@oh-my-pi/pi-ai 13.7.6 → 13.9.1

package/CHANGELOG.md

@@ -2,7 +2,54 @@
 
 ## [Unreleased]
 
+## [13.9.1] - 2026-03-05
+### Breaking Changes
+
+- Removed `THINKING_LEVELS`, `ALL_THINKING_LEVELS`, `ALL_THINKING_MODES`, `THINKING_MODE_DESCRIPTIONS`, and `THINKING_MODE_LABELS` exports
+- Renamed `formatThinking()` to `getThinkingMetadata()` with changed return type from string to `ThinkingMetadata` object
+- Renamed `getAvailableThinkingLevel()` to `getAvailableThinkingLevels()` and added default parameter
+- Renamed `getAvailableThinkingEffort()` to `getAvailableThinkingEfforts()` and added default parameter
+
+### Added
+
+- Added `ThinkingMetadata` type to provide structured access to thinking mode information (value, label, description)
+
+## [13.9.0] - 2026-03-05
+### Added
+
+- Exported new thinking module with `ThinkingEffort`, `ThinkingLevel`, and `ThinkingMode` types for managing reasoning effort levels
+- Added `getAvailableThinkingEffort()` function to determine supported thinking effort levels based on model capabilities
+- Added `parseThinkingEffort()`, `parseThinkingLevel()`, and `parseThinkingMode()` functions for parsing thinking configuration strings
+- Added `THINKING_LEVELS`, `ALL_THINKING_LEVELS`, and `ALL_THINKING_MODES` constants for iterating over available thinking options
+- Added `THINKING_MODE_DESCRIPTIONS` and `THINKING_MODE_LABELS` for displaying thinking modes in user interfaces
+- Added `formatThinking()` function to format thinking modes as compact display labels
+
+### Changed
+
+- Refactored thinking level handling to distinguish between `ThinkingEffort` (provider-level, no "off") and `ThinkingLevel` (user-facing, includes "off")
+- Updated `ThinkingBudgets` type to use `ThinkingEffort` instead of `ThinkingLevel` for more precise token budget configuration
+- Improved reasoning option handling to explicitly support "off" value for disabling reasoning across all providers
+- Simplified thinking effort mapping logic by centralizing provider-specific clamping behavior
+
+## [13.7.8] - 2026-03-04
+
+### Added
+
+- Added ZenMux provider support with mixed API routing: Anthropic-owned models discovered from `https://zenmux.ai/api/v1/models` now use the Anthropic transport (`https://zenmux.ai/api/anthropic`), while other ZenMux models use the OpenAI-compatible transport.
+
+## [13.7.7] - 2026-03-04
+
+### Changed
+
+- Modified response ID normalization to preserve existing item ID prefixes when truncating oversized IDs
+- Updated tool call ID normalization to use `fc_` prefix for generated item IDs instead of `item_` prefix
+
+### Fixed
+
+- Fixed handling of reasoning item IDs to remain untouched during response normalization while function call IDs are properly normalized
+
 ## [13.7.2] - 2026-03-04
+
 ### Added
 
 - Added support for Kagi API key authentication via `login kagi` command
@@ -15,6 +62,7 @@
 - Tool schema compilation is now cached per schema identity, eliminating redundant recompilation on every tool call
 
 ## [13.6.0] - 2026-03-03
+
 ### Added
 
 - Added Anthropic Foundry gateway mode controlled by `CLAUDE_CODE_USE_FOUNDRY`, with support for `FOUNDRY_BASE_URL`, `ANTHROPIC_FOUNDRY_API_KEY`, `ANTHROPIC_CUSTOM_HEADERS`, and optional mTLS material (`CLAUDE_CODE_CLIENT_CERT`, `CLAUDE_CODE_CLIENT_KEY`, `NODE_EXTRA_CA_CERTS`)
@@ -27,6 +75,7 @@
 - Anthropic auth base-URL fallback now prefers `FOUNDRY_BASE_URL` when `CLAUDE_CODE_USE_FOUNDRY` is enabled
 
 ## [13.5.8] - 2026-03-02
+
 ### Fixed
 
 - Fixed schema compatibility issue where patternProperties in tool parameters caused failures when converting to legacy Antigravity format
@@ -43,6 +92,7 @@
 - Anthropic cache-control normalization now removes later `ttl: "1h"` entries when a default/5m block has already appeared earlier in evaluation order
 
 ## [13.5.3] - 2026-03-01
+
 ### Fixed
 
 - Fixed tool argument coercion to handle malformed JSON with trailing wrapper braces by parsing leading JSON containers

package/README.md

@@ -68,6 +68,7 @@ Unified LLM API with automatic model discovery, provider configuration, token an
 - **zAI** (requires `ZAI_API_KEY`)
 - **MiniMax Coding Plan** (requires `MINIMAX_CODE_API_KEY` or `MINIMAX_CODE_CN_API_KEY`)
 - **Xiaomi MiMo** (requires `XIAOMI_API_KEY`)
+- **ZenMux** (requires `ZENMUX_API_KEY`)
 - **Qwen Portal** (supports `QWEN_OAUTH_TOKEN` or `QWEN_PORTAL_API_KEY`)
 - **Cloudflare AI Gateway** (requires `CLOUDFLARE_AI_GATEWAY_API_KEY` and provider-specific gateway base URL)
 - **Ollama** (local OpenAI-compatible runtime; optional `OLLAMA_API_KEY`)
@@ -929,6 +930,7 @@ In Node.js environments, you can set environment variables to avoid passing API
 | zAI            | `ZAI_API_KEY`                                                                |
 | MiniMax Code   | `MINIMAX_CODE_API_KEY` (international) or `MINIMAX_CODE_CN_API_KEY` (China) |
 | Xiaomi MiMo    | `XIAOMI_API_KEY`                                                             |
+| ZenMux         | `ZENMUX_API_KEY`                                                             |
 | vLLM           | `VLLM_API_KEY`                                                               |
 | Cloudflare AI Gateway | `CLOUDFLARE_AI_GATEWAY_API_KEY`                                      |
 | GitHub Copilot | `COPILOT_GITHUB_TOKEN` or `GH_TOKEN` or `GITHUB_TOKEN`                      |
@@ -950,6 +952,8 @@ Provider endpoint defaults for the current OpenAI-compatible integrations:
 - Hugging Face Inference: `https://router.huggingface.co/v1`
 - Venice: `https://api.venice.ai/api/v1`
 - Xiaomi MiMo: `https://api.xiaomimimo.com/anthropic`
+- ZenMux (OpenAI): `https://zenmux.ai/api/v1`
+- ZenMux (Anthropic models): `https://zenmux.ai/api/anthropic`
 - vLLM: `http://127.0.0.1:8000/v1`
 - Ollama: local OpenAI-compatible runtime
 - LiteLLM: `http://localhost:4000/v1`

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-ai",
-	"version": "13.7.6",
+	"version": "13.9.1",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://github.com/can1357/oh-my-pi",
 	"author": "Can Boluk",
@@ -41,7 +41,7 @@
 		"@aws-sdk/client-bedrock-runtime": "^3",
 		"@bufbuild/protobuf": "^2.11",
 		"@google/genai": "^1.43",
-		"@oh-my-pi/pi-utils": "13.7.6",
+		"@oh-my-pi/pi-utils": "13.9.1",
 		"@sinclair/typebox": "^0.34",
 		"@smithy/node-http-handler": "^4.4",
 		"ajv": "^8.18",

package/src/auth-storage.ts

@@ -64,6 +64,7 @@ import { loginVenice } from "./utils/oauth/venice";
 import { loginVllm } from "./utils/oauth/vllm";
 import { loginXiaomi } from "./utils/oauth/xiaomi";
 import { loginZai } from "./utils/oauth/zai";
+import { loginZenMux } from "./utils/oauth/zenmux";
 
 // ─────────────────────────────────────────────────────────────────────────────
 // Credential Types
@@ -488,12 +489,7 @@ export class AuthStorage {
 	 */
 	#getHashedIndex(sessionId: string, total: number): number {
 		if (total <= 1) return 0;
-		let hash = 2166136261; // FNV offset basis
-		for (let i = 0; i < sessionId.length; i++) {
-			hash ^= sessionId.charCodeAt(i);
-			hash = Math.imul(hash, 16777619); // FNV prime
-		}
-		return (hash >>> 0) % total;
+		return Bun.hash.xxHash32(sessionId) % total;
 	}
 
 	/**
@@ -920,6 +916,11 @@ export class AuthStorage {
 				await saveApiKeyCredential(apiKey);
 				return;
 			}
+			case "zenmux": {
+				const apiKey = await loginZenMux(ctrl);
+				await saveApiKeyCredential(apiKey);
+				return;
+			}
 			default: {
 				const customProvider = getOAuthProvider(provider);
 				if (!customProvider) {
@@ -1490,12 +1491,33 @@ export class AuthStorage {
 		const order = this.#getCredentialOrder(providerKey, sessionId, credentials.length);
 		const strategy = this.#rankingStrategyResolver?.(provider);
 		const checkUsage = strategy !== undefined && credentials.length > 1;
-		const candidates = checkUsage
-			? await this.#rankOAuthSelections({ providerKey, provider, order, credentials, options, strategy })
+		const sessionCredential = this.#getSessionCredential(provider, sessionId);
+		const sessionPreferredIndex = sessionCredential?.type === "oauth" ? sessionCredential.index : undefined;
+		// Skip ranking only when the session already has a working preferred credential — re-ranking
+		// mid-session causes account switches that cold-start the server-side prompt cache. New sessions
+		// (no preference) and sessions whose preferred is blocked still rank, so we pick the account
+		// with the most headroom proactively and fall back intelligently when rate-limited.
+		const sessionPreferredIsAvailable =
+			sessionPreferredIndex !== undefined && !this.#isCredentialBlocked(providerKey, sessionPreferredIndex);
+		const shouldRank = checkUsage && !sessionPreferredIsAvailable;
+		const candidates = shouldRank
+			? await this.#rankOAuthSelections({ providerKey, provider, order, credentials, options, strategy: strategy! })
 			: order
 					.map(idx => credentials[idx])
 					.filter((selection): selection is { credential: OAuthCredential; index: number } => Boolean(selection))
 					.map(selection => ({ selection, usage: null, usageChecked: false }));
+
+		if (sessionPreferredIndex !== undefined) {
+			const sessionPreferredCandidate = candidates.findIndex(
+				candidate =>
+					!this.#isCredentialBlocked(providerKey, candidate.selection.index) &&
+					candidate.selection.index === sessionPreferredIndex,
+			);
+			if (sessionPreferredCandidate > 0) {
+				const [preferred] = candidates.splice(sessionPreferredCandidate, 1);
+				candidates.unshift(preferred);
+			}
+		}
 		const fallback = candidates[0];
 
 		for (const candidate of candidates) {

package/src/cli.ts

@@ -15,6 +15,7 @@ import { loginNanoGPT } from "./utils/oauth/nanogpt";
 import { loginOpenAICodex } from "./utils/oauth/openai-codex";
 import type { OAuthCredentials, OAuthProvider } from "./utils/oauth/types";
 import { loginZai } from "./utils/oauth/zai";
+import { loginZenMux } from "./utils/oauth/zenmux";
 
 const PROVIDERS = getOAuthProviders();
 
@@ -220,6 +221,22 @@ async function login(provider: OAuthProvider): Promise<void> {
 				return;
 			}
 
+			case "zenmux": {
+				const apiKey = await loginZenMux({
+					onAuth(info) {
+						const { url, instructions } = info;
+						console.log(`\nOpen this URL in your browser:\n${url}`);
+						if (instructions) console.log(instructions);
+						console.log();
+					},
+					onPrompt(p) {
+						return promptFn(`${p.message}${p.placeholder ? ` (${p.placeholder})` : ""}:`);
+					},
+				});
+				storage.saveApiKey(provider, apiKey);
+				console.log(`\nAPI key saved to ~/.omp/agent/agent.db`);
+				return;
+			}
 			case "minimax-code": {
 				const apiKey = await loginMiniMaxCode({
 					onAuth(info) {
@@ -294,6 +311,7 @@ Providers:
   minimax-code      MiniMax Coding Plan (International)
   minimax-code-cn   MiniMax Coding Plan (China)
   cursor            Cursor (Claude, GPT, etc.)
+  zenmux            ZenMux
 
 Examples:
   bunx @oh-my-pi/pi-ai login              # interactive provider selection

package/src/index.ts

@@ -21,6 +21,7 @@ export * from "./providers/openai-responses";
 export * from "./providers/synthetic";
 export * from "./rate-limit-utils";
 export * from "./stream";
+export * from "./thinking";
 export * from "./types";
 export * from "./usage";
 export * from "./usage/claude";