pleumrouter 0.5.3 → 0.5.5

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pleumrouter",
-  "version": "0.5.3",
+  "version": "0.5.5",
   "description": "Official PleumRouter terminal CLI — launch any coding agent against PleumRouter.",
   "type": "module",
   "bin": {

package/src/models.mjs

@@ -128,7 +128,20 @@ export function renderModelInfo(m, styler = style) {
     ["max output", fmtCtx(m.max_output_tokens || m.max_output)],
     [t("price/1M"), `${t("in")} ${pin} · ${t("out")} ${pout} ${styler.dim(t("(incl. {pct}% markup)", { pct: m.effective_markup_percent ?? 0 }))}`],
   ];
-  if (m.supports_thinking) facts.push(["thinking", m.thinking_forced ? t("forced (can't be turned off)") : t("supported")]);
+  if (m.supports_thinking) {
+    facts.push([
+      "thinking",
+      m.thinking_forced
+        ? t("forced (can't be turned off)")
+        : t("supported (reasoning_effort: minimal|none|low|medium|high|xhigh|max)"),
+    ]);
+  }
+  if (Array.isArray(m.supported_parameters) && m.supported_parameters.includes("service_tier")) {
+    facts.push(["service_tier", "auto|default|flex|background|priority"]);
+  }
+  if (Array.isArray(m.supported_parameters) && m.supported_parameters.includes("reasoning_mode")) {
+    facts.push(["reasoning_mode", "standard|pro"]);
+  }
   for (const [k, v] of facts) lines.push(`${styler.dim(padDisplay(k, 10))} ${v}`);
 
   const u = m.usage;

package/src/recipes.mjs

@@ -16,6 +16,11 @@ export const RECIPES = {
     bin: "claude",
     type: "env",
     install: "npm i -g @anthropic-ai/claude-code",
+    // 앱 랭킹 귀속 메타 — 백엔드 featured_agents / UA fallback과 동기화.
+    attribution: {
+      title: "Claude Code",
+      referer: "https://docs.anthropic.com/en/docs/claude-code",
+    },
     // 함정: ANTHROPIC_BASE_URL은 반드시 루트(/v1 없이) — CLI가 /v1/messages를 스스로 덧붙인다.
     // /v1을 넣으면 /v1/v1/messages가 되어 실패.
     env: (key) => ({
@@ -30,6 +35,10 @@ export const RECIPES = {
     bin: "codex",
     type: "codex-home",
     install: "npm i -g @openai/codex",
+    attribution: {
+      title: "Codex",
+      referer: "https://github.com/openai/codex",
+    },
     env: (key) => ({ PLEUM_API_KEY: key }),
     args: () => [],
   },
@@ -38,6 +47,10 @@ export const RECIPES = {
     bin: "aider",
     type: "env",
     install: "python -m pip install aider-install && aider-install",
+    attribution: {
+      title: "Aider",
+      referer: "https://aider.chat",
+    },
     env: (key) => ({ OPENAI_API_BASE: V1, OPENAI_API_KEY: key }),
     args: (m) => (m ? ["--model", `openai/${m}`] : []),
   },
@@ -47,6 +60,10 @@ export const RECIPES = {
     type: "env",
     install: "https://block.github.io/goose/docs/getting-started/installation",
     note: "Goose sets its model in its own config — use the openai/<id> format in `goose configure`.",
+    attribution: {
+      title: "Goose",
+      referer: "https://block.github.io/goose",
+    },
     env: (key) => ({ OPENAI_API_BASE: V1, OPENAI_API_KEY: key }),
     args: () => [],
   },
@@ -56,6 +73,10 @@ export const RECIPES = {
     type: "env",
     install: "python -m pip install openhands-ai",
     note: "For OpenHands, prefix the model name with openai/ in its config.",
+    attribution: {
+      title: "OpenHands",
+      referer: "https://www.all-hands.dev",
+    },
     env: (key) => ({ OPENAI_API_BASE: V1, OPENAI_API_KEY: key }),
     args: () => [],
   },
@@ -67,6 +88,10 @@ export const RECIPES = {
     // opencode는 커스텀 엔드포인트를 opencode.json provider 블록으로만 읽는다 — env var는 없다.
     // 대신 OPENCODE_CONFIG_CONTENT(파일 대신 JSON 문자열을 그대로 읽는 env var)로 파일 없이 주입한다.
     note: "Generates a temporary PleumRouter provider config for this run only — your own opencode.json is untouched.",
+    attribution: {
+      title: "opencode",
+      referer: "https://opencode.ai",
+    },
     env: (key, model) => ({
       PLEUM_API_KEY: key,
       OPENCODE_CONFIG_CONTENT: JSON.stringify({
@@ -90,6 +115,10 @@ export const RECIPES = {
     // omp는 커스텀 OpenAI 호환 엔드포인트는 YAML 설정 파일로만 읽지만, 내장 anthropic
     // provider만은 claude와 동일한 ANTHROPIC_BASE_URL env var로 리디렉션된다.
     note: "Needs the Bun runtime (bun >= 1.3.14) on PATH. Only Claude-family models are routed via env vars.",
+    attribution: {
+      title: "omp (Oh My Pi)",
+      referer: "https://omp.sh",
+    },
     env: (key) => ({ ANTHROPIC_BASE_URL: ROOT, ANTHROPIC_API_KEY: key }),
     args: (m) => (m ? ["--model", `anthropic/${m}`] : []),
   },
@@ -99,6 +128,10 @@ export const RECIPES = {
     type: "env",
     install: "curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash",
     note: "Pick the model after it starts with the `hermes model` command.",
+    attribution: {
+      title: "Hermes Agent",
+      referer: "https://hermes-agent.nousresearch.com",
+    },
     env: (key) => ({ OPENAI_BASE_URL: V1, OPENAI_API_KEY: key }),
     args: () => [],
   },
@@ -107,6 +140,10 @@ export const RECIPES = {
     bin: "qwen",
     type: "env",
     install: "npm install -g @qwen-code/qwen-code@latest",
+    attribution: {
+      title: "Qwen Code",
+      referer: "https://github.com/QwenLM/qwen-code",
+    },
     env: (key, model) => ({ OPENAI_BASE_URL: V1, OPENAI_API_KEY: key, OPENAI_MODEL: model || "gpt-5.4-mini" }),
     args: () => [],
   },
@@ -116,6 +153,10 @@ export const RECIPES = {
     type: "env",
     install: "npm install -g @github/copilot",
     note: "Needs GitHub Copilot CLI 2026-04 or later for BYOK provider routing support.",
+    attribution: {
+      title: "GitHub Copilot CLI",
+      referer: "https://docs.github.com/en/copilot/how-tos/use-copilot-agents/use-copilot-cli",
+    },
     env: (key, model) => ({
       COPILOT_PROVIDER_BASE_URL: V1,
       COPILOT_PROVIDER_TYPE: "openai",

package/src/run.mjs

@@ -10,12 +10,20 @@ import { t } from "./i18n.mjs";
 
 const IMG_EXT = /\.(png|jpe?g|gif|webp|bmp)$/i;
 
-// `pleum run [model] ["프롬프트"...] [--verbose] [--skip-cache] [--tag a,b]`
+// `pleum run [model] ["프롬프트"...] [--verbose] [--skip-cache] [--tag a,b]
+//   [--reasoning-effort LEVEL] [--service-tier TIER] [--reasoning-mode MODE]`
 // 인자 파싱(순수 함수 — 테스트 대상).
+const REASONING_EFFORTS = new Set(["minimal", "none", "low", "medium", "high", "xhigh", "max"]);
+const SERVICE_TIERS = new Set(["auto", "default", "flex", "background", "priority"]);
+const REASONING_MODES = new Set(["standard", "pro"]);
+
 export function parseRunArgs(rest = []) {
   let verbose = false;
   let skipCache = false;
   let tags = [];
+  let reasoningEffort = null;
+  let serviceTier = null;
+  let reasoningMode = null;
   const pos = [];
   for (let i = 0; i < rest.length; i++) {
     const a = rest[i];
@@ -27,9 +35,36 @@ export function parseRunArgs(rest = []) {
         tags = v.split(",").map((s) => s.trim()).filter(Boolean).slice(0, 10);
         i++;
       }
+    } else if (a === "--reasoning-effort") {
+      const v = rest[i + 1];
+      if (typeof v === "string" && REASONING_EFFORTS.has(v)) {
+        reasoningEffort = v;
+        i++;
+      }
+    } else if (a === "--service-tier") {
+      const v = rest[i + 1];
+      if (typeof v === "string" && SERVICE_TIERS.has(v)) {
+        serviceTier = v;
+        i++;
+      }
+    } else if (a === "--reasoning-mode") {
+      const v = rest[i + 1];
+      if (typeof v === "string" && REASONING_MODES.has(v)) {
+        reasoningMode = v;
+        i++;
+      }
     } else pos.push(a);
   }
-  return { model: pos[0], prompt: pos.slice(1).join(" "), verbose, skipCache, tags };
+  return {
+    model: pos[0],
+    prompt: pos.slice(1).join(" "),
+    verbose,
+    skipCache,
+    tags,
+    reasoningEffort,
+    serviceTier,
+    reasoningMode,
+  };
 }
 
 // 입력에서 존재하는 이미지 파일 경로를 찾아 멀티모달 content로 만든다(없으면 문자열 그대로).
@@ -70,11 +105,21 @@ function isAbort(e) {
 
 export async function run(rest) {
   const args = Array.isArray(rest) ? rest : rest == null ? [] : [rest];
-  const { verbose, prompt: inlinePrompt, model: argModel, skipCache, tags } = parseRunArgs(args);
+  const {
+    verbose,
+    prompt: inlinePrompt,
+    model: argModel,
+    skipCache,
+    tags,
+    reasoningEffort,
+    serviceTier,
+    reasoningMode,
+  } = parseRunArgs(args);
   let model = argModel;
   const piped = await readStdin();
   const oneShot = Boolean(inlinePrompt || piped);
   const key = requireKey();
+  const callOpts = { tags, skipCache, reasoningEffort, serviceTier, reasoningMode };
 
   if (!model && !oneShot) {
     if (!process.stdin.isTTY) {
@@ -100,7 +145,7 @@ export async function run(rest) {
     }
     const messages = [{ role: "user", content }];
     try {
-      await streamChat(key, model, messages, { verbose, quiet: true, tags, skipCache });
+      await streamChat(key, model, messages, { verbose, quiet: true, ...callOpts });
     } catch (e) {
       console.error(`${style.red("✗")} ${e.message}`);
       process.exit(1);
@@ -108,10 +153,16 @@ export async function run(rest) {
     process.exit(0);
   }
 
-  await repl(key, model, verbose, { tags, skipCache });
+  await repl(key, model, verbose, callOpts);
 }
 
-async function repl(key, initialModel, initialVerbose, { tags = [], skipCache = false } = {}) {
+async function repl(key, initialModel, initialVerbose, {
+  tags = [],
+  skipCache = false,
+  reasoningEffort = null,
+  serviceTier = null,
+  reasoningMode = null,
+} = {}) {
   let model = initialModel;
   let system = "";
   let verbose = initialVerbose;
@@ -128,7 +179,15 @@ async function repl(key, initialModel, initialVerbose, { tags = [], skipCache =
     const ac = new AbortController();
     currentAbort = ac;
     try {
-      const reply = await streamChat(key, model, req, { verbose, signal: ac.signal, tags, skipCache });
+      const reply = await streamChat(key, model, req, {
+        verbose,
+        signal: ac.signal,
+        tags,
+        skipCache,
+        reasoningEffort,
+        serviceTier,
+        reasoningMode,
+      });
       if (reply) messages.push({ role: "assistant", content: reply });
       else messages.pop(); // 응답 없음(즉시 중단 등) — user 턴 되돌림
     } catch (e) {
@@ -344,7 +403,17 @@ async function handleSlash(text, ctx) {
 
 // SSE 스트리밍 채팅. fetchImpl 주입 가능(테스트용 — 네트워크 없이 SSE 파싱 검증).
 // 비용은 마지막 SSE 청크의 cost.krw — 스트리밍 응답엔 X-Cost-Krw 헤더가 없다(백엔드 계약).
-export async function streamChat(key, model, messages, { verbose = false, quiet = false, signal, tags = [], skipCache = false, fetchImpl = fetch } = {}) {
+export async function streamChat(key, model, messages, {
+  verbose = false,
+  quiet = false,
+  signal,
+  tags = [],
+  skipCache = false,
+  reasoningEffort = null,
+  serviceTier = null,
+  reasoningMode = null,
+  fetchImpl = fetch,
+} = {}) {
   const t0 = Date.now();
   const sp = spinner(t("{model} thinking…", { model }));
   let res;
@@ -354,6 +423,9 @@ export async function streamChat(key, model, messages, { verbose = false, quiet
     if (skipCache) headers["X-Pleum-Skip-Cache"] = "1";
     const bodyObj = { model, messages, stream: true, stream_options: { include_usage: true } };
     if (tags.length > 0) bodyObj.tags = tags.slice(0, 10);
+    if (reasoningEffort) bodyObj.reasoning_effort = reasoningEffort;
+    if (serviceTier) bodyObj.service_tier = serviceTier;
+    if (reasoningMode) bodyObj.reasoning_mode = reasoningMode;
     res = await fetchImpl(`${V1}/chat/completions`, {
       method: "POST",
       headers,