@xsft/panels 0.3.2 → 0.3.3

package/README.md

@@ -31,7 +31,7 @@ Run the command in the folder of the product. Each judge starts in that folder,
 
 On a terminal, `panels run` shows the score table, and each judge runs in its interactive harness in a pseudo-terminal inside the command. Each judge writes `judges/N/answer.json` and stays open. The table writes the report when each judge has an answer or an error, and the command exits from `report.json` when you quit. Without a terminal, the judges run in the background (`claude -p`, `codex exec`).
 
-Keys in the table: `1` to `9` open the screen of that judge, and `q` quits and asks `y` while a judge still runs. In a judge, each key goes to the judge, and `ctrl-]` returns to the table.
+Keys in the table: `1` to `9` open the screen of that judge, and `q` quits and asks `y` while a judge still runs. In a judge, each key goes to the judge, and `ctrl-q` returns to the table. A harness that has not seen the folder asks whether to trust it. The first judge starts alone; press `1`, answer, and the other judges start.
 
 ## Run folder
 

package/lib/table.js

@@ -8,19 +8,23 @@ import xterm from '@xterm/headless';
 import { check, MINUTES_PER_CRITERION, parseAnswer, writeJudge } from './judge.js';
 import { buildReport, renderMarkdown } from './report.js';
 
-// More parallel sessions on one account get 429 answers from the model api.
-const LIVE_JUDGES = 4;
+const LIVE_JUDGES = 10;
+// A harness asks this in a folder it has not seen. Only the first judge starts until its screen no longer asks,
+// so the engineer answers once and the harness saves the answer for the other judges.
+const TRUST_PROMPT = /trust this folder/i;
+const FIRST_JUDGE_HEAD_START = 3000;
 const ORANGE = '\x1b[38;2;255;106;61m';
 const GREY = '\x1b[90m';
 const BOLD = '\x1b[1m';
 const RESET = '\x1b[0m';
-// ctrl-] returns from a judge to the table. A terminal in the kitty keyboard mode sends it as CSI 93;5u.
-const BACK = ['\x1d', '\x1b[93;5u'];
+// ctrl-q returns from a judge to the table. A terminal in the kitty keyboard mode sends it as CSI 113;5u.
+const BACK = ['\x11', '\x1b[113;5u'];
 // Turns off each mode that the screen of a judge can turn on, so the table gets the keys and the screen as before.
 const TABLE_MODES = '\x1b[?1049h\x1b[r\x1b[0m\x1b[4l\x1b[?6l\x1b[?7h\x1b[?1l\x1b>\x1b[?1000l\x1b[?1002l\x1b[?1003l\x1b[?1006l\x1b[?1004l\x1b[?2004l\x1b[<u\x1b[?25l\x1b[2J';
 
 /**
- * The lines of the table. rows hold { rank, name, state, result }, and state is queued, running, done, or error.
+ * The lines of the table. rows hold { rank, name, state, result, trust }, and state is queued, running, done, or error.
+ * trust is true while the harness of a running judge asks whether to trust the folder.
  * final is null until the report exists, then { status, path }. confirm is true while q waits for y.
  */
 const SPIN = ['⠋', '⠙', '⠹', '⠸', '⠼', '⠴', '⠦', '⠧', '⠇', '⠏'];
@@ -31,7 +35,8 @@ export function render({ title, rows, final, confirm, width, frame = 0 }) {
   const lines = [` ${BOLD}panels run${RESET}  ${title}`, rule];
   for (const r of rows) {
     const head = `${String(r.rank).padStart(3)}  ${r.name.padEnd(pad)}`;
-    if (r.state === 'running') lines.push(`${ORANGE}${head}${SPIN[frame % SPIN.length]}${RESET}`);
+    if (r.state === 'running' && r.trust) lines.push(`${ORANGE}${head}press ${r.rank} and answer the trust prompt${RESET}`);
+    else if (r.state === 'running') lines.push(`${ORANGE}${head}${SPIN[frame % SPIN.length]}${RESET}`);
     else if (r.state === 'queued') lines.push(`${head}${GREY}queued${RESET}`);
     else if (r.state === 'error') lines.push(`${head}error`);
     else lines.push(`${head}${r.result.score} ${GREY}b${r.result.band}${RESET}`);
@@ -80,7 +85,14 @@ export function runTable({ panel, runDir, harness, cwd, target, brief, sources }
     }
     if (judges[i] === undefined) return { state: 'queued' };
     if (Date.now() - judges[i].started > MINUTES_PER_CRITERION * 60_000) return { state: 'error', error: `no answer.json after ${MINUTES_PER_CRITERION} minutes` };
-    return { state: 'running' };
+    return { state: 'running', trust: asksTrust(judges[i]) };
+  }
+
+  function asksTrust(j) {
+    const buffer = j.term.buffer.active;
+    const lines = [];
+    for (let y = 0; y < buffer.length; y++) lines.push(buffer.getLine(y).translateToString(true));
+    return TRUST_PROMPT.test(lines.join(' '));
   }
 
   function start(rank) {
@@ -113,7 +125,7 @@ export function runTable({ panel, runDir, harness, cwd, target, brief, sources }
   function drawFooter() {
     if (!ui.view) return;
     const { cursorX, cursorY } = judges[ui.view - 1].term.buffer.active;
-    out.write(`\x1b[${out.rows};1H${RESET}${GREY} judge ${ui.view}   ctrl-] scores${RESET}\x1b[K\x1b[${cursorY + 1};${cursorX + 1}H`);
+    out.write(`\x1b[${out.rows};1H${RESET}${GREY} judge ${ui.view}   ctrl-q scores${RESET}\x1b[K\x1b[${cursorY + 1};${cursorX + 1}H`);
   }
 
   // Waits for a pause in the output, so the footer does not land inside an escape sequence that a chunk split.
@@ -153,7 +165,10 @@ export function runTable({ panel, runDir, harness, cwd, target, brief, sources }
     if (!final) {
       rows = criteria.map((c, i) => ({ ...c, ...state(c, i) }));
       const live = rows.filter((r) => r.state === 'running').length;
-      for (const r of rows.filter((x) => x.state === 'queued').slice(0, LIVE_JUDGES - live)) {
+      const first = judges[0];
+      const hold = first && !first.exited && (Date.now() - first.started < FIRST_JUDGE_HEAD_START || rows[0].trust);
+      for (const r of rows.filter((x) => x.state === 'queued').slice(0, first ? LIVE_JUDGES - live : 1)) {
+        if (hold) break;
         start(r.rank);
         r.state = 'running';
       }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xsft/panels",
-  "version": "0.3.2",
+  "version": "0.3.3",
   "private": false,
   "description": "Run a panel of AI judges from cross.software/panels against a target.",
   "type": "module",