@prometheus-ai/ai 0.5.4 → 0.5.8

package/src/types.ts

@@ -1,8 +1,6 @@
-import type { ZodType, z } from "zod/v4";
-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 "@prometheus-ai/catalog/effort";
+export * from "@prometheus-ai/catalog/types";
+
 import type {
 	DeleteArgs,
 	DeleteResult,
@@ -19,7 +17,16 @@ import type {
 	ShellResult,
 	WriteArgs,
 	WriteResult,
-} from "./providers/cursor/gen/agent_pb";
+} from "@prometheus-ai/catalog/discovery/cursor-gen/agent_pb";
+import type { Effort } from "@prometheus-ai/catalog/effort";
+import type { Api, FetchImpl, KnownApi, Model, Provider, ThinkingBudgets, Usage } from "@prometheus-ai/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 { StopDetails } from "./providers/anthropic-wire";
+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";
@@ -29,21 +36,22 @@ import type { OpenAICompletionsOptions } from "./providers/openai-completions";
 import type { OpenAIResponsesOptions } from "./providers/openai-responses";
 import type { AssistantMessageEventStream } from "./utils/event-stream";
 
+export type { StopDetails } from "./providers/anthropic-wire";
 export type { AssistantMessageEventStream } from "./utils/event-stream";
 
-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 & {});
+/**
+ * Ceiling on the output-token count prometheus requests from any OpenAI-family endpoint
+ * (openai-responses, azure/xai responses, and openai-completions). Mirrors
+ * Anthropic's {@link CLAUDE_CODE_MAX_OUTPUT_TOKENS}.
+ *
+ * Catalog `maxTokens` frequently reflects a model's context window rather than a
+ * given upstream's real per-request output cap. OpenRouter, for instance,
+ * advertises 131072 output tokens for `z-ai/glm-4.7`, but the Cerebras upstream
+ * only allows ~131072 tokens total — so requesting the full ceiling overflows
+ * with a 400. Requested output is clamped to this value (and to `model.maxTokens`).
+ */
+export const OPENAI_MAX_OUTPUT_TOKENS = 64000;
+
 export interface ApiOptionsMap {
 	"anthropic-messages": AnthropicOptions;
 	"bedrock-converse-stream": BedrockOptions;
@@ -69,93 +77,6 @@ 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 =
-	| "alibaba-coding-plan"
-	| "amazon-bedrock"
-	| "anthropic"
-	| "google"
-	| "google-gemini-cli"
-	| "google-antigravity"
-	| "google-vertex"
-	| "openai"
-	| "openai-codex"
-	| "kimi-code"
-	| "minimax-code"
-	| "minimax-code-cn"
-	| "github-copilot"
-	| "fireworks"
-	| "firepass"
-	| "gitlab-duo"
-	| "cursor"
-	| "deepseek"
-	| "xai"
-	| "xai-oauth"
-	| "groq"
-	| "cerebras"
-	| "openrouter"
-	| "kilo"
-	| "vercel-ai-gateway"
-	| "zai"
-	| "zhipu-coding-plan"
-	| "mistral"
-	| "minimax"
-	| "opencode-go"
-	| "opencode-zen"
-	| "synthetic"
-	| "cloudflare-ai-gateway"
-	| "huggingface"
-	| "litellm"
-	| "moonshot"
-	| "nvidia"
-	| "nanogpt"
-	| "ollama"
-	| "ollama-cloud"
-	| "qianfan"
-	| "qwen-portal"
-	| "together"
-	| "venice"
-	| "vllm"
-	| "xiaomi"
-	| "xiaomi-token-plan-sgp"
-	| "xiaomi-token-plan-ams"
-	| "xiaomi-token-plan-cn"
-	| "wafer-pass"
-	| "wafer-serverless"
-	| "zenmux"
-	| "lm-studio";
-export type Provider = KnownProvider | string;
-
-import type { Effort } from "./model-thinking";
-
-/** Token budgets for each thinking level (token-based providers only) */
-export type ThinkingBudgets = { [key in Effort]?: number };
-
 export interface TokenTaskBudget {
 	type: "tokens";
 	total: number;
@@ -267,15 +188,6 @@ 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;
@@ -298,12 +210,6 @@ export interface StreamOptions {
 	maxTokens?: number;
 	signal?: AbortSignal;
 	apiKey?: string;
-	/**
-	 * Called when a provider returns 401 before any replay-unsafe assistant
-	 * event has been emitted. Returning a different key retries the provider
-	 * request once.
-	 */
-	onAuthError?: (provider: string, apiKey: string, error: unknown) => Promise<string | undefined>;
 	cacheRetention?: CacheRetention;
 	/**
 	 * Additional headers to include in provider requests.
@@ -415,7 +321,21 @@ export interface StreamOptions {
 }
 
 // Unified options with reasoning passed to streamSimple() and completeSimple()
-export interface SimpleStreamOptions extends StreamOptions {
+export interface SimpleStreamOptions extends Omit<StreamOptions, "apiKey"> {
+	/**
+	 * API key for the request: either a static bearer string, or an
+	 * {@link ApiKeyResolver} that mints/rotates the key across the central
+	 * a/b/c auth-retry policy. `streamSimple`/`completeSimple` resolve a
+	 * resolver to a string before per-provider dispatch, so providers only
+	 * ever see the resolved {@link StreamOptions.apiKey} string.
+	 */
+	apiKey?: ApiKey;
+	/**
+	 * Legacy static-key retry hook. New callers should pass an {@link ApiKeyResolver}
+	 * as `apiKey`; `streamSimple` adapts this hook into that resolver shape for
+	 * older Prometheus runtime paths.
+	 */
+	onAuthError?: (provider: string, oldKey: string, error: unknown) => Promise<string | undefined> | string | undefined;
 	reasoning?: Effort;
 	/**
 	 * Force-disable reasoning for the request even when the model supports it.
@@ -497,6 +417,12 @@ export interface ImageContent {
 	type: "image";
 	data: string; // base64 encoded image data
 	mimeType: string; // e.g., "image/jpeg", "image/png"
+	/**
+	 * OpenAI-only resolution hint. `"original"` preserves native resolution
+	 * (required for snapcompact frames, whose glyphs do not survive the
+	 * default `auto` downscale). Providers without a detail knob ignore it.
+	 */
+	detail?: "auto" | "low" | "high" | "original";
 }
 
 export interface ToolCall {
@@ -516,53 +442,6 @@ 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 {
@@ -605,8 +484,17 @@ export interface AssistantMessage {
 	provider: Provider;
 	model: string;
 	responseId?: string; // Provider-specific response/message identifier when the upstream API exposes one
+	/**
+	 * Name of the upstream provider an aggregator routed this request to, as
+	 * reported in the response (e.g. OpenRouter's top-level `provider` field:
+	 * `"OpenAI"`, `"Anthropic"`, `"Together"`). Distinct from `provider`, which
+	 * is the configured gateway we called (`"openrouter"`). Undefined for direct
+	 * providers that expose no such field.
+	 */
+	upstreamProvider?: string;
 	usage: Usage;
 	stopReason: StopReason;
+	stopDetails?: StopDetails | null;
 	errorMessage?: string;
 	/** HTTP status surfaced by the provider when the request failed. Populated by every provider's catch block alongside `errorMessage` so consumers (auth retry, telemetry, UI) can branch without regex-scraping the message. */
 	errorStatus?: number;
@@ -636,6 +524,12 @@ export interface ToolResultMessage<TDetails = any> {
 	attribution?: MessageAttribution;
 	/** Timestamp when output was pruned (ms since epoch). Undefined if unpruned. */
 	prunedAt?: number;
+	/**
+	 * Tool-declared: this result carried no information worth retaining once
+	 * consumed (zero matches, elapsed wait). Compaction passes may elide it.
+	 * Never set together with isError.
+	 */
+	useless?: boolean;
 	timestamp: number; // Unix timestamp in milliseconds
 }
 
@@ -748,218 +642,3 @@ 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 Prometheus 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+ 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;
-}
-
-/**
- * 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 Prometheus cannot discover — sending the wrong value triggers 400s
-	 * from the upstream provider.
-	 */
-	omitMaxOutputTokens?: boolean;
-	headers?: Record<string, string>;
-	/**
-	 * Streaming transport override. When `"prometheus-native"`, `streamSimple` routes
-	 * the request to the model's `baseUrl` via the auth-gateway's
-	 * `POST /v1/prometheus/stream` endpoint instead of dispatching the per-API
-	 * provider client. The `baseUrl` must point at an `prometheus auth-gateway`
-	 * (or compatible) host; `headers.Authorization` (or `apiKey` resolved by
-	 * the registry) carries the gateway bearer.
-	 *
-	 * Used by containerized prometheus installs (e.g. roboprometheus 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?: "prometheus-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;
-}

package/src/usage/claude.ts

@@ -1,4 +1,6 @@
 import { scheduler } from "node:timers/promises";
+import { toNumber } from "@prometheus-ai/catalog/utils";
+import { claudeCodeVersion } from "../providers/anthropic";
 import type {
 	CredentialRankingStrategy,
 	UsageAmount,
@@ -10,7 +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;
@@ -24,7 +26,7 @@ const CLAUDE_HEADERS = {
 	"anthropic-beta":
 		"claude-code-20250219,oauth-2025-04-20,interleaved-thinking-2025-05-14,redact-thinking-2026-02-12,context-management-2025-06-27,prompt-caching-scope-2026-01-05,mid-conversation-system-2026-04-07,advanced-tool-use-2025-11-20,effort-2025-11-24,extended-cache-ttl-2025-04-11",
 	"content-type": "application/json",
-	"user-agent": "claude-cli/2.1.160 (external, cli)",
+	"user-agent": `claude-cli/${claudeCodeVersion} (external, cli)`,
 	connection: "keep-alive",
 } as const;
 

package/src/usage/github-copilot.ts

@@ -3,6 +3,9 @@
  *
  * Normalizes Copilot quota usage into the shared UsageReport schema.
  */
+
+import { toBoolean, toNumber } from "@prometheus-ai/catalog/utils";
+import { OPENCODE_HEADERS } from "@prometheus-ai/catalog/wire/github-copilot";
 import type {
 	UsageAmount,
 	UsageFetchContext,
@@ -13,8 +16,7 @@ import type {
 	UsageStatus,
 	UsageWindow,
 } from "../usage";
-import { isRecord, toBoolean, toNumber } from "../utils";
-import { OPENCODE_HEADERS } from "../utils/oauth/github-copilot";
+import { isRecord } from "../utils";
 
 type CopilotQuotaDetail = {
 	entitlement: number;