@diegopetrucci/pi-oracle 0.1.12 → 0.1.14

package/index.ts

@@ -114,6 +114,7 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"claude-sonnet-4",
 		"claude-3-7-sonnet",
 	],
+	"ant-ling": ["Ling-2.6-1T", "Ling-2.6-flash"],
 	"azure-openai-responses": [
 		"gpt-5.5-pro",
 		"gpt-5.5",
@@ -146,6 +147,8 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"workers-ai/@cf/zai-org/glm-4.7-flash",
 	],
 	"cloudflare-workers-ai": [
+		"@cf/moonshotai/kimi-k2.7-code",
+		"@cf/zai-org/glm-5.2",
 		"@cf/moonshotai/kimi-k2.6",
 		"@cf/nvidia/nemotron-3-120b-a12b",
 		"@cf/moonshotai/kimi-k2.5",
@@ -155,9 +158,14 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 	deepseek: ["deepseek-v4-pro", "deepseek-v4-flash"],
 	fireworks: [
 		"accounts/fireworks/models/deepseek-v4-pro",
-		"accounts/fireworks/models/kimi-k2p6",
+		"accounts/fireworks/models/kimi-k2p7-code",
+		"accounts/fireworks/routers/kimi-k2p7-code-fast",
+		"accounts/fireworks/models/glm-5p2",
+		"accounts/fireworks/models/minimax-m3",
 		"accounts/fireworks/models/glm-5p1",
+		"accounts/fireworks/models/kimi-k2p6",
 		"accounts/fireworks/models/minimax-m2p7",
+		"accounts/fireworks/models/qwen3p7-plus",
 		"accounts/fireworks/models/qwen3p6-plus",
 		"accounts/fireworks/models/gpt-oss-120b",
 	],
@@ -217,9 +225,9 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"MiniMaxAI/MiniMax-M2.5",
 		"Qwen/Qwen3-Coder-Next",
 	],
-	"kimi-coding": ["k2p6", "kimi-k2-thinking", "kimi-for-coding"],
-	minimax: ["MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
-	"minimax-cn": ["MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
+	"kimi-coding": ["k2p7", "kimi-k2-thinking", "kimi-for-coding"],
+	minimax: ["MiniMax-M3", "MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
+	"minimax-cn": ["MiniMax-M3", "MiniMax-M2.7-highspeed", "MiniMax-M2.7"],
 	mistral: [
 		"mistral-medium-2604",
 		"mistral-medium-3.5",
@@ -285,11 +293,16 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 	],
 	"opencode-go": [
 		"deepseek-v4-pro",
+		"glm-5.2",
 		"glm-5.1",
+		"qwen3.7-max",
+		"qwen3.7-plus",
 		"qwen3.6-plus",
 		"mimo-v2.5-pro",
 		"mimo-v2-pro",
+		"minimax-m3",
 		"minimax-m2.7",
+		"kimi-k2.7-code",
 		"kimi-k2.6",
 		"kimi-k2.5",
 	],
@@ -310,20 +323,26 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"google/gemini-3.1-pro-preview-customtools",
 		"google/gemini-3.1-pro-preview",
 		"google/gemini-2.5-pro",
+		"moonshotai/kimi-k2.7-code",
 		"moonshotai/kimi-k2.6",
 		"moonshotai/kimi-k2-thinking",
 		"deepseek/deepseek-v4-pro",
 		"deepseek/deepseek-r1",
 		"deepseek/deepseek-v3.2",
+		"minimax/minimax-m3",
 		"minimax/minimax-m2.7",
 		"minimax/minimax-m2.1",
+		"z-ai/glm-5.2",
 		"z-ai/glm-5.1",
 	],
 	together: [
 		"deepseek-ai/DeepSeek-V4-Pro",
 		"zai-org/GLM-5.1",
+		"moonshotai/Kimi-K2.7-Code",
 		"moonshotai/Kimi-K2.6",
+		"Qwen/Qwen3.7-Max",
 		"Qwen/Qwen3.6-Plus",
+		"MiniMaxAI/MiniMax-M3",
 		"MiniMaxAI/MiniMax-M2.7",
 		"Qwen/Qwen3.5-397B-A17B",
 		"Qwen/Qwen3-Coder-Next-FP8",
@@ -347,15 +366,21 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"openai/gpt-5.4",
 		"openai/gpt-5.1-codex",
 		"openai/gpt-5-codex",
+		"moonshotai/kimi-k2.7-code",
+		"moonshotai/kimi-k2.7-code-highspeed",
 		"moonshotai/kimi-k2.6",
 		"moonshotai/kimi-k2-thinking",
 		"deepseek/deepseek-v4-pro",
 		"deepseek/deepseek-v3.2-thinking",
+		"alibaba/qwen3.7-max",
+		"alibaba/qwen3.7-plus",
 		"alibaba/qwen3.5-plus",
 		"alibaba/qwen3-max-thinking",
 		"google/gemini-3.1-pro-preview",
 		"google/gemini-3-flash",
 		"xai/grok-4.3",
+		"minimax/minimax-m3",
+		"zai/glm-5.2",
 		"zai/glm-5.1",
 	],
 	xai: [
@@ -390,7 +415,18 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"mimo-v2-omni",
 		"mimo-v2-flash",
 	],
+	nvidia: [
+		"nvidia/nemotron-3-ultra-550b-a55b",
+		"nvidia/nemotron-3-super-120b-a12b",
+		"moonshotai/kimi-k2.6",
+		"z-ai/glm-5.1",
+		"qwen/qwen3.5-122b-a10b",
+		"openai/gpt-oss-120b",
+		"nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
+		"nvidia/nemotron-3-nano-30b-a3b",
+	],
 	zai: [
+		"glm-5.2",
 		"glm-5.1",
 		"glm-5-turbo",
 		"glm-5v-turbo",
@@ -401,8 +437,17 @@ const PROVIDER_MODEL_PREFERENCES: Record<string, string[]> = {
 		"glm-4.5v",
 		"glm-4.5-air",
 	],
-	moonshotai: ["kimi-k2.6", "kimi-k2-thinking-turbo", "kimi-k2-thinking", "kimi-k2.5"],
-	"moonshotai-cn": ["kimi-k2.6", "kimi-k2-thinking-turbo", "kimi-k2-thinking", "kimi-k2.5"],
+	"zai-coding-cn": [
+		"glm-5.2",
+		"glm-5.1",
+		"glm-5-turbo",
+		"glm-5v-turbo",
+		"glm-5",
+		"glm-4.7",
+		"glm-4.5-air",
+	],
+	moonshotai: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2-thinking-turbo", "kimi-k2-thinking", "kimi-k2.5"],
+	"moonshotai-cn": ["kimi-k2.7-code", "kimi-k2.7-code-highspeed", "kimi-k2.6", "kimi-k2-thinking-turbo", "kimi-k2-thinking", "kimi-k2.5"],
 };
 
 const ORACLE_SYSTEM_PROMPT = [
@@ -557,6 +602,26 @@ function extractTextFromContent(content: unknown): string {
 	return parts.join("\n\n").trim();
 }
 
+// Errors that typically resolve on retry: provider overload, rate limiting,
+// transient 5xx, gateway/network failures, and request timeouts. Keep this list
+// pattern-based so we recognize variants across providers without enumerating them.
+const TRANSIENT_ERROR_PATTERN =
+	/\b(overload(?:ed)?|rate[ _-]?limit(?:ed)?|too many requests|429|500|502|503|504|bad gateway|service unavailable|gateway timeout|temporarily unavailable|timed?[ _-]?out|timeout|econnreset|econnrefused|etimedout|enetunreach|socket hang up|fetch failed)\b/i;
+
+function isTransientErrorMessage(message: string): boolean {
+	return TRANSIENT_ERROR_PATTERN.test(message);
+}
+
+function formatOracleModelError(stopReason: "error" | "aborted", errorMessage: string | undefined): string {
+	const trimmed = errorMessage?.trim();
+	if (stopReason === "aborted") {
+		return trimmed ? `Oracle model turn aborted: ${trimmed}` : "Oracle model turn aborted.";
+	}
+	if (!trimmed) return "Oracle model error (no detail provided by provider).";
+	const base = `Oracle model error: ${trimmed}`;
+	return isTransientErrorMessage(trimmed) ? `${base} (transient; retry may succeed)` : base;
+}
+
 function parseVersionScore(text: string): number {
 	const matches = text.match(/\d+(?:\.\d+){0,2}/g) ?? [];
 	let best = 0;
@@ -817,6 +882,9 @@ async function runOracle(
 	let finalOutput = "";
 	let stderr = "";
 
+	let lastStopReason: string | undefined;
+	let lastErrorMessage: string | undefined;
+
 	const details: OracleDetails = {
 		...selection,
 		includeBash,
@@ -888,6 +956,11 @@ async function runOracle(
 				if (text) finalOutput = text;
 				currentText = "";
 
+				const stopReason = event.message.stopReason;
+				lastStopReason = typeof stopReason === "string" ? stopReason : undefined;
+				const errorMessageField = event.message.errorMessage;
+				lastErrorMessage = typeof errorMessageField === "string" ? errorMessageField : undefined;
+
 				const messageUsage = event.message.usage;
 				if (messageUsage) {
 					usage.turns += 1;
@@ -946,6 +1019,19 @@ async function runOracle(
 		return { ok: false, error: "Oracle was aborted.", details };
 	}
 
+	// The pi subprocess in `--mode json` does not promote an errored assistant
+	// turn to a non-zero exit code or stderr; the error is only carried on the
+	// streamed assistant message via stopReason/errorMessage. Surface that here
+	// so callers can distinguish transient provider errors from a genuinely
+	// empty response.
+	if (lastStopReason === "error" || lastStopReason === "aborted") {
+		return {
+			ok: false,
+			error: formatOracleModelError(lastStopReason, lastErrorMessage),
+			details,
+		};
+	}
+
 	if (exitCode !== 0) {
 		return {
 			ok: false,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@diegopetrucci/pi-oracle",
-  "version": "0.1.12",
+  "version": "0.1.14",
   "description": "An Amp-style oracle extension for pi that consults the strongest reasoning model on your current provider.",
   "keywords": [
     "pi-package",