donsetch 0.5.0-beta.1

package/README.md

@@ -0,0 +1,84 @@
+# DonSeTch
+
+> Web fetch, search and crawl for AI agents. Zero API keys.
+
+[![CI](https://github.com/dondai44423/donsetch/actions/workflows/ci.yml/badge.svg)](https://github.com/dondai44423/donsetch/actions/workflows/ci.yml)
+
+DonSeTch is an MCP server that gives AI agents web research capabilities:
+fetch any URL, search across 10+ engines, and crawl multi-page docs —
+all with Chrome-true TLS fingerprinting and zero API keys.
+
+## Install
+
+```bash
+npm install -g donsetch
+```
+
+The postinstall script downloads the prebuilt binary for your platform
+from [GitHub Releases](https://github.com/dondai44423/donsetch/releases).
+
+| Platform | Binary |
+|---|---|
+| Linux x86_64 | `donsetch-linux-x64.tar.gz` |
+| macOS arm64 | `donsetch-darwin-arm64.tar.gz` |
+| Windows x86_64 | `donsetch-win32-x64.tar.gz` |
+
+## Usage
+
+```bash
+# Fetch a URL (Chrome-true TLS, anti-bot bypass)
+donsetch fetch https://example.com
+
+# Search (10+ keyless backends, semantic reranking)
+donsetch search "rust async patterns"
+
+# Crawl (sitemap-aware, focus-filtered)
+donsetch crawl https://docs.example.com --focus "api reference"
+
+# Start MCP server (for Claude Code, Cursor, Pi, etc.)
+donsetch mcp
+```
+
+## MCP Configuration
+
+Add to your MCP client config:
+
+```json
+{
+  "mcpServers": {
+    "donsetch": {
+      "command": "donsetch",
+      "args": ["mcp"]
+    }
+  }
+}
+```
+
+Or use `npx` without global install:
+
+```json
+{
+  "mcpServers": {
+    "donsetch": {
+      "command": "npx",
+      "args": ["donsetch", "mcp"]
+    }
+  }
+}
+```
+
+Three tools: `web_fetch`, `web_search`, `web_crawl`.
+
+## Build from source
+
+Requirements: Rust 1.75+, Go 1.22+, NASM, LLVM/Clang (for BoringSSL bindgen).
+
+```bash
+git clone https://github.com/dondai44423/donsetch
+cd donsetch
+cargo build --release
+```
+
+## License
+
+AGPL-3.0 — Copyright (c) 2026 Bishesh Bhandari

package/bin/donsetch.js

@@ -0,0 +1,69 @@
+#!/usr/bin/env node
+'use strict';
+
+// donsetch CLI wrapper: spawn the native binary with forwarded stdio.
+//
+// For MCP server usage (`donsetch mcp`), the MCP client spawns this
+// wrapper as a subprocess and communicates via stdin/stdout (JSON-RPC).
+// stdio: 'inherit' pipes the native binary's stdio directly to the
+// parent process, so the MCP protocol passes through unmodified.
+//
+// Signal forwarding: when the MCP client sends SIGTERM/SIGINT to this
+// wrapper, we forward it to the native binary so it can clean up
+// (close connections, save ghost state, etc.) before exiting.
+
+const { spawn } = require('child_process');
+const { existsSync } = require('fs');
+const { join } = require('path');
+
+// ── resolve binary path ─────────────────────────────────────────
+const binaryName = process.platform === 'win32' ? 'donsetch.exe' : 'donsetch';
+const binDir = join(__dirname, '..', 'binaries');
+const binaryPath = join(binDir, binaryName);
+
+if (!existsSync(binaryPath)) {
+  process.stderr.write('donsetch: binary not found. Run `npm rebuild donsetch` to reinstall.\n');
+  process.exit(1);
+}
+
+// ── spawn native binary ─────────────────────────────────────────
+const child = spawn(binaryPath, process.argv.slice(2), {
+  stdio: 'inherit',
+  cwd: process.cwd(),
+  env: process.env,
+  windowsHide: true,
+});
+
+// ── forward signals to child ────────────────────────────────────
+// When the parent receives SIGTERM/SIGINT (e.g. MCP client
+// disconnecting), forward to the child so it can clean up.
+let killed = false;
+
+function forwardSignal(sig) {
+  if (!killed) {
+    killed = true;
+    try { child.kill(sig); } catch (_) {}
+    // Force exit after 5s if child doesn't respond
+    setTimeout(() => process.exit(1), 5000).unref();
+  }
+}
+
+process.on('SIGTERM', () => forwardSignal('SIGTERM'));
+process.on('SIGINT', () => forwardSignal('SIGINT'));
+process.on('SIGHUP', () => forwardSignal('SIGHUP'));
+
+// ── exit with child's code ──────────────────────────────────────
+child.on('exit', (code, signal) => {
+  if (signal) {
+    // Re-raise the signal so the parent's exit reflects it
+    try { process.kill(process.pid, signal); } catch (_) {}
+    process.exit(128 + 15); // 128 + SIGTERM
+  } else {
+    process.exit(code || 0);
+  }
+});
+
+child.on('error', (err) => {
+  process.stderr.write(`donsetch: failed to spawn binary: ${err.message}\n`);
+  process.exit(1);
+});

package/install.js

@@ -0,0 +1,152 @@
+#!/usr/bin/env node
+'use strict';
+
+// donsetch postinstall: download the prebuilt binary for this platform
+// from GitHub Releases, verify SHA256, and extract to ./binaries/.
+//
+// The binary is NOT bundled in the npm package — it's fetched at
+// install time from the GitHub release matching this package version.
+// This keeps the npm registry clean (no 35 MB binary tarballs) and
+// uses the same release artifacts that manual users download.
+//
+// Supported platforms:
+//   linux-x64      Linux x86_64
+//   darwin-arm64   macOS Apple Silicon
+//   win32-x64      Windows x86_64
+
+const https = require('https');
+const crypto = require('crypto');
+const fs = require('fs');
+const path = require('path');
+const { execSync } = require('child_process');
+
+const REPO = 'dondai44423/donsetch';
+const VERSION = require('./package.json').version;
+const TAG = `v${VERSION}`;
+
+// ── platform mapping ────────────────────────────────────────────
+const PLATFORMS = {
+  'linux-x64':    { asset: 'donsetch-linux-x64.tar.gz',    binary: 'donsetch'     },
+  'darwin-arm64': { asset: 'donsetch-darwin-arm64.tar.gz', binary: 'donsetch'     },
+  'win32-x64':    { asset: 'donsetch-win32-x64.tar.gz',    binary: 'donsetch.exe' },
+};
+
+const platKey = `${process.platform}-${process.arch}`;
+const plat = PLATFORMS[platKey];
+
+if (!plat) {
+  console.error(`donsetch: unsupported platform ${platKey}`);
+  console.error('');
+  console.error('Supported platforms:');
+  console.error('  linux-x64      Linux x86_64');
+  console.error('  darwin-arm64   macOS Apple Silicon');
+  console.error('  win32-x64      Windows x86_64');
+  console.error('');
+  console.error('Build from source: https://github.com/' + REPO);
+  process.exit(1);
+}
+
+const binDir = path.join(__dirname, 'binaries');
+const binaryPath = path.join(binDir, plat.binary);
+
+// Skip if already installed (npm cache reuse, reinstall, etc.)
+if (fs.existsSync(binaryPath)) {
+  console.log(`donsetch: binary already present (${plat.binary})`);
+  process.exit(0);
+}
+
+fs.mkdirSync(binDir, { recursive: true });
+
+const baseUrl = `https://github.com/${REPO}/releases/download/${TAG}`;
+const assetUrl = `${baseUrl}/${plat.asset}`;
+const checksumUrl = `${baseUrl}/${plat.asset}.sha256`;
+
+// ── download with redirect following ────────────────────────────
+function download(url, dest) {
+  return new Promise((resolve, reject) => {
+    function get(u) {
+      const opts = { headers: { 'Accept': 'application/octet-stream', 'User-Agent': 'donsetch-npm-installer' } };
+      https.get(u, opts, (res) => {
+        if ([301, 302, 303, 307, 308].includes(res.statusCode)) {
+          res.resume();
+          get(res.headers.location);
+          return;
+        }
+        if (res.statusCode !== 200) {
+          res.resume();
+          reject(new Error(`HTTP ${res.statusCode} for ${u}`));
+          return;
+        }
+        const file = fs.createWriteStream(dest);
+        res.pipe(file);
+        file.on('finish', () => { file.close(); resolve(); });
+        file.on('error', (err) => {
+          try { fs.unlinkSync(dest); } catch (_) {}
+          reject(err);
+        });
+      }).on('error', reject);
+    }
+    get(url);
+  });
+}
+
+// ── main ────────────────────────────────────────────────────────
+async function main() {
+  const tarball = path.join(binDir, plat.asset);
+  const checksumFile = path.join(binDir, 'checksum.sha256');
+
+  // 1. Download the binary tarball
+  console.log(`donsetch: downloading ${plat.asset} from ${TAG}...`);
+  await download(assetUrl, tarball);
+
+  // 2. Download the SHA256 checksum
+  console.log('donsetch: verifying checksum...');
+  await download(checksumUrl, checksumFile);
+
+  // 3. Verify SHA256
+  const expectedHash = fs.readFileSync(checksumFile, 'utf8').trim().split(/\s+/)[0];
+  const actualHash = crypto.createHash('sha256').update(fs.readFileSync(tarball)).digest('hex');
+
+  if (actualHash !== expectedHash) {
+    try { fs.unlinkSync(tarball); } catch (_) {}
+    try { fs.unlinkSync(checksumFile); } catch (_) {}
+    console.error(`donsetch: SHA256 mismatch!`);
+    console.error(`  expected: ${expectedHash}`);
+    console.error(`  actual:   ${actualHash}`);
+    console.error('The download may have been corrupted or tampered with.');
+    process.exit(1);
+  }
+
+  // 4. Extract (tar is built into Linux, macOS, and Windows 10+)
+  console.log('donsetch: extracting...');
+  execSync(`tar xzf "${tarball}" -C "${binDir}"`, { stdio: 'inherit' });
+
+  // 5. Cleanup
+  try { fs.unlinkSync(tarball); } catch (_) {}
+  try { fs.unlinkSync(checksumFile); } catch (_) {}
+
+  // 6. Make executable (Unix only)
+  if (process.platform !== 'win32') {
+    fs.chmodSync(binaryPath, 0o755);
+  }
+
+  // 7. Verify the binary exists after extraction
+  if (!fs.existsSync(binaryPath)) {
+    console.error(`donsetch: expected ${plat.binary} not found after extraction`);
+    console.error(`  looked at: ${binaryPath}`);
+    console.error('  contents of binaries/:', fs.readdirSync(binDir).join(', '));
+    process.exit(1);
+  }
+
+  console.log(`donsetch: installed ${plat.binary} to ${binaryPath}`);
+  console.log(`donsetch: run \`donsetch\` to see available commands.`);
+}
+
+main().catch((err) => {
+  console.error(`donsetch: install failed: ${err.message}`);
+  console.error('');
+  console.error('You can build from source:');
+  console.error('  git clone https://github.com/' + REPO);
+  console.error('  cd donsetch && cargo build --release');
+  process.exit(1);
+});

package/package.json

@@ -0,0 +1,42 @@
+{
+  "name": "donsetch",
+  "version": "0.5.0-beta.1",
+  "description": "Web fetch, search and crawl for AI agents. Zero API keys. Chrome-true TLS.",
+  "license": "AGPL-3.0-only",
+  "author": "Bishesh Bhandari",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/dondai44423/donsetch"
+  },
+  "homepage": "https://github.com/dondai44423/donsetch#readme",
+  "bugs": {
+    "url": "https://github.com/dondai44423/donsetch/issues"
+  },
+  "keywords": [
+    "mcp",
+    "web-fetch",
+    "search",
+    "crawl",
+    "ai-agent",
+    "scraper",
+    "tls",
+    "chrome",
+    "anti-bot",
+    "pdf",
+    "extraction"
+  ],
+  "bin": {
+    "donsetch": "bin/donsetch.js"
+  },
+  "scripts": {
+    "postinstall": "node install.js"
+  },
+  "files": [
+    "install.js",
+    "bin/donsetch.js",
+    "README.md"
+  ],
+  "engines": {
+    "node": ">=14"
+  }
+}