claude-overnight 1.25.38 → 1.25.39

package/dist/coach.js

@@ -5,9 +5,10 @@ import { execSync } from "child_process";
 import { homedir } from "os";
 import chalk from "chalk";
 import { runPlannerQuery, attemptJsonParse } from "./planner-query.js";
+import { renderWaitingIndicator } from "./render.js";
 import { createTurn, beginTurn, endTurn } from "./turns.js";
 import { selectKey, ask } from "./cli.js";
-import { envFor } from "./providers.js";
+import { envFor, isCursorProxyProvider, ensureCursorProxyRunning, PROXY_DEFAULT_URL } from "./providers.js";
 // ── URL fetching for plan links in the objective ──
 const URL_REGEX = /https?:\/\/[^\s<>"{}|\\^`\[\]]+/g;
 async function fetchUrlContent(url, timeoutMs = 5_000) {
@@ -56,7 +57,7 @@ export function saveUserSettings(s) {
 }
 // ── Coach model (separate from DEFAULT_MODEL so the coach can stay cheap) ──
 export const COACH_MODEL = "claude-haiku-4-5";
-const COACH_TIMEOUT_MS = 15_000;
+const COACH_TIMEOUT_MS = 60_000;
 const COACH_SOFT_STATUS_MS = 5_000;
 // ── Raw schema matching the SKILL.md invocation contract ──
 const COACH_SCHEMA = {
@@ -342,18 +343,31 @@ export async function runSetupCoach(rawObjective, cwd, ctx) {
     }
     const userMessage = renderRepoFacts(facts, rawObjective, ctx.providers, ctx.cliFlags, planContent);
     const prompt = `${skill}\n\n---\n\n${userMessage}\n\nRespond with the JSON object defined in "Invocation contract" only.`;
-    const model = ctx.coachModel ?? COACH_MODEL;
+    // cursor "auto" maps to a slow thinking-class model for large prompts (182s observed).
+    // composer-2-fast gives the same quality for structured JSON at ~8s.
+    const CURSOR_FAST_MODEL = "composer-2-fast";
+    let model = ctx.coachModel ?? COACH_MODEL;
     const startedAt = Date.now();
     const spinner = setInterval(() => {
-        const elapsed = Math.round((Date.now() - startedAt) / 1000);
-        const hint = elapsed >= Math.round(COACH_SOFT_STATUS_MS / 1000) ? " · still thinking…" : "";
-        process.stdout.write(`\x1B[2K\r  ${chalk.cyan("⚡")} ${chalk.dim(`coach ${elapsed}s${hint}`)}`);
-    }, 500);
+        const indicator = renderWaitingIndicator("coach", startedAt, { style: "thinking" });
+        process.stdout.write(`\x1B[2K\r  ${indicator}`);
+    }, 120);
+    if (ctx.coachProvider && isCursorProxyProvider(ctx.coachProvider)) {
+        const proxyUrl = ctx.coachProvider.baseURL || PROXY_DEFAULT_URL;
+        const proxyUp = await ensureCursorProxyRunning(proxyUrl);
+        if (!proxyUp) {
+            clearInterval(spinner);
+            process.stdout.write(`\x1B[2K\r`);
+            console.log(chalk.dim("  coach skipped: proxy failed to start"));
+            return null;
+        }
+        if (model === "auto")
+            model = CURSOR_FAST_MODEL;
+    }
     let raw;
     const turn = createTurn("coach", "Coach", "coach-0", model);
     beginTurn(turn);
     try {
-        const coachEnv = ctx.coachProvider ? envFor(ctx.coachProvider) : undefined;
         const queryPromise = runPlannerQuery(prompt, {
             cwd,
             model,
@@ -362,7 +376,7 @@ export async function runSetupCoach(rawObjective, cwd, ctx) {
             transcriptName: "coach",
             maxTurns: 3,
             tools: [],
-            env: coachEnv,
+            env: ctx.coachProvider ? envFor(ctx.coachProvider) : undefined,
             turnId: turn.id,
         }, () => { });
         const timeout = new Promise((_, reject) => {

package/dist/planner-query.js

@@ -81,7 +81,49 @@ async function throttlePlanner(onLog, aborted) {
 const NUDGE_MS = 15 * 60 * 1000;
 const HARD_TIMEOUT_MS = 30 * 60 * 1000;
 const WALL_CLOCK_LIMIT_MS = 45 * 60 * 1000;
+// ── Cursor proxy: direct HTTP bypass ──
+//
+// When the env routes to a cursor proxy (CURSOR_API_KEY present, no ANTHROPIC_API_KEY),
+// the claude-agent-sdk wrapper is harmful, not helpful:
+//   - The SDK spawns a `claude` subprocess that makes 4+ sequential HTTP calls to the proxy.
+//   - Each call spawns a fresh cursor-agent subprocess (~15s overhead each).
+//   - Total: 4 × 15s = 60s for what should be a single 10-15s completion.
+// The SDK features (local tool loop, session resume, rate-limit headers) do not apply:
+//   - cursor-agent runs its own internal tool loop; local tool_use never fires.
+//   - cursor proxy doesn't expose session IDs or rate-limit headers.
+// One direct POST is always correct and 4-10× faster.
+function isCursorProxyEnv(env) {
+    return !!env?.CURSOR_API_KEY && !env?.ANTHROPIC_API_KEY;
+}
+async function runViaDirectFetch(prompt, opts, onLog) {
+    const env = opts.env ?? _envResolver?.(opts.model);
+    const baseUrl = (env?.ANTHROPIC_BASE_URL ?? "http://127.0.0.1:8765").replace(/\/$/, "");
+    const authToken = env?.ANTHROPIC_AUTH_TOKEN ?? "";
+    const MAX_RETRIES = 3;
+    const BACKOFF = [30_000, 60_000, 120_000];
+    for (let attempt = 0; attempt <= MAX_RETRIES; attempt++) {
+        const res = await fetch(`${baseUrl}/v1/messages`, {
+            method: "POST",
+            headers: { "Content-Type": "application/json", "Authorization": `Bearer ${authToken}` },
+            body: JSON.stringify({ model: opts.model, max_tokens: 8192, messages: [{ role: "user", content: prompt }] }),
+        });
+        if (res.status === 429 && attempt < MAX_RETRIES) {
+            const waitMs = BACKOFF[attempt];
+            onLog(`Cursor proxy rate limited — waiting ${Math.round(waitMs / 1000)}s`, "event");
+            await new Promise(r => setTimeout(r, waitMs));
+            continue;
+        }
+        if (!res.ok)
+            throw new Error(`Cursor proxy ${res.status}: ${(await res.text().catch(() => ""))}`);
+        const data = await res.json();
+        return data.content?.[0]?.text ?? "";
+    }
+    throw new Error("Cursor proxy direct fetch failed after retries");
+}
 export async function runPlannerQuery(prompt, opts, onLog) {
+    const env = opts.env ?? _envResolver?.(opts.model);
+    if (isCursorProxyEnv(env))
+        return runViaDirectFetch(prompt, opts, onLog);
     const MAX_RETRIES = 3;
     const BACKOFF = [30_000, 60_000, 120_000];
     let currentPrompt = prompt;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-overnight",
-  "version": "1.25.38",
+  "version": "1.25.39",
   "description": "Parallel Claude agents in git worktrees with a usage cap that reserves headroom for your interactive Claude Code. Crash-safe resume. Provider-agnostic model catalog (Anthropic, Cursor, OpenAI, Gemini, DeepSeek, Llama, Qwen) with capability-based task scoping.",
   "type": "module",
   "bin": {

package/plugins/claude-overnight/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-overnight",
-  "version": "1.25.38",
+  "version": "1.25.39",
   "description": "Claude Code skill for understanding, installing, and inspecting claude-overnight runs  -- parallel Claude agents in git worktrees with thinking waves, multi-wave steering, and crash-safe resume. Supports Cursor API Proxy, Qwen, OpenRouter.",
   "author": {
     "name": "Francesco Fornace"