@lawrence369/loop-cli 0.1.0

package/dist/utils/pty-filter.js

@@ -0,0 +1,187 @@
+/**
+ * PTY output classification and filtering.
+ *
+ * Ported from iterloop's pty-session.ts — classifies raw terminal output lines
+ * into content, status updates, or ignorable TUI noise so that only meaningful
+ * text reaches the reviewer and transcript.
+ */
+// ── Status keywords ──────────────────────────────────
+const STATUS_KEYWORDS = [
+    "harmonizing",
+    "thinking",
+    "planning",
+    "generating",
+    "searching",
+    "analyzing",
+    "burrowing",
+    "contemplating",
+    "flibbertigibbeting",
+    "running stop hook",
+];
+// Note: ⏺ is NOT a spinner char — Claude CLI uses it as a content output marker
+const SPINNER_CHARS = /^[\s✳✶✻✽✢·⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏●○◆◇▪▸▹►☐☑✓✗✔✘⏵]/;
+// ── Line classifier ──────────────────────────────────
+/**
+ * Classify a single (ANSI-stripped) line of PTY output.
+ *
+ * - `"content"` — meaningful AI output to keep
+ * - `"status"`  — transient status / spinner text
+ * - `"ignore"`  — TUI chrome, empty lines, noise
+ */
+export function classifyLine(line, engine) {
+    const trimmed = line.trim();
+    if (!trimmed)
+        return "ignore";
+    // Claude CLI content marker: ⏺ followed by text = actual AI output
+    if (/^⏺/.test(trimmed)) {
+        return trimmed.length > 1 ? "content" : "ignore";
+    }
+    // Short fragments (< 5 chars) from TUI cursor repositioning are artifacts
+    if (trimmed.length < 5)
+        return "ignore";
+    // ── Known CLI UI elements to ignore ──
+    // Prompt marker (with or without text)
+    if (/^❯/.test(trimmed))
+        return "ignore";
+    // Permission mode indicator
+    if (/^⏵⏵/.test(trimmed))
+        return "ignore";
+    // Update notices
+    if (/update\s*avai|brew\s*upgrade/i.test(trimmed))
+        return "ignore";
+    // Keyboard hints
+    if (/shift\+tab/i.test(trimmed))
+        return "ignore";
+    // Mode indicator
+    if (/fast\s*mode/i.test(trimmed))
+        return "ignore";
+    // Truncated UI text
+    if (/^…\s/.test(trimmed) && trimmed.length < 80)
+        return "ignore";
+    // TUI box layout: lines inside boxes (start with │ or box corners)
+    if (/^[│╭╰]/.test(trimmed))
+        return "ignore";
+    // Status bar indicators (▪▪▪ pattern)
+    if (/▪▪▪/.test(trimmed))
+        return "ignore";
+    // Block element art (CLI logos, decorative)
+    if (/^[\s▐▛▜▝▘█▌▪·↯]+$/.test(trimmed))
+        return "ignore";
+    // Lines that are primarily horizontal rules/separators
+    if (/^[─═\s▪·↯]+$/.test(trimmed))
+        return "ignore";
+    // Box-drawing / UI chrome (pure box chars)
+    if (/^[\s─│┌┐└┘├┤┬┴┼═╔╗╚╝╠╣╦╩╬┃┗┛┏┓╭╮╰╯▐▛▜▝]+$/.test(trimmed))
+        return "ignore";
+    // Pure spinner characters
+    if (/^[\s✳✶✻✽✢·⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏]+$/.test(trimmed))
+        return "ignore";
+    // ── Status detection ──
+    const lower = trimmed.toLowerCase();
+    for (const kw of STATUS_KEYWORDS) {
+        if (lower.includes(kw))
+            return "status";
+    }
+    // Lines starting with spinner chars — only classify as status if it looks like
+    // a real spinner update (very short text after the spinner prefix)
+    if (SPINNER_CHARS.test(trimmed) && trimmed.length < 80) {
+        const withoutPrefix = trimmed.replace(/^[\s✳✶✻✽✢·⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏●○◆◇▪▸▹►☐☑✓✗✔✘⏵]+\s*/, "");
+        if (withoutPrefix.length > 0 && withoutPrefix.length < 60)
+            return "status";
+        if (withoutPrefix.length === 0)
+            return "ignore";
+    }
+    // For Claude engine: only ⏺-prefixed lines are true content.
+    // Everything else that reaches here is TUI rendering noise.
+    if (engine === "claude")
+        return "ignore";
+    return "content";
+}
+// ── Output filtering ─────────────────────────────────
+/**
+ * Filter raw PTY output to extract only meaningful content.
+ *
+ * Strips TUI chrome, spinners, box-drawing, update notices, and other noise.
+ * For Claude output, strips the ⏺ content marker prefix.
+ *
+ * Result is capped at ~50 KB (last 50 KB on overflow).
+ */
+export function filterOutput(raw, _engine) {
+    const lines = raw.split("\n");
+    const filtered = lines
+        .map((line) => {
+        // Strip ⏺ content marker prefix from Claude CLI output
+        const t = line.trim();
+        if (t.startsWith("⏺"))
+            return t.slice(1).trimStart();
+        return line;
+    })
+        .filter((line) => {
+        const trimmed = line.trim();
+        if (trimmed === "")
+            return true;
+        // Short TUI fragments
+        if (trimmed.length < 5)
+            return false;
+        if (/^[\s⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏●○◆◇✳✶✻✽✢·]+$/.test(trimmed))
+            return false;
+        if (/^[\s─│┌┐└┘├┤┬┴┼═╔╗╚╝╠╣╦╩╬┃┗┛┏┓╭╮╰╯▐▛▜▝]+$/.test(trimmed))
+            return false;
+        // Known UI noise
+        if (/^❯/.test(trimmed))
+            return false;
+        if (/^⏵⏵/.test(trimmed))
+            return false;
+        if (/^[│╭╰]/.test(trimmed))
+            return false;
+        if (/▪▪▪/.test(trimmed))
+            return false;
+        if (/^[\s▐▛▜▝▘█▌▪·↯]+$/.test(trimmed))
+            return false;
+        if (/^[─═\s▪·↯]+$/.test(trimmed))
+            return false;
+        if (/update\s*avai|brew\s*upgrade/i.test(trimmed))
+            return false;
+        if (/shift\+tab/i.test(trimmed))
+            return false;
+        if (/fast\s*mode/i.test(trimmed))
+            return false;
+        if (/^…\s/.test(trimmed) && trimmed.length < 80)
+            return false;
+        // Status lines
+        const lower = trimmed.toLowerCase();
+        if (STATUS_KEYWORDS.some((kw) => lower.includes(kw)) && trimmed.length < 80)
+            return false;
+        return true;
+    });
+    // Limit to last ~50 KB
+    const MAX_BYTES = 50 * 1024;
+    let result = filtered.join("\n");
+    if (Buffer.byteLength(result) > MAX_BYTES) {
+        const buf = Buffer.from(result);
+        result = buf.subarray(buf.length - MAX_BYTES).toString("utf-8");
+        const firstNewline = result.indexOf("\n");
+        if (firstNewline > 0) {
+            result = result.slice(firstNewline + 1);
+        }
+    }
+    return result.trim();
+}
+// ── Deduplication ────────────────────────────────────
+/**
+ * Remove duplicate consecutive lines (TUI re-renders can emit the same
+ * content multiple times).  Comparison ignores whitespace differences.
+ */
+export function deduplicateLines(lines) {
+    const result = [];
+    let lastNorm = "";
+    for (const line of lines) {
+        const norm = line.replace(/\s+/g, "");
+        if (norm !== lastNorm || norm === "") {
+            result.push(line);
+            lastNorm = norm;
+        }
+    }
+    return result;
+}
+//# sourceMappingURL=pty-filter.js.map

package/package.json

@@ -0,0 +1,71 @@
+{
+  "name": "@lawrence369/loop-cli",
+  "version": "0.1.0",
+  "description": "Iterative Multi-Engine AI Orchestration",
+  "type": "module",
+  "main": "./dist/index.js",
+  "types": "./dist/index.d.ts",
+  "bin": {
+    "loop": "./dist/index.js",
+    "lclaude": "./dist/bin/lclaude.js",
+    "lgemini": "./dist/bin/lgemini.js",
+    "lcodex": "./dist/bin/lcodex.js"
+  },
+  "scripts": {
+    "dev": "tsx src/index.ts",
+    "build": "tsc",
+    "start": "node dist/index.js",
+    "test": "vitest run",
+    "test:watch": "vitest",
+    "test:coverage": "vitest run --coverage"
+  },
+  "engines": {
+    "node": ">=18"
+  },
+  "files": [
+    "dist/**/*.js",
+    "dist/**/*.d.ts",
+    "skills/",
+    "README.md",
+    "LICENSE",
+    "CHANGELOG.md"
+  ],
+  "keywords": [
+    "cli",
+    "ai",
+    "orchestration",
+    "multi-agent",
+    "claude",
+    "gemini",
+    "codex",
+    "iteration",
+    "loop"
+  ],
+  "author": "loop-cli contributors",
+  "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/lawrence3699/loop.git"
+  },
+  "homepage": "https://github.com/lawrence3699/loop",
+  "bugs": {
+    "url": "https://github.com/lawrence3699/loop/issues"
+  },
+  "dependencies": {
+    "@clack/prompts": "^0.10.0",
+    "blessed": "^0.1.81",
+    "commander": "^12.0.0",
+    "gradient-string": "^3.0.0",
+    "gray-matter": "^4.0.3",
+    "node-pty": "^1.0.0",
+    "strip-ansi": "^7.2.0"
+  },
+  "devDependencies": {
+    "@types/blessed": "^0.1.25",
+    "@types/node": "^22.0.0",
+    "tsx": "^4.0.0",
+    "typescript": "^5.5.0",
+    "vitest": "^3.0.0",
+    "@vitest/coverage-v8": "^3.0.0"
+  }
+}

package/skills/loop/SKILL.md

@@ -0,0 +1,19 @@
+---
+name: loop
+description: Core iteration loop protocol
+---
+
+# Loop Protocol
+
+You are participating in an iterative review loop. Your output will be reviewed and scored.
+
+## As Executor
+- Focus on the task requirements
+- Address ALL reviewer feedback from previous iterations
+- Show your work: list files changed and commands executed
+
+## As Reviewer
+- Score the output from 1-10
+- List specific issues found
+- Provide actionable suggestions
+- Write "APPROVED" on its own line if score >= 9

package/skills/plan/SKILL.md

@@ -0,0 +1,9 @@
+---
+name: plan
+description: Plan management
+---
+
+# Plan Protocol
+
+Track decisions and iteration history in the shared plan file.
+Reference previous iteration feedback when making changes.

package/skills/review/SKILL.md

@@ -0,0 +1,14 @@
+---
+name: review
+description: Code review workflow
+---
+
+# Review Protocol
+
+When reviewing code output:
+1. Check correctness: Does it solve the stated task?
+2. Check completeness: Are all requirements addressed?
+3. Check quality: Is the code clean, well-structured, and maintainable?
+4. Check edge cases: Are error cases handled?
+
+Score 1-10. Write "APPROVED" if satisfied.