, fetchImpl: FetchImpl): Promise<unknown> {
 	const response = await fetchImpl(url, init);
 	if (!response.ok) {
@@ -2,7 +2,7 @@
 * Antigravity OAuth flow (Gemini 3, Claude, GPT-OSS via Google Cloud)
 * Uses different OAuth credentials than google-gemini-cli for access to additional models.
 */
-import { getAntigravityUserAgent } from "../../providers/google-gemini-headers";
+import { getAntigravityUserAgent } from "@oh-my-pi/pi-catalog/wire/gemini-headers";
 import { runGoogleOAuthLogin } from "./google-oauth-shared";
 import type { OAuthController, OAuthCredentials } from "./types";

@@ -3,8 +3,8 @@
 * Standard Gemini models only (gemini-2.0-flash, gemini-2.5-*)
 */

+import { getGeminiCliHeaders } from "@oh-my-pi/pi-catalog/wire/gemini-headers";
 import { $env } from "@oh-my-pi/pi-utils";
-import { getGeminiCliHeaders } from "../../providers/google-gemini-headers";
 import { runGoogleOAuthLogin } from "./google-oauth-shared";
 import type { OAuthController, OAuthCredentials } from "./types";

@@ -1,6 +1,5 @@
-import { ollamaCloudModelManagerOptions } from "../provider-models/ollama";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const OLLAMA_CLOUD_KEYS_URL = "https://ollama.com/settings/keys";

@@ -32,9 +31,5 @@ export async function loginOllamaCloud(options: OAuthController): Promise<string
 export const ollamaCloudProvider = {
 	id: "ollama-cloud",
 	name: "Ollama Cloud",
-	defaultModel: "gpt-oss:120b",
-	createModelManagerOptions: (config: ModelManagerConfig) => ollamaCloudModelManagerOptions(config),
-	catalogDiscovery: { label: "Ollama Cloud", envVars: ["OLLAMA_CLOUD_API_KEY"], oauthProvider: "ollama-cloud" },
-	envKeys: "OLLAMA_CLOUD_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginOllamaCloud(cb),
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { ollamaModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const OLLAMA_DOCS_URL = "https://github.com/ollama/ollama/blob/main/docs/api.md";

@@ -39,9 +38,5 @@ export async function loginOllama(options: OAuthController): Promise<string> {
 export const ollamaProvider = {
 	id: "ollama",
 	name: "Ollama (Local OpenAI-compatible)",
-	defaultModel: "gpt-oss:20b",
-	createModelManagerOptions: (config: ModelManagerConfig) => ollamaModelManagerOptions(config),
-	allowUnauthenticated: true,
 	login: loginOllama,
-	envKeys: "OLLAMA_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -4,9 +4,6 @@ import type { ProviderDefinition } from "./types";
 export const openaiCodexProvider = {
 	id: "openai-codex",
 	name: "ChatGPT Plus/Pro (Codex Subscription)",
-	defaultModel: "gpt-5.4",
-	specialModelManager: true,
-	envKeys: "OPENAI_CODEX_OAUTH_TOKEN",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginOpenAICodex } = await import("./oauth/openai-codex");
@@ -1,10 +1,6 @@
-import { openaiModelManagerOptions } from "../provider-models/openai-compat";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const openaiProvider = {
 	id: "openai",
 	name: "OpenAI",
-	defaultModel: "gpt-5.4",
-	createModelManagerOptions: (config: ModelManagerConfig) => openaiModelManagerOptions(config),
-	envKeys: "OPENAI_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,13 +1,9 @@
-import { opencodeGoModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const opencodeGoProvider = {
 	id: "opencode-go",
 	name: "OpenCode Go",
-	defaultModel: "kimi-k2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) => opencodeGoModelManagerOptions(config),
-	envKeys: "OPENCODE_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginOpenCode } = await import("./oauth/opencode");
@@ -1,13 +1,9 @@
-import { opencodeZenModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const opencodeZenProvider = {
 	id: "opencode-zen",
 	name: "OpenCode Zen",
-	defaultModel: "claude-sonnet-4-6",
-	createModelManagerOptions: (config: ModelManagerConfig) => opencodeZenModelManagerOptions(config),
-	envKeys: "OPENCODE_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginOpenCode } = await import("./oauth/opencode");
@@ -1,7 +1,6 @@
-import { openrouterModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 /** OpenRouter login flow (API key paste, validated via /auth/key).
 *
@@ -25,9 +24,5 @@ export const loginOpenRouter = createApiKeyLogin({
 export const openrouterProvider = {
 	id: "openrouter",
 	name: "OpenRouter",
-	defaultModel: "openai/gpt-5.4",
-	createModelManagerOptions: (config: ModelManagerConfig) => openrouterModelManagerOptions(config),
-	catalogDiscovery: { label: "OpenRouter", envVars: ["OPENROUTER_API_KEY"], allowUnauthenticated: true },
-	envKeys: "OPENROUTER_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginOpenRouter(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { qianfanModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://console.bce.baidu.com/qianfan/ais/console/apiKey";
 const API_BASE_URL = "https://qianfan.baidubce.com/v2";
@@ -46,9 +45,5 @@ export async function loginQianfan(options: OAuthController): Promise<string> {
 export const qianfanProvider = {
 	id: "qianfan",
 	name: "Qianfan",
-	defaultModel: "deepseek-v3.2",
-	createModelManagerOptions: (config: ModelManagerConfig) => qianfanModelManagerOptions(config),
-	catalogDiscovery: { label: "Qianfan", envVars: ["QIANFAN_API_KEY"] },
-	envKeys: "QIANFAN_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginQianfan(cb),
 } as const satisfies ProviderDefinition;
@@ -1,8 +1,6 @@
-import { $pickenv } from "@oh-my-pi/pi-utils";
-import { qwenPortalModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://chat.qwen.ai";
 const API_BASE_URL = "https://portal.qwen.ai/v1";
@@ -47,13 +45,5 @@ export async function loginQwenPortal(options: OAuthController): Promise<string>
 export const qwenPortalProvider = {
 	id: "qwen-portal",
 	name: "Qwen Portal",
-	defaultModel: "coder-model",
-	createModelManagerOptions: (config: ModelManagerConfig) => qwenPortalModelManagerOptions(config),
-	catalogDiscovery: {
-		label: "Qwen Portal",
-		envVars: ["QWEN_OAUTH_TOKEN", "QWEN_PORTAL_API_KEY"],
-		oauthProvider: "qwen-portal",
-	},
-	envKeys: () => $pickenv("QWEN_OAUTH_TOKEN", "QWEN_PORTAL_API_KEY"),
 	login: (cb: OAuthLoginCallbacks) => loginQwenPortal(cb),
 } as const satisfies ProviderDefinition;
@@ -1,3 +1,4 @@
+import type { KnownProvider } from "@oh-my-pi/pi-catalog";
 import { aimlApiProvider } from "./aimlapi";
 import { alibabaCodingPlanProvider } from "./alibaba-coding-plan";
 import { amazonBedrockProvider } from "./amazon-bedrock";
@@ -137,7 +138,12 @@ export function getProviderDefinition(id: string): ProviderDefinition | undefine
 	return BY_ID.get(id);
 }

-/** Chat-model providers (those carrying a `defaultModel`). */
-export type KnownProviderId = Extract<RegistryDef, { defaultModel: string }>["id"];
+/** Compile-time completeness: every catalog chat-model provider must have a registry definition. */
+type _MissingCatalogProviders = Exclude<KnownProvider, RegistryDef["id"]>;
+type _CheckRegistryComplete = _MissingCatalogProviders extends never
+	? true
+	: ["registry is missing catalog providers", _MissingCatalogProviders];
+true satisfies _CheckRegistryComplete;
+
 /** Loginable providers (those carrying a `login` flow). */
 export type OAuthProviderUnion = Extract<RegistryDef, { login: object }>["id"];
@@ -1,6 +1,5 @@
-import { syntheticModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginSynthetic = createApiKeyLogin({
 	providerLabel: "Synthetic",
@@ -18,10 +17,5 @@ export const loginSynthetic = createApiKeyLogin({
 export const syntheticProvider = {
 	id: "synthetic",
 	name: "Synthetic",
-	defaultModel: "hf:zai-org/GLM-5.1",
-	createModelManagerOptions: (config: ModelManagerConfig) => syntheticModelManagerOptions(config),
-	dynamicModelsAuthoritative: true,
-	catalogDiscovery: { label: "Synthetic", envVars: ["SYNTHETIC_API_KEY"] },
-	envKeys: "SYNTHETIC_API_KEY",
 	login: loginSynthetic,
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { togetherModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginTogether = createApiKeyLogin({
 	providerLabel: "Together",
@@ -19,9 +18,5 @@ export const loginTogether = createApiKeyLogin({
 export const togetherProvider = {
 	id: "together",
 	name: "Together",
-	defaultModel: "moonshotai/Kimi-K2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) => togetherModelManagerOptions(config),
-	catalogDiscovery: { label: "Together", envVars: ["TOGETHER_API_KEY"] },
-	envKeys: "TOGETHER_API_KEY",
 	login: (cb: Parameters<typeof loginTogether>[0]) => loginTogether(cb),
 } as const satisfies ProviderDefinition;
@@ -1,20 +1,16 @@
 /**
- * Single-source provider model. Every provider — model providers, gateways,
- * search/tool credentials, and login-only flows — is described by one
- * {@link ProviderDefinition}. The legacy scattered structures (the
- * `KnownProvider`/`OAuthProvider` unions, `PROVIDER_DESCRIPTORS`,
- * `serviceProviderMap`, `builtInOAuthProviders`, the refresh/login switches,
- * and the CLI callback maps) are all *derived* from the registry of these
- * definitions. Adding a provider is one new file in `./providers/` plus one
- * line in `./registry.ts`.
+ * Single-source provider auth model. Every provider — model providers,
+ * gateways, search/tool credentials, and login-only flows — is described by
+ * one {@link ProviderDefinition}. The legacy scattered structures (the
+ * `OAuthProvider` union, `serviceProviderMap`, `builtInOAuthProviders`, the
+ * refresh/login switches, and the CLI callback maps) are all *derived* from
+ * the registry of these definitions. Adding a provider is one new file in
+ * `./providers/` plus one line in `./registry.ts`. Model-catalog metadata
+ * (default model, model-manager factory, catalog discovery) lives in
+ * `@oh-my-pi/pi-catalog`'s descriptor table.
 */
-import type { ModelManagerOptions } from "../model-manager";
-import type { Api, FetchImpl } from "../types";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "./oauth/types";

-/** Config passed to a provider's runtime model-manager factory. */
-export type ModelManagerConfig = { apiKey?: string; baseUrl?: string; fetch?: FetchImpl };
-
 /**
 * API-key environment fallback: either a single env var name (e.g.
 * `"OPENAI_API_KEY"`) or a resolver that inspects several env vars / probes
@@ -22,53 +18,13 @@ export type ModelManagerConfig = { apiKey?: string; baseUrl?: string; fetch?: Fe
 */
 export type KeyResolver = string | (() => string | undefined);

-/** Catalog discovery configuration for providers that support endpoint-based model listing. */
-export interface CatalogDiscoveryConfig {
-	/** Human-readable name for log messages. */
-	label: string;
-	/** Environment variables to check for API keys during catalog generation. */
-	envVars: readonly string[];
-	/** OAuth provider for credential refresh during catalog generation. */
-	oauthProvider?: string;
-	/** When true, catalog discovery proceeds even without credentials. */
-	allowUnauthenticated?: boolean;
-}
-
-/** Unified provider descriptor used by both runtime discovery and catalog generation. */
-export interface ProviderDescriptor {
-	providerId: string;
-	createModelManagerOptions(config: ModelManagerConfig): ModelManagerOptions<Api>;
-	/** Preferred model ID when no explicit selection is made. */
-	defaultModel: string;
-	/** When true, the runtime creates a model manager even without a valid API key (e.g. ollama). */
-	allowUnauthenticated?: boolean;
-	/** When true, successful runtime discovery replaces bundled provider models instead of merging fallback-only IDs. */
-	dynamicModelsAuthoritative?: boolean;
-	/** Catalog discovery configuration. Only providers with this field participate in generate-models.ts. */
-	catalogDiscovery?: CatalogDiscoveryConfig;
-}
-
-/** A provider descriptor that has catalog discovery configured. */
-export type CatalogProviderDescriptor = ProviderDescriptor & { catalogDiscovery: CatalogDiscoveryConfig };
-
-/** Type guard for descriptors with catalog discovery. */
-export function isCatalogDescriptor(d: ProviderDescriptor): d is CatalogProviderDescriptor {
-	return d.catalogDiscovery != null;
-}
-
-/** Whether catalog discovery may run without provider credentials. */
-export function allowsUnauthenticatedCatalogDiscovery(descriptor: CatalogProviderDescriptor): boolean {
-	return descriptor.catalogDiscovery.allowUnauthenticated ?? descriptor.allowUnauthenticated ?? false;
-}
-
 /**
- * Declarative description of a single provider. All fields are optional except
- * `id`/`name`; presence of a field opts the provider into a derived structure:
+ * Declarative description of a single provider's auth/login wiring. All
+ * fields are optional except `id`/`name`; presence of a field opts the
+ * provider into a derived structure:
 *
- * - `defaultModel` present ⇒ member of `KnownProvider` (a chat-model provider).
- * - `createModelManagerOptions` present (and not `specialModelManager`) ⇒
- *   appears in `PROVIDER_DESCRIPTORS` for runtime model discovery.
- * - `envKeys` present ⇒ env-var fallback in `getEnvApiKey`.
+ * - `envKeys` present ⇒ env-var fallback in `getEnvApiKey`, overriding the
+ *   catalog table's `envVars` for that provider.
 * - `login` present ⇒ member of `OAuthProvider`, shown in the `/login` list
 *   (unless `showInLoginList === false`) and dispatchable via `AuthStorage.login`.
 * - `callbackPort` present ⇒ entry in the auth-broker `CALLBACK_PORTS` map.
@@ -84,25 +40,7 @@ export interface ProviderDefinition {
 	readonly available?: boolean;
 	/** Whether to surface in the interactive login list. Defaults to true when `login` is present. */
 	readonly showInLoginList?: boolean;
-	// --- model discovery ---
-	/** Preferred model ID when no explicit selection is made. Presence ⇒ `KnownProvider` member. */
-	readonly defaultModel?: string;
-	/** Runtime model-manager factory. Omitted for login-only tools and catalog-only providers. */
-	readonly createModelManagerOptions?: (config: ModelManagerConfig) => ModelManagerOptions<Api>;
-	/** When true, the runtime creates a model manager even without a valid API key. */
-	readonly allowUnauthenticated?: boolean;
-	/** When true, successful runtime discovery replaces bundled provider models. */
-	readonly dynamicModelsAuthoritative?: boolean;
-	/** Catalog discovery configuration for generate-models.ts. */
-	readonly catalogDiscovery?: CatalogDiscoveryConfig;
-	/**
-	 * Providers whose model manager is constructed bespoke in the coding-agent
-	 * runtime (`google-antigravity`/`google-gemini-cli`/`openai-codex`). Excluded
-	 * from the derived `PROVIDER_DESCRIPTORS`; the registry supplies only their
-	 * identity/login/refresh/default-model metadata.
-	 */
-	readonly specialModelManager?: boolean;
-	// --- env-var fallback ---
+	// --- env-var fallback (the catalog table's `envVars` supplies plain names; set this only for computed resolvers) ---
 	readonly envKeys?: KeyResolver;
 	// --- interactive login (OAuthProviderInterface-compatible) ---
 	readonly login?: (callbacks: OAuthLoginCallbacks) => Promise<OAuthCredentials | string>;
@@ -1,7 +1,6 @@
-import { veniceModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://venice.ai/settings/api";
 const API_BASE_URL = "https://api.venice.ai/api/v1";
@@ -52,9 +51,5 @@ export async function loginVenice(options: OAuthController): Promise<string> {
 export const veniceProvider = {
 	id: "venice",
 	name: "Venice",
-	defaultModel: "llama-3.3-70b",
-	createModelManagerOptions: (config: ModelManagerConfig) => veniceModelManagerOptions(config),
-	catalogDiscovery: { label: "Venice", envVars: ["VENICE_API_KEY"], allowUnauthenticated: true },
-	envKeys: "VENICE_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginVenice(cb),
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { vercelAiGatewayModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://vercel.com/d?to=%2F%5Bteam%5D%2F%7E%2Fai-gateway%2Fapi-keys&title=AI+Gateway+API+Keys";

@@ -34,9 +33,5 @@ export async function loginVercelAiGateway(options: OAuthController): Promise<st
 export const vercelAiGatewayProvider = {
 	id: "vercel-ai-gateway",
 	name: "Vercel AI Gateway",
-	defaultModel: "anthropic/claude-sonnet-4-6",
-	createModelManagerOptions: (config: ModelManagerConfig) => vercelAiGatewayModelManagerOptions(config),
-	catalogDiscovery: { label: "Vercel AI Gateway", envVars: ["VERCEL_AI_GATEWAY_API_KEY"], allowUnauthenticated: true },
-	envKeys: "AI_GATEWAY_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginVercelAiGateway(cb),
 } as const satisfies ProviderDefinition;
@@ -1,6 +1,5 @@
-import { vllmModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthController, OAuthLoginCallbacks, OAuthProvider } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const PROVIDER_ID: OAuthProvider = "vllm";
 const AUTH_URL = "https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html";
@@ -30,9 +29,5 @@ export async function loginVllm(options: OAuthController): Promise<string> {
 export const vllmProvider = {
 	id: "vllm",
 	name: "vLLM (Local OpenAI-compatible)",
-	defaultModel: "gpt-oss-20b",
-	createModelManagerOptions: (config: ModelManagerConfig) => vllmModelManagerOptions(config),
-	catalogDiscovery: { label: "vLLM", envVars: ["VLLM_API_KEY"], allowUnauthenticated: true },
-	envKeys: "VLLM_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginVllm(cb),
 } as const satisfies ProviderDefinition;
@@ -1,14 +1,9 @@
-import { waferPassModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const waferPassProvider = {
 	id: "wafer-pass",
 	name: "Wafer Pass (flat-rate subscription)",
-	defaultModel: "GLM-5.1",
-	createModelManagerOptions: (config: ModelManagerConfig) => waferPassModelManagerOptions(config),
-	catalogDiscovery: { label: "Wafer Pass", envVars: ["WAFER_PASS_API_KEY"], oauthProvider: "wafer-pass" },
-	envKeys: "WAFER_PASS_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginWaferPass } = await import("./oauth/wafer");
@@ -1,18 +1,9 @@
-import { waferServerlessModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const waferServerlessProvider = {
 	id: "wafer-serverless",
 	name: "Wafer Serverless (pay-as-you-go)",
-	defaultModel: "GLM-5.1",
-	createModelManagerOptions: (config: ModelManagerConfig) => waferServerlessModelManagerOptions(config),
-	catalogDiscovery: {
-		label: "Wafer Serverless",
-		envVars: ["WAFER_SERVERLESS_API_KEY"],
-		oauthProvider: "wafer-serverless",
-	},
-	envKeys: "WAFER_SERVERLESS_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginWaferServerless } = await import("./oauth/wafer");
@@ -1,19 +1,9 @@
-import { $pickenv } from "@oh-my-pi/pi-utils";
-import { xaiOAuthModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const xaiOauthProvider = {
 	id: "xai-oauth",
 	name: "xAI Grok OAuth (SuperGrok Subscription)",
-	defaultModel: "grok-4.3",
-	createModelManagerOptions: (config: ModelManagerConfig) => xaiOAuthModelManagerOptions(config),
-	catalogDiscovery: {
-		label: "xAI Grok OAuth (SuperGrok)",
-		envVars: ["XAI_OAUTH_TOKEN", "XAI_API_KEY"],
-		oauthProvider: "xai-oauth",
-	},
-	envKeys: () => $pickenv("XAI_OAUTH_TOKEN", "XAI_API_KEY"),
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginXAIOAuth } = await import("./oauth/xai-oauth");
@@ -1,10 +1,6 @@
-import { xaiModelManagerOptions } from "../provider-models/openai-compat";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const xaiProvider = {
 	id: "xai",
 	name: "xAI",
-	defaultModel: "grok-4-fast-non-reasoning",
-	createModelManagerOptions: (config: ModelManagerConfig) => xaiModelManagerOptions(config),
-	envKeys: "XAI_API_KEY",
 } as const satisfies ProviderDefinition;
@@ -1,14 +1,9 @@
-import { xiaomiModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const xiaomiTokenPlanAmsProvider = {
 	id: "xiaomi-token-plan-ams",
 	name: "Xiaomi Token Plan (Europe)",
-	defaultModel: "mimo-v2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) =>
-		xiaomiModelManagerOptions({ ...config, providerId: "xiaomi-token-plan-ams", tokenPlanRegion: "ams" }),
-	envKeys: "XIAOMI_TOKEN_PLAN_AMS_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginXiaomiTokenPlan } = await import("./oauth/xiaomi");
@@ -1,14 +1,9 @@
-import { xiaomiModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const xiaomiTokenPlanCnProvider = {
 	id: "xiaomi-token-plan-cn",
 	name: "Xiaomi Token Plan (China)",
-	defaultModel: "mimo-v2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) =>
-		xiaomiModelManagerOptions({ ...config, providerId: "xiaomi-token-plan-cn", tokenPlanRegion: "cn" }),
-	envKeys: "XIAOMI_TOKEN_PLAN_CN_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginXiaomiTokenPlan } = await import("./oauth/xiaomi");
@@ -1,14 +1,9 @@
-import { xiaomiModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const xiaomiTokenPlanSgpProvider = {
 	id: "xiaomi-token-plan-sgp",
 	name: "Xiaomi Token Plan (Singapore)",
-	defaultModel: "mimo-v2.5",
-	createModelManagerOptions: (config: ModelManagerConfig) =>
-		xiaomiModelManagerOptions({ ...config, providerId: "xiaomi-token-plan-sgp", tokenPlanRegion: "sgp" }),
-	envKeys: "XIAOMI_TOKEN_PLAN_SGP_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginXiaomiTokenPlan } = await import("./oauth/xiaomi");
@@ -1,14 +1,9 @@
-import { xiaomiModelManagerOptions } from "../provider-models/openai-compat";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const xiaomiProvider = {
 	id: "xiaomi",
 	name: "Xiaomi MiMo",
-	defaultModel: "mimo-v2-flash",
-	createModelManagerOptions: (config: ModelManagerConfig) => xiaomiModelManagerOptions(config),
-	catalogDiscovery: { label: "Xiaomi", envVars: ["XIAOMI_API_KEY"] },
-	envKeys: "XIAOMI_API_KEY",
 	login: async (cb: OAuthLoginCallbacks) => {
 		// Lazy import: keep heavy OAuth flow modules out of the eager registry graph.
 		const { loginXiaomi } = await import("./oauth/xiaomi");
@@ -1,7 +1,6 @@
-import { zaiModelManagerOptions } from "../provider-models/special";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://z.ai/manage-apikey/apikey-list";
 const API_BASE_URL = "https://api.z.ai/api/coding/paas/v4";
@@ -45,9 +44,5 @@ export async function loginZai(options: OAuthController): Promise<string> {
 export const zaiProvider = {
 	id: "zai",
 	name: "Z.AI (GLM Coding Plan)",
-	defaultModel: "glm-5.1",
-	createModelManagerOptions: (config: ModelManagerConfig) => zaiModelManagerOptions(config),
-	catalogDiscovery: { label: "zAI", envVars: ["ZAI_API_KEY"] },
-	envKeys: "ZAI_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginZai(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { zenmuxModelManagerOptions } from "../provider-models/openai-compat";
 import { createApiKeyLogin } from "./api-key-login";
 import type { OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 export const loginZenMux = createApiKeyLogin({
 	providerLabel: "ZenMux",
@@ -19,9 +18,5 @@ export const loginZenMux = createApiKeyLogin({
 export const zenmuxProvider = {
 	id: "zenmux",
 	name: "ZenMux",
-	defaultModel: "anthropic/claude-opus-4.6",
-	createModelManagerOptions: (config: ModelManagerConfig) => zenmuxModelManagerOptions(config),
-	catalogDiscovery: { label: "ZenMux", envVars: ["ZENMUX_API_KEY"] },
-	envKeys: "ZENMUX_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginZenMux(cb),
 } as const satisfies ProviderDefinition;
@@ -1,7 +1,6 @@
-import { zhipuCodingPlanModelManagerOptions } from "../provider-models/openai-compat";
 import { validateOpenAICompatibleApiKey } from "./api-key-validation";
 import type { OAuthController, OAuthLoginCallbacks } from "./oauth/types";
-import type { ModelManagerConfig, ProviderDefinition } from "./types";
+import type { ProviderDefinition } from "./types";

 const AUTH_URL = "https://bigmodel.cn/coding-plan/personal/overview";
 const API_BASE_URL = "https://open.bigmodel.cn/api/coding/paas/v4";
@@ -45,9 +44,5 @@ export async function loginZhipuCodingPlan(options: OAuthController): Promise<st
 export const zhipuCodingPlanProvider = {
 	id: "zhipu-coding-plan",
 	name: "Zhipu Coding Plan (智谱)",
-	defaultModel: "glm-5.1",
-	createModelManagerOptions: (config: ModelManagerConfig) => zhipuCodingPlanModelManagerOptions(config),
-	catalogDiscovery: { label: "Zhipu Coding Plan", envVars: ["ZHIPU_API_KEY"] },
-	envKeys: "ZHIPU_API_KEY",
 	login: (cb: OAuthLoginCallbacks) => loginZhipuCodingPlan(cb),
 } as const satisfies ProviderDefinition;
@@ -1,13 +1,14 @@
-import { $env, extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
-import { getCustomApi } from "./api-registry";
-import { AUTH_RETRY_STEPS, isApiKeyResolver, resolveRetryKey } from "./auth-retry";
-import type { Effort } from "./effort";
+import type { Effort } from "@oh-my-pi/pi-catalog/effort";
 import {
 	mapEffortToAnthropicAdaptiveEffort,
 	mapEffortToGoogleThinkingLevel,
 	modelOmitsReasoningEffort,
 	requireSupportedEffort,
-} from "./model-thinking";
+} from "@oh-my-pi/pi-catalog/model-thinking";
+import { CATALOG_PROVIDERS, type ProviderCatalogEntry } from "@oh-my-pi/pi-catalog/provider-models";
+import { $env, $pickenv, extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
+import { getCustomApi } from "./api-registry";
+import { AUTH_RETRY_STEPS, isApiKeyResolver, resolveRetryKey } from "./auth-retry";
 import type { BedrockOptions } from "./providers/amazon-bedrock";
 import type { AnthropicOptions } from "./providers/anthropic";
 import type { CursorOptions } from "./providers/cursor";
@@ -166,7 +167,20 @@ const LEGACY_ENV_KEYS: Record<string, KeyResolver> = {
 	brave: "BRAVE_API_KEY",
 };

+/**
+ * Env fallbacks derived from the catalog table — the single source for plain
+ * provider env-var names. Registry defs override with computed resolvers
+ * (Foundry/ADC/Bedrock probes); legacy non-provider keys merge last.
+ */
+const CATALOG_ENTRY_ENV_KEYS = (CATALOG_PROVIDERS as readonly ProviderCatalogEntry[]).flatMap(provider => {
+	const envVars = provider.envVars;
+	if (!envVars || envVars.length === 0) return [];
+	const resolver: KeyResolver = envVars.length === 1 ? envVars[0] : () => $pickenv(...envVars);
+	return [[provider.id, resolver] as [string, KeyResolver]];
+});
+
 const serviceProviderMap: Record<string, KeyResolver> = {
+	...Object.fromEntries(CATALOG_ENTRY_ENV_KEYS),
 	...Object.fromEntries(
 		PROVIDER_REGISTRY.flatMap(provider =>
 			provider.envKeys != null ? [[provider.id, provider.envKeys] as [string, KeyResolver]] : [],
@@ -1,9 +1,6 @@
-import type { ZodType, z } from "zod/v4";
-import type { ApiKey } from "./auth-retry";
-import type { BedrockOptions } from "./providers/amazon-bedrock";
-import type { AnthropicOptions } from "./providers/anthropic";
-import type { AzureOpenAIResponsesOptions } from "./providers/azure-openai-responses";
-import type { CursorOptions } from "./providers/cursor";
+export * from "@oh-my-pi/pi-catalog/effort";
+export * from "@oh-my-pi/pi-catalog/types";
+
 import type {
 	DeleteArgs,
 	DeleteResult,
@@ -20,7 +17,15 @@ import type {
 	ShellResult,
 	WriteArgs,
 	WriteResult,
-} from "./providers/cursor/gen/agent_pb";
+} from "@oh-my-pi/pi-catalog/discovery/cursor-gen/agent_pb";
+import type { Effort } from "@oh-my-pi/pi-catalog/effort";
+import type { Api, FetchImpl, KnownApi, Model, Provider, ThinkingBudgets, Usage } from "@oh-my-pi/pi-catalog/types";
+import type { ZodType, z } from "zod/v4";
+import type { ApiKey } from "./auth-retry";
+import type { BedrockOptions } from "./providers/amazon-bedrock";
+import type { AnthropicOptions } from "./providers/anthropic";
+import type { AzureOpenAIResponsesOptions } from "./providers/azure-openai-responses";
+import type { CursorOptions } from "./providers/cursor";
 import type { GoogleOptions } from "./providers/google";
 import type { GoogleGeminiCliOptions } from "./providers/google-gemini-cli";
 import type { GoogleVertexOptions } from "./providers/google-vertex";
@@ -28,7 +33,6 @@ import type { OllamaChatOptions } from "./providers/ollama";
 import type { OpenAICodexResponsesOptions } from "./providers/openai-codex-responses";
 import type { OpenAICompletionsOptions } from "./providers/openai-completions";
 import type { OpenAIResponsesOptions } from "./providers/openai-responses";
-import type { KnownProviderId } from "./registry";
 import type { AssistantMessageEventStream } from "./utils/event-stream";

 export type { AssistantMessageEventStream } from "./utils/event-stream";
@@ -46,19 +50,6 @@ export type { AssistantMessageEventStream } from "./utils/event-stream";
 */
 export const OPENAI_MAX_OUTPUT_TOKENS = 64000;

-export type KnownApi =
-	| "openai-completions"
-	| "openai-responses"
-	| "openai-codex-responses"
-	| "azure-openai-responses"
-	| "anthropic-messages"
-	| "bedrock-converse-stream"
-	| "google-generative-ai"
-	| "google-gemini-cli"
-	| "google-vertex"
-	| "ollama-chat"
-	| "cursor-agent";
-export type Api = KnownApi | (string & {});
 export interface ApiOptionsMap {
 	"anthropic-messages": AnthropicOptions;
 	"bedrock-converse-stream": BedrockOptions;
@@ -84,44 +75,6 @@ export type OptionsForApi<TApi extends Api> =
 	| StreamOptions
 	| (TApi extends keyof ApiOptionsMap ? ApiOptionsMap[TApi] : never);

-/** Canonical thinking transport used by a model. */
-export type ThinkingControlMode =
-	| "effort"
-	| "budget"
-	| "google-level"
-	| "anthropic-adaptive"
-	| "anthropic-budget-effort";
-
-/** Per-model thinking capabilities used to clamp and map user-facing effort levels. */
-export interface ThinkingConfig {
-	/** Least intensive supported user-facing effort level. */
-	minLevel: Effort;
-	/** Most intensive supported user-facing effort level. */
-	maxLevel: Effort;
-	/**
-	 * Optional explicit list of supported levels. When present, takes precedence over
-	 * the `minLevel`..`maxLevel` range — used to encode discrete sets with gaps
-	 * (e.g. Gemini 3 Pro supports `low` and `high` but not `medium`).
-	 */
-	levels?: readonly Effort[];
-	/** Optional default effort applied when this model is selected. Falls back to global default if absent. */
-	defaultLevel?: Effort;
-	/** Provider-specific transport used to encode the selected effort. */
-	mode: ThinkingControlMode;
-}
-
-export type KnownProvider = KnownProviderId;
-// `Provider` is any provider-id string; `KnownProvider` enumerates the built-in model
-// providers. Kept structurally `string` (the prior `KnownProvider | string` already
-// collapsed to `string`) so the registry-derived `KnownProvider` can reference the model
-// types below without forming a circular type-alias reference.
-export type Provider = string;
-
-import type { Effort } from "./effort";
-
-/** Token budgets for each thinking level (token-based providers only) */
-export type ThinkingBudgets = { [key in Effort]?: number };
-
 export interface TokenTaskBudget {
 	type: "tokens";
 	total: number;
@@ -233,15 +186,6 @@ export interface RawSseEvent {
 	raw: string[];
 }

-/**
- * `fetch`-compatible function. Accepts any callable matching the standard
- * fetch signature; `preconnect` is optional because non-Bun runtimes (browsers,
- * test mocks) won't expose it.
- */
-export type FetchImpl = ((input: string | URL | Request, init?: RequestInit) => Promise<Response>) & {
-	preconnect?: typeof globalThis.fetch.preconnect;
-};
-
 export interface StreamOptions {
 	temperature?: number;
 	topP?: number;
@@ -484,53 +428,6 @@ export interface ToolCall {
 	customWireName?: string;
 }

-export interface Usage {
-	/** Non-cached input tokens (matches the bucket the provider bills as new input). */
-	input: number;
-	/** Total output tokens for the turn, including thinking, assistant text, and tool-call argument tokens. */
-	output: number;
-	/** Tokens read from the prompt cache. */
-	cacheRead: number;
-	/** Tokens written to the prompt cache (cache creation). */
-	cacheWrite: number;
-	/** Sum of input + output + cacheRead + cacheWrite. */
-	totalTokens: number;
-	/** Copilot premium-request counter, when applicable. */
-	premiumRequests?: number;
-	/**
-	 * Reasoning/thinking tokens included in `output`, when the provider reports them
-	 * (OpenAI `output_tokens_details.reasoning_tokens`, Google `thoughtsTokenCount`).
-	 * Always a subset of `output` — non-reasoning output is `output - reasoningTokens`.
-	 *
-	 * Providers that don't expose this leave it undefined rather than guessing;
-	 * `undefined` means unknown, NOT zero.
-	 */
-	reasoningTokens?: number;
-	/**
-	 * Cache-write TTL breakdown (Anthropic only). When set, the components sum to
-	 * `cacheWrite`. Absent providers do not populate this.
-	 */
-	cttl?: {
-		ephemeral5m?: number;
-		ephemeral1h?: number;
-	};
-	/**
-	 * Server-side tool invocations made during this turn (Anthropic web_search /
-	 * web_fetch, OpenAI built-in tools when reported). Counts requests, not tokens.
-	 */
-	server?: {
-		webSearch?: number;
-		webFetch?: number;
-	};
-	cost: {
-		input: number;
-		output: number;
-		cacheRead: number;
-		cacheWrite: number;
-		total: number;
-	};
-}
-
 export type StopReason = "stop" | "length" | "toolUse" | "error" | "aborted";

 export interface OpenAIResponsesHistoryPayload {
@@ -724,227 +621,3 @@ export type AssistantMessageEvent =
 			reason: Extract<StopReason, "aborted" | "error">;
 			error: AssistantMessage;
 	  };
-
-/**
- * Compatibility settings for openai-completions API.
- * Use this to override URL-based auto-detection for custom providers.
- */
-export interface OpenAICompat {
-	/** Whether the provider supports the `store` field. Default: auto-detected from URL. */
-	supportsStore?: boolean;
-	/** Whether the provider supports the `developer` role (vs `system`). Default: auto-detected from URL. */
-	supportsDeveloperRole?: boolean;
-	/**
-	 * Whether the provider's chat-completions endpoint accepts multiple
-	 * leading `system`/`developer` messages. When false, ordered system
-	 * prompts are coalesced into a single message joined by `\n\n` so
-	 * strict chat templates (e.g. Qwen-served via vLLM, MiniMax) accept
-	 * the request. Default: detected per provider/baseUrl. Canonical
-	 * OpenAI/Azure/OpenRouter/Cerebras/Together/Fireworks/Groq/DeepSeek/
-	 * Mistral/xAI/Z.ai/GitHub Copilot/Zenmux are treated as `true`;
-	 * unknown or strict-template hosts default to `false`. Setting this
-	 * to `true` preserves separate blocks, which is preferred for
-	 * KV-cache reuse when the trailing prompt changes between calls.
-	 */
-	supportsMultipleSystemMessages?: boolean;
-	/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */
-	supportsReasoningEffort?: boolean;
-	/** Optional mapping from pi-ai reasoning levels to provider/model-specific `reasoning_effort` values. */
-	reasoningEffortMap?: Partial<Record<Effort, string>>;
-	/** Whether the provider supports `stream_options: { include_usage: true }` for token usage in streaming responses. Default: true. */
-	supportsUsageInStreaming?: boolean;
-	/** Which field to use for max tokens. Default: auto-detected from URL. */
-	maxTokensField?: "max_completion_tokens" | "max_tokens";
-	/** Whether tool results require the `name` field. Default: auto-detected from URL. */
-	requiresToolResultName?: boolean;
-	/** Whether a user message after tool results requires an assistant message in between. Default: auto-detected from URL. */
-	requiresAssistantAfterToolResult?: boolean;
-	/** Whether thinking blocks must be converted to text blocks with <thinking> delimiters. Default: auto-detected from URL. */
-	requiresThinkingAsText?: boolean;
-	/** Whether tool call IDs must be normalized to Mistral format (exactly 9 alphanumeric chars). Default: auto-detected from URL. */
-	requiresMistralToolIds?: boolean;
-	/** Format for reasoning/thinking parameter. "openai" uses reasoning_effort, "openrouter" uses reasoning: { effort }, "zai" uses thinking: { type: "enabled" | "disabled" } (also used by Moonshot Kimi), "qwen" uses top-level enable_thinking, and "qwen-chat-template" uses chat_template_kwargs.enable_thinking. Default: "openai". */
-	thinkingFormat?: "openai" | "openrouter" | "zai" | "qwen" | "qwen-chat-template";
-	/** Optional `thinking.keep` value for Z.ai/Moonshot-style thinking params. Set false to suppress auto-detected keep. Default: auto-detected. */
-	thinkingKeep?: "all" | false;
-	/** Which reasoning content field to emit on assistant messages. Default: auto-detected. */
-	reasoningContentField?: "reasoning_content" | "reasoning" | "reasoning_text";
-	/** Whether assistant tool-call messages must include reasoning content. Default: false. */
-	requiresReasoningContentForToolCalls?: boolean;
-	/** Whether the provider accepts a synthetic placeholder (e.g. ".") for missing reasoning_content on tool-call turns. Default: true. Set to false for providers like DeepSeek that validate the exact reasoning_content value. */
-	allowsSyntheticReasoningContentForToolCalls?: boolean;
-	/** Whether assistant tool-call messages must include non-empty content. Default: false. */
-	requiresAssistantContentForToolCalls?: boolean;
-	/** Whether the provider supports the `tool_choice` parameter. Default: true. */
-	supportsToolChoice?: boolean;
-	/**
-	 * Drop reasoning fields (`reasoning_effort`, OpenRouter `reasoning`) for
-	 * the request when `tool_choice` forces a tool call. Mirrors the Anthropic
-	 * `disableThinkingIfToolChoiceForced` rule for backends like Kimi that
-	 * 400 with `tool_choice 'specified' is incompatible with thinking
-	 * enabled` whenever both are present. Default: auto-detected (Kimi).
-	 */
-	disableReasoningOnForcedToolChoice?: boolean;
-	/**
-	 * Drop reasoning fields (`reasoning_effort`, OpenRouter `reasoning`) for
-	 * any request that sends `tool_choice`. Use for providers/models that accept
-	 * tools and `tool_choice`, but reject `tool_choice` while thinking is enabled.
-	 * Default: auto-detected (DeepSeek reasoning models).
-	 */
-	disableReasoningOnToolChoice?: boolean;
-	/** OpenRouter-specific routing preferences. Only used when baseUrl points to OpenRouter. */
-	openRouterRouting?: OpenRouterRouting;
-	/** Vercel AI Gateway routing preferences. Only used when baseUrl points to Vercel AI Gateway. */
-	vercelGatewayRouting?: VercelGatewayRouting;
-	/** Extra fields to include in request body (e.g. gateway routing hints for OpenClaw-style proxies). */
-	extraBody?: Record<string, unknown>;
-	/** Whether chat-completions payloads should include provider-specific prompt-cache markers. */
-	cacheControlFormat?: "anthropic" | undefined;
-	/** Whether the provider supports the `strict` field in tool definitions. Default: auto-detected per provider/baseUrl (conservative for unknown providers). */
-	supportsStrictMode?: boolean;
-	/** Whether tool schemas must be sent either all strict or all non-strict. Undefined keeps the existing per-tool mixed behavior. */
-	toolStrictMode?: "all_strict" | "none";
-}
-
-/**
- * Compatibility settings for anthropic-messages API.
- * Use this to disable features that strict-by-default Anthropic accepts but
- * that proxy gateways (Vertex AI, AWS Bedrock-style fronts, etc.) reject.
- */
-export interface AnthropicCompat {
-	/**
-	 * Drop the top-level `strict: true` field on tool definitions. Vertex AI's
-	 * Anthropic-compatible endpoint rejects unknown tool fields with
-	 * `tools.<n>.custom.strict: Extra inputs are not permitted`.
-	 */
-	disableStrictTools?: boolean;
-	/**
-	 * Map adaptive thinking (`thinking: { type: "adaptive" }`) to
-	 * `{ type: "enabled", budget_tokens }`. Vertex AI rejects the `adaptive`
-	 * tag with `Input tag 'adaptive' ... does not match any of the expected
-	 * tags: 'disabled', 'enabled'`.
-	 */
-	disableAdaptiveThinking?: boolean;
-	/** Whether tools may include Anthropic's per-tool eager_input_streaming flag. Default: true. */
-	supportsEagerToolInputStreaming?: boolean;
-	/** Whether long prompt-cache retention (`ttl: "1h"`) is supported. Default: true for canonical Anthropic API. */
-	supportsLongCacheRetention?: boolean;
-	/**
-	 * Whether mid-conversation `role: "system"` messages are accepted in the
-	 * `messages` array (Claude Opus 4.8+ and Claude Fable/Mythos 5 on the
-	 * first-party Claude API and Claude Platform on AWS). When unset,
-	 * auto-detected from the model id and base URL. Not available on Bedrock,
-	 * Vertex AI, or Microsoft Foundry.
-	 */
-	supportsMidConversationSystem?: boolean;
-	/**
-	 * Whether the model accepts a forced `tool_choice` (`{ type: "any" }` or
-	 * `{ type: "tool", name }`). Claude Fable/Mythos 5 reject forced tool use
-	 * outright ("tool_choice forces tool use is not compatible with this model");
-	 * the request builder downgrades forced choices to `auto` when this is false.
-	 * When unset, auto-detected from the model id. Default: true.
-	 */
-	supportsForcedToolChoice?: boolean;
-}
-
-/**
- * OpenRouter provider routing preferences.
- * Controls which upstream providers OpenRouter routes requests to.
- * @see https://openrouter.ai/docs/provider-routing
- */
-export interface OpenRouterRouting {
-	/** List of provider slugs to exclusively use for this request (e.g., ["amazon-bedrock", "anthropic"]). */
-	only?: string[];
-	/** List of provider slugs to try in order (e.g., ["anthropic", "openai"]). */
-	order?: string[];
-}
-
-/**
- * Vercel AI Gateway routing preferences.
- * Controls which upstream providers the gateway routes requests to.
- * @see https://vercel.com/docs/ai-gateway/models-and-providers/provider-options
- */
-export interface VercelGatewayRouting {
-	/** List of provider slugs to exclusively use for this request (e.g., ["bedrock", "anthropic"]). */
-	only?: string[];
-	/** List of provider slugs to try in order (e.g., ["anthropic", "openai"]). */
-	order?: string[];
-}
-
-// Model interface for the unified model system
-export interface Model<TApi extends Api = any> {
-	id: string;
-	name: string;
-	api: TApi;
-	provider: Provider;
-	baseUrl: string;
-	reasoning: boolean;
-	input: ("text" | "image")[];
-	cost: {
-		input: number; // $/million tokens
-		output: number; // $/million tokens
-		cacheRead: number; // $/million tokens
-		cacheWrite: number; // $/million tokens
-	};
-	/** Premium Copilot requests charged per user-initiated request (defaults to 1). */
-	premiumMultiplier?: number;
-	contextWindow: number;
-	maxTokens: number;
-	/**
-	 * When `true`, providers MUST omit `max_output_tokens` (Responses) /
-	 * `max_tokens` / `max_completion_tokens` (Completions) from the outbound
-	 * request and let the upstream API decide the per-response cap. `maxTokens`
-	 * is still used locally for budgeting (compaction, context promotion); only
-	 * the wire field is suppressed.
-	 *
-	 * Use this for proxies (notably Ollama) that forward to a backend whose true
-	 * output limit OMP cannot discover — sending the wrong value triggers 400s
-	 * from the upstream provider.
-	 */
-	omitMaxOutputTokens?: boolean;
-	headers?: Record<string, string>;
-	/**
-	 * Streaming transport override. When `"pi-native"`, `streamSimple` routes
-	 * the request to the model's `baseUrl` via the auth-gateway's
-	 * `POST /v1/pi/stream` endpoint instead of dispatching the per-API
-	 * provider client. The `baseUrl` must point at an `omp auth-gateway`
-	 * (or compatible) host; `headers.Authorization` (or `apiKey` resolved by
-	 * the registry) carries the gateway bearer.
-	 *
-	 * Used by containerized omp installs (e.g. robomp slots) to route every
-	 * LLM call through a sidecar gateway that holds the real provider
-	 * credentials. The model's other metadata (pricing, context window,
-	 * thinking config, …) still resolves locally; only the streaming
-	 * dispatch is redirected.
-	 */
-	transport?: "pi-native";
-	/** Hint that websocket transport should be preferred when supported by the provider implementation. */
-	preferWebsockets?: boolean;
-	/** Preferred model to switch to when context promotion is triggered (model id or provider/id). */
-	contextPromotionTarget?: string;
-	/** Provider-assigned priority value (lower = higher priority). */
-	priority?: number;
-	/** Canonical thinking capability metadata for this model. */
-	thinking?: ThinkingConfig;
-	/** Compatibility overrides per API. If not set, auto-detected from baseUrl. */
-	compat?: TApi extends "openai-completions" | "openai-responses"
-		? OpenAICompat
-		: TApi extends "anthropic-messages"
-			? AnthropicCompat
-			: never;
-	/**
-	 * Which shape to use when exposing the Codex `apply_patch` tool to this model.
-	 * Generated catalog policy sets `"freeform"` for first-party GPT-5 Responses
-	 * models that support OpenAI custom tools with a Lark grammar. The freeform
-	 * variant sends a raw patch string with no JSON envelope.
-	 * - `"function"` or undefined: JSON function-tool with `{input: string}` (spec §1.2).
-	 */
-	applyPatchToolType?: "freeform" | "function";
-	/**
-	 * Force OAuth-style request shaping for providers whose API key prefix doesn't
-	 * match an OAuth token (e.g. routing Anthropic traffic through a proxy that
-	 * expects Claude Code framing). When true, the streaming layer sets
-	 * `options.isOAuth = true` for the underlying provider call.
-	 */
-	isOAuth?: boolean;
-}
@@ -1,4 +1,5 @@
 import { scheduler } from "node:timers/promises";
+import { toNumber } from "@oh-my-pi/pi-catalog/utils";
 import { claudeCodeVersion } from "../providers/anthropic";
 import type {
 	CredentialRankingStrategy,
@@ -11,7 +12,7 @@ import type {
 	UsageStatus,
 	UsageWindow,
 } from "../usage";
-import { isRecord, toNumber } from "../utils";
+import { isRecord } from "../utils";

 const DEFAULT_ENDPOINT = "https://api.anthropic.com/api/oauth";
 const FIVE_HOURS_MS = 5 * 60 * 60 * 1000;
@@ -4,7 +4,8 @@
 * Normalizes Copilot quota usage into the shared UsageReport schema.
 */

-import { OPENCODE_HEADERS } from "../registry/oauth/github-copilot";
+import { toBoolean, toNumber } from "@oh-my-pi/pi-catalog/utils";
+import { OPENCODE_HEADERS } from "@oh-my-pi/pi-catalog/wire/github-copilot";
 import type {
 	UsageAmount,
 	UsageFetchContext,
@@ -15,7 +16,7 @@ import type {
 	UsageStatus,
 	UsageWindow,
 } from "../usage";
-import { isRecord, toBoolean, toNumber } from "../utils";
+import { isRecord } from "../utils";

 type CopilotQuotaDetail = {
 	entitlement: number;
@@ -1,4 +1,4 @@
-import { getAntigravityUserAgent } from "../providers/google-gemini-headers";
+import { getAntigravityUserAgent } from "@oh-my-pi/pi-catalog/wire/gemini-headers";
 import type {
 	CredentialRankingStrategy,
 	UsageAmount,
@@ -1,5 +1,5 @@
 import { Buffer } from "node:buffer";
-import { CODEX_BASE_URL } from "../providers/openai-codex/constants";
+import { CODEX_BASE_URL } from "@oh-my-pi/pi-catalog/wire/codex";
 import type {
 	CredentialRankingStrategy,
 	UsageAmount,
@@ -1,3 +1,4 @@
+import { toNumber } from "@oh-my-pi/pi-catalog/utils";
 import type {
 	UsageAmount,
 	UsageFetchContext,
@@ -8,7 +9,7 @@ import type {
 	UsageStatus,
 	UsageWindow,
 } from "../usage";
-import { isRecord, toNumber } from "../utils";
+import { isRecord } from "../utils";

 const DEFAULT_ENDPOINT = "https://api.z.ai";
 const QUOTA_PATH = "/api/monitor/usage/quota/limit";
--- a/Show More
+++ b/Show More