@link-assistant/hive-mind 2.0.21 → 2.0.22

package/CHANGELOG.md

@@ -1,5 +1,11 @@
 # @link-assistant/hive-mind
 
+## 2.0.22
+
+### Patch Changes
+
+- 4d65c05: Make disk admission safer by default: the disk usage queue gate now waits at 80%, the absolute free-space default is 10240 MB, and isolation defaults to Docker.
+
 ## 2.0.21
 
 ### Patch Changes

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@link-assistant/hive-mind",
-  "version": "2.0.21",
+  "version": "2.0.22",
   "description": "AI-powered issue solver and hive mind for collaborative problem solving",
   "main": "src/hive.mjs",
   "type": "module",

package/src/config.lib.mjs

@@ -111,7 +111,7 @@ export const githubLimits = {
 
 // Memory and disk configurations
 export const systemLimits = {
-  minDiskSpaceMb: parseIntWithDefault('HIVE_MIND_MIN_DISK_SPACE_MB', 2048),
+  minDiskSpaceMb: parseIntWithDefault('HIVE_MIND_MIN_DISK_SPACE_MB', 10240),
   defaultPageSizeKb: parseIntWithDefault('HIVE_MIND_DEFAULT_PAGE_SIZE_KB', 16),
 };
 

package/src/hive.mjs

@@ -1444,7 +1444,7 @@ if (isRunningDirectly) {
     } else {
       const systemCheck = await checkSystem(
         {
-          minDiskSpaceMB: argv.minDiskSpace || 2048,
+          minDiskSpaceMB: argv.minDiskSpace || 10240,
           minMemoryMB: 256,
           exitOnFailure: true,
         },

package/src/memory-check.mjs

@@ -26,7 +26,7 @@ const lib = await import('./lib.mjs');
 const { log: libLog, setLogFile } = lib;
 
 // Function to check available disk space
-export const checkDiskSpace = async (minSpaceMB = 2048, options = {}) => {
+export const checkDiskSpace = async (minSpaceMB = 10240, options = {}) => {
   const log = options.log || libLog;
 
   try {
@@ -289,7 +289,7 @@ export const getResourceSnapshot = async () => {
 
 // Combined system check function
 export const checkSystem = async (requirements = {}, options = {}) => {
-  const { minMemoryMB = 256, minDiskSpaceMB = 2048, exitOnFailure = false } = requirements;
+  const { minMemoryMB = 256, minDiskSpaceMB = 10240, exitOnFailure = false } = requirements;
 
   // Note: log is passed through options to checkDiskSpace and checkRAM
   const results = {
@@ -334,7 +334,7 @@ const createMemoryCheckYargsConfig = yargsInstance =>
       alias: 'd',
       type: 'number',
       description: 'Minimum required disk space in MB',
-      default: 2048,
+      default: 10240,
     })
     .option('exit-on-failure', {
       alias: 'e',

package/src/queue-config.lib.mjs

@@ -22,6 +22,7 @@ import { ensureUseM } from './use-m-bootstrap.lib.mjs';
  *
  * @see https://github.com/link-assistant/hive-mind/issues/1242
  * @see https://github.com/link-assistant/hive-mind/issues/1253
+ * @see https://github.com/link-assistant/hive-mind/issues/1981
  */
 
 // Use use-m to dynamically import modules
@@ -236,9 +237,8 @@ function getThresholdConfig(linoKey, envVarThreshold, envVarStrategy, defaultThr
  * - 'enqueue': Block and wait in queue
  * - 'dequeue-one-at-a-time': Allow one command, block subsequent
  *
- * BREAKING CHANGE: Disk threshold default strategy changed from 'dequeue-one-at-a-time' to 'reject'
- * because the queue is lost on server restart anyway, so there's no point in queueing.
- * To restore old behavior: HIVE_MIND_DISK_STRATEGY=dequeue-one-at-a-time
+ * Issue #1981: disk now defaults to the normal wait/enqueue path at 80% used.
+ * Operators can still choose immediate rejection with HIVE_MIND_DISK_STRATEGY=reject.
  */
 export const QUEUE_CONFIG = {
   // Threshold configurations with value and strategy
@@ -246,10 +246,8 @@ export const QUEUE_CONFIG = {
   thresholds: {
     ram: getThresholdConfig('ram', 'HIVE_MIND_RAM_THRESHOLD', 'HIVE_MIND_RAM_STRATEGY', 0.65, 'enqueue'),
     cpu: getThresholdConfig('cpu', 'HIVE_MIND_CPU_THRESHOLD', 'HIVE_MIND_CPU_STRATEGY', 0.65, 'enqueue'),
-    // BREAKING: disk default changed from 'dequeue-one-at-a-time' to 'reject'
-    // Queue is in RAM and lost on restart - no point enlarging it when disk is full
-    // See: https://github.com/link-assistant/hive-mind/issues/1253
-    disk: getThresholdConfig('disk', 'HIVE_MIND_DISK_THRESHOLD', 'HIVE_MIND_DISK_STRATEGY', 0.9, 'reject'),
+    // Issue #1981: wait instead of immediately rejecting when disk crosses 80%.
+    disk: getThresholdConfig('disk', 'HIVE_MIND_DISK_THRESHOLD', 'HIVE_MIND_DISK_STRATEGY', 0.8, 'enqueue'),
     claude5Hour: getThresholdConfig('claude5Hour', 'HIVE_MIND_CLAUDE_5_HOUR_SESSION_THRESHOLD', 'HIVE_MIND_CLAUDE_5_HOUR_SESSION_STRATEGY', 0.65, 'dequeue-one-at-a-time'),
     claudeWeekly: getThresholdConfig('claudeWeekly', 'HIVE_MIND_CLAUDE_WEEKLY_THRESHOLD', 'HIVE_MIND_CLAUDE_WEEKLY_STRATEGY', 0.97, 'dequeue-one-at-a-time'),
     codex5Hour: getThresholdConfig('codex5Hour', 'HIVE_MIND_CODEX_5_HOUR_SESSION_THRESHOLD', 'HIVE_MIND_CODEX_5_HOUR_SESSION_STRATEGY', 0.65, 'dequeue-one-at-a-time'),
@@ -265,7 +263,7 @@ export const QUEUE_CONFIG = {
   // These are derived from thresholds.{metric}.value
   RAM_THRESHOLD: getThresholdConfig('ram', 'HIVE_MIND_RAM_THRESHOLD', 'HIVE_MIND_RAM_STRATEGY', 0.65, 'enqueue').value,
   CPU_THRESHOLD: getThresholdConfig('cpu', 'HIVE_MIND_CPU_THRESHOLD', 'HIVE_MIND_CPU_STRATEGY', 0.65, 'enqueue').value,
-  DISK_THRESHOLD: getThresholdConfig('disk', 'HIVE_MIND_DISK_THRESHOLD', 'HIVE_MIND_DISK_STRATEGY', 0.9, 'reject').value,
+  DISK_THRESHOLD: getThresholdConfig('disk', 'HIVE_MIND_DISK_THRESHOLD', 'HIVE_MIND_DISK_STRATEGY', 0.8, 'enqueue').value,
   CLAUDE_5_HOUR_SESSION_THRESHOLD: getThresholdConfig('claude5Hour', 'HIVE_MIND_CLAUDE_5_HOUR_SESSION_THRESHOLD', 'HIVE_MIND_CLAUDE_5_HOUR_SESSION_STRATEGY', 0.65, 'dequeue-one-at-a-time').value,
   CLAUDE_WEEKLY_THRESHOLD: getThresholdConfig('claudeWeekly', 'HIVE_MIND_CLAUDE_WEEKLY_THRESHOLD', 'HIVE_MIND_CLAUDE_WEEKLY_STRATEGY', 0.97, 'dequeue-one-at-a-time').value,
   CODEX_5_HOUR_SESSION_THRESHOLD: getThresholdConfig('codex5Hour', 'HIVE_MIND_CODEX_5_HOUR_SESSION_THRESHOLD', 'HIVE_MIND_CODEX_5_HOUR_SESSION_STRATEGY', 0.65, 'dequeue-one-at-a-time').value,

package/src/solve.config.lib.mjs

@@ -289,8 +289,8 @@ export const SOLVE_OPTION_DEFINITIONS = {
   },
   'min-disk-space': {
     type: 'number',
-    description: 'Minimum required disk space in MB (default: 2048)',
-    default: 2048,
+    description: 'Minimum required disk space in MB (default: 10240)',
+    default: 10240,
   },
   'log-dir': {
     type: 'string',

package/src/solve.mjs

@@ -248,7 +248,7 @@ if (argv.planModel) {
 const skipToolConnectionCheck = argv.dryRun || argv.skipToolConnectionCheck || argv.toolConnectionCheck === false;
 const { cascadePlaywrightMcpDisable, ensureSolvePlaywrightMcpReady } = await import('./playwright-mcp.lib.mjs');
 await cascadePlaywrightMcpDisable(argv, log);
-if (!(await performSystemChecks(argv.minDiskSpace || 2048, skipToolConnectionCheck, argv.model, argv))) {
+if (!(await performSystemChecks(argv.minDiskSpace || 10240, skipToolConnectionCheck, argv.model, argv))) {
   await safeExit(1, 'System checks failed');
 }
 // Playwright MCP preflight is local/free and stays independent from paid tool connection checks.

package/src/solve.validation.lib.mjs

@@ -54,7 +54,7 @@ const { parseResetTime: parseResetTimeToDate } = usageLimitLib;
 const { validateClaudeConnection } = claudeLib;
 
 // Wrapper function for disk space check using imported module
-const checkDiskSpace = async (minSpaceMB = 2048) => {
+const checkDiskSpace = async (minSpaceMB = 10240) => {
   const result = await memoryCheck.checkDiskSpace(minSpaceMB, { log });
   return result.success;
 };
@@ -216,7 +216,7 @@ export const validateContinueOnlyOnFeedback = async (argv, isPrUrl, isIssueUrl)
 // Perform all system checks (disk space, memory, tool connection, GitHub permissions)
 // Note: skipToolConnection only skips the connection check, not model validation
 // Model validation should be done separately before calling this function
-export const performSystemChecks = async (minDiskSpace = 2048, skipToolConnection = false, model = 'sonnet', argv = {}) => {
+export const performSystemChecks = async (minDiskSpace = 10240, skipToolConnection = false, model = 'sonnet', argv = {}) => {
   // Check disk space before proceeding
   const hasEnoughSpace = await checkDiskSpace(minDiskSpace);
   if (!hasEnoughSpace) {

package/src/task.config.lib.mjs

@@ -61,7 +61,7 @@ export const createYargsConfig = yargsInstance =>
       type: 'string',
       description: 'agent-commander isolation mode',
       choices: ['screen', 'none', 'docker'],
-      default: 'screen',
+      default: 'docker',
     })
     .option('screen-name', {
       type: 'string',

package/src/task.mjs

@@ -35,7 +35,7 @@ if (earlyArgs.length === 0 || earlyArgs.includes('--help') || earlyArgs.includes
   console.log('  --split-count      Number of issues to split into [default: 2]');
   console.log('  --tool             AI tool for agent-commander read-only mode (claude, codex, opencode, agent, qwen, gemini) [default: claude]');
   console.log('  --model, -m        Model to use');
-  console.log('  --isolation        agent-commander isolation mode [default: screen]');
+  console.log('  --isolation        agent-commander isolation mode [default: docker]');
   console.log('  --dry-run          Print split output without creating GitHub issues');
   console.log('  --verbose, -v      Enable verbose logging');
   console.log('  --output-format    Output format (text or json) [default: text]');

package/src/telegram-bot.mjs

@@ -102,7 +102,7 @@ const config = yargs(hideBin(process.argv))
   .option('autoStartScreenWatchMessage', { type: 'boolean', description: 'Experimental: auto-start separate /terminal_watch messages for public /solve sessions', alias: 'auto-start-screen-watch-message', default: getenv('TELEGRAM_AUTO_START_SCREEN_WATCH_MESSAGE', getenv('TELEGRAM_AUTO_WATCH_MESSAGE', 'false')) === 'true' })
   // Issue #594: bot-owner toggle for --show-limits virtual option in /solve and /hive.
   .option('showLimits', { type: 'boolean', description: 'Experimental: allow /solve and /hive callers to use --show-limits to embed Claude/Codex usage at start, end, and delta in the completion message', alias: 'show-limits', default: getenv('TELEGRAM_SHOW_LIMITS', 'true') !== 'false' })
-  .option('isolation', { type: 'string', description: "Isolation backend (screen/tmux/docker). Defaults to 'screen' so Telegram-bot work sessions survive bot restarts; pass --isolation '' (or set TELEGRAM_ISOLATION='') to disable.", default: getenv('TELEGRAM_ISOLATION', 'screen') })
+  .option('isolation', { type: 'string', description: "Isolation backend (screen/tmux/docker). Defaults to 'docker' so Telegram-bot work sessions run in Docker isolation; pass --isolation '' (or set TELEGRAM_ISOLATION='') to disable.", default: getenv('TELEGRAM_ISOLATION', 'docker') })
   .help('h')
   .alias('h', 'help')
   .parserConfiguration({

package/src/telegram-solve-queue.lib.mjs

@@ -724,10 +724,11 @@ export class SolveQueue {
    * Default strategies:
    * - RAM: enqueue
    * - CPU: enqueue
-   * - DISK: reject (changed from dequeue-one-at-a-time - queue lost on restart)
+   * - DISK: enqueue (waits until disk drops below the threshold)
    *
    * See: https://github.com/link-assistant/hive-mind/issues/1155
    * See: https://github.com/link-assistant/hive-mind/issues/1253
+   * See: https://github.com/link-assistant/hive-mind/issues/1981
    *
    * @param {number} totalProcessing - Total processing count (queue + external claude processes)
    * @returns {Promise<{ok: boolean, reasons: string[], oneAtATime: boolean, rejected: boolean, rejectReason: string|null}>}

package/src/telegram-start-stop-command.lib.mjs

@@ -584,7 +584,7 @@ export function registerStartStopCommands(bot, options) {
       // immediately and was dispatched to a detached session) but the session
       // monitor still tracks a running isolated session for this URL, forward
       // CTRL+C to its start-command UUID. This is the common case for tasks
-      // that begin executing right away with `--isolation screen`.
+      // that begin executing right away with an isolation backend.
       const queueHasTask = lookup.action === 'cancel-queued' || lookup.action === 'stop-running';
       if (!queueHasTask && runningSession?.stoppable && runningSession.sessionId) {
         VERBOSE && console.log(`[VERBOSE] /stop: forwarding CTRL+C to tracked session ${runningSession.sessionId} for ${url} (queue action=${lookup.action})`);
@@ -597,7 +597,7 @@ export function registerStartStopCommands(bot, options) {
         // running-but-non-stoppable (non-isolation) session, say so; otherwise
         // fall back to the UUID hint.
         if (runningSession) {
-          await ctx.reply(`⚠️ Found a running task for ${url}, but it was not started with an isolation backend, so \`/stop\` cannot forward CTRL+C to it.\n\nNext time you can run the command with \`--isolation screen\` to make this task interruptible via \`/stop\`.`, {
+          await ctx.reply(`⚠️ Found a running task for ${url}, but it was not started with an isolation backend, so \`/stop\` cannot forward CTRL+C to it.\n\nNext time run it with the default isolation backend or pass \`--isolation docker\` to make this task interruptible via \`/stop\`.`, {
             parse_mode: 'Markdown',
             reply_to_message_id: message.message_id,
           });
@@ -615,13 +615,13 @@ export function registerStartStopCommands(bot, options) {
         // have forwarded CTRL+C above). If it tracked a non-isolation session,
         // explain why it can't be stopped; otherwise report not found.
         if (runningSession) {
-          await ctx.reply(`⚠️ Found a running task for ${url}, but it was not started with an isolation backend, so \`/stop\` cannot forward CTRL+C to it.\n\nNext time you can run the command with \`--isolation screen\` to make this task interruptible via \`/stop\`.`, {
+          await ctx.reply(`⚠️ Found a running task for ${url}, but it was not started with an isolation backend, so \`/stop\` cannot forward CTRL+C to it.\n\nNext time run it with the default isolation backend or pass \`--isolation docker\` to make this task interruptible via \`/stop\`.`, {
             parse_mode: 'Markdown',
             reply_to_message_id: message.message_id,
           });
           return;
         }
-        await ctx.reply(`ℹ️ No queued or running task found for ${url}.\n\nIf the task is running with \`--isolation screen\`, try \`/stop <UUID>\` (the UUID is shown in the bot's session-id message).`, {
+        await ctx.reply(`ℹ️ No queued or running task found for ${url}.\n\nIf the task is running with an isolation backend, try \`/stop <UUID>\` (the UUID is shown in the bot's session-id message).`, {
           parse_mode: 'Markdown',
           reply_to_message_id: message.message_id,
         });
@@ -655,7 +655,7 @@ export function registerStartStopCommands(bot, options) {
       // running-not-isolated: a started, non-isolated screen session. We
       // could shell out to `screen -X -S <name> stuff $'\003'`, but that's
       // brittle and out of scope for #1780. Tell the user how to recover.
-      await ctx.reply(`⚠️ Found a running task for ${url}, but it was not started with an isolation backend, so \`/stop\` cannot forward CTRL+C to it.\n\nNext time you can run the command with \`--isolation screen\` to make this task interruptible via \`/stop\`.`, {
+      await ctx.reply(`⚠️ Found a running task for ${url}, but it was not started with an isolation backend, so \`/stop\` cannot forward CTRL+C to it.\n\nNext time run it with the default isolation backend or pass \`--isolation docker\` to make this task interruptible via \`/stop\`.`, {
         parse_mode: 'Markdown',
         reply_to_message_id: message.message_id,
       });