@evident-ai/cli 3.3.1-dev.dd703c3 → 3.4.0

package/README.md

@@ -105,6 +105,11 @@ Options:
   keeps retrying) if no usable login is found; `off` disables it entirely — no
   Claude credential is ever read. An unrecognized value falls back to `auto` with
   a warning. Env: `EVIDENT_CLAUDE_USAGE_REPORTING`.
+- `--no-resource-usage-reporting` — Don't report this machine's CPU
+  utilization, total and available memory, core count, disk space, and the
+  OpenCode session-store size to Evident, so it shows on the runner page. On
+  by default; nothing else about the machine leaves it. Env:
+  `EVIDENT_RESOURCE_USAGE_REPORTING=off` (the flag wins if both are set).
 - `--json` — Output in JSON format (forces non-interactive mode).
 - `--session-cleanup-max-age <duration>` — Delete OpenCode sessions idle longer
   than this window (format `<number><unit>`, unit one of `s, m, h, d` — e.g.
@@ -171,6 +176,9 @@ targets the **production** Evident platform by default.
 - `EVIDENT_TUNNEL_URL` — Override the tunnel relay URL (equivalent to `--tunnel`).
 - `EVIDENT_CLAUDE_USAGE_REPORTING` — Equivalent to `--claude-usage-reporting`; the
   flag wins if both are set.
+- `EVIDENT_RESOURCE_USAGE_REPORTING` — Equivalent to
+  `--no-resource-usage-reporting`; set it to `off` to disable. The flag wins
+  if both are set.
 - `EVIDENT_LOG_LEVEL` — Equivalent to `--log-level`; the flag wins if both are
   set, and `-v`/`--verbose` also outranks this env var.
 - `EVIDENT_OPENCODE_START_TIMEOUT` — Equivalent to `--opencode-start-timeout`

package/dist/index.js

@@ -746,6 +746,35 @@ async function reportClaudeUsage(agentId, authHeader, snapshot) {
     return { ok: false, error: describeBestEffortError(error2) };
   }
 }
+async function reportResourceUsage(agentId, authHeader, usage) {
+  try {
+    const apiUrl = getApiUrlConfig();
+    const response = await fetch(`${apiUrl}/runners/${agentId}/resource-usage`, {
+      method: "POST",
+      headers: { Authorization: authHeader, "Content-Type": "application/json" },
+      body: JSON.stringify({
+        cpu_percent: usage.cpuPercent,
+        cpu_count: usage.cpuCount,
+        memory_total_bytes: usage.memoryTotalBytes,
+        memory_available_bytes: usage.memoryAvailableBytes,
+        disk_total_bytes: usage.diskTotalBytes,
+        disk_free_bytes: usage.diskFreeBytes,
+        opencode_db_bytes: usage.opencodeDbBytes
+      }),
+      signal: AbortSignal.timeout(BEST_EFFORT_NOTIFY_TIMEOUT_MS)
+    });
+    if (!response.ok) {
+      const serverMessage = await readErrorMessage(response);
+      return {
+        ok: false,
+        error: `HTTP ${response.status}${serverMessage ? `: ${serverMessage}` : ""}`
+      };
+    }
+    return { ok: true };
+  } catch (error2) {
+    return { ok: false, error: describeBestEffortError(error2) };
+  }
+}
 async function getAgentInfo(agentId, authHeader) {
   const apiUrl = getApiUrlConfig();
   try {
@@ -2962,6 +2991,21 @@ function writeTunnelReadyMarker(path, agentId) {
   }
 }
 
+// src/lib/reporting-schedule.ts
+function jitteredDelayMs(baseMs, jitterFraction, random = Math.random) {
+  const jitterRangeMs = baseMs * jitterFraction;
+  return baseMs - jitterRangeMs + random() * (2 * jitterRangeMs);
+}
+function firstReportDelayMs(random = Math.random) {
+  return 5e3 + random() * 1e4;
+}
+function reportFailureLogLevel(consecutiveFailures, reescalationTicks) {
+  return consecutiveFailures === 1 || consecutiveFailures % reescalationTicks === 0 ? "warn" : "debug";
+}
+function failureStreakSuffix(consecutiveFailures) {
+  return consecutiveFailures > 1 ? ` (${consecutiveFailures} consecutive failures)` : "";
+}
+
 // src/lib/claude-usage-reporting.ts
 var VALID_MODES = ["auto", "on", "off"];
 function resolveClaudeUsageReportingMode(flagValue, env) {
@@ -2984,13 +3028,167 @@ function resolveClaudeUsageReportingMode(flagValue, env) {
 var BASE_REPORT_DELAY_MS = 10 * 6e4;
 var REPORT_DELAY_JITTER_FRACTION = 0.2;
 function nextReportDelayMs(random = Math.random) {
-  const jitterRangeMs = BASE_REPORT_DELAY_MS * REPORT_DELAY_JITTER_FRACTION;
-  return BASE_REPORT_DELAY_MS - jitterRangeMs + random() * (2 * jitterRangeMs);
+  return jitteredDelayMs(BASE_REPORT_DELAY_MS, REPORT_DELAY_JITTER_FRACTION, random);
 }
-var FIRST_REPORT_DELAY_MS = 5e3 + Math.random() * 1e4;
+var FIRST_REPORT_DELAY_MS = firstReportDelayMs();
 var CLAUDE_USAGE_FAILURE_REESCALATION_TICKS = 6;
 function claudeUsageFailureLogLevel(consecutiveFailures) {
-  return consecutiveFailures === 1 || consecutiveFailures % CLAUDE_USAGE_FAILURE_REESCALATION_TICKS === 0 ? "warn" : "debug";
+  return reportFailureLogLevel(consecutiveFailures, CLAUDE_USAGE_FAILURE_REESCALATION_TICKS);
+}
+
+// src/lib/resource-usage-reporting.ts
+var ENABLED_VALUES = /* @__PURE__ */ new Set(["on", "true", "1"]);
+var DISABLED_VALUES = /* @__PURE__ */ new Set(["off", "false", "0"]);
+function resolveResourceUsageReportingEnabled(flagValue, env) {
+  if (flagValue === false) {
+    return { enabled: false, warnings: [] };
+  }
+  const raw = env.EVIDENT_RESOURCE_USAGE_REPORTING;
+  if (raw === void 0 || raw === "") {
+    return { enabled: true, warnings: [] };
+  }
+  const normalized = raw.trim().toLowerCase();
+  if (DISABLED_VALUES.has(normalized)) {
+    return { enabled: false, warnings: [] };
+  }
+  if (ENABLED_VALUES.has(normalized)) {
+    return { enabled: true, warnings: [] };
+  }
+  return {
+    enabled: true,
+    warnings: [
+      `Ignoring invalid EVIDENT_RESOURCE_USAGE_REPORTING "${raw}": expected on or off; leaving reporting on`
+    ]
+  };
+}
+
+// src/lib/resource-usage.ts
+import { cpus, totalmem, freemem } from "os";
+import { statfsSync as statfsSync2 } from "fs";
+
+// src/lib/ecs-task-metadata.ts
+var ECS_METADATA_TIMEOUT_MS = 2e3;
+function parseEcsTaskLimits(payload) {
+  if (typeof payload !== "object" || payload === null) return null;
+  const limits = payload.Limits;
+  if (typeof limits !== "object" || limits === null) return null;
+  const cpu = limits.CPU;
+  const memory = limits.Memory;
+  if (typeof cpu !== "number" || !Number.isFinite(cpu) || cpu <= 0) return null;
+  if (typeof memory !== "number" || !Number.isFinite(memory) || memory <= 0) return null;
+  return {
+    cpuCount: Math.max(1, Math.round(cpu)),
+    memoryTotalBytes: memory * 1024 * 1024
+  };
+}
+async function readEcsTaskLimits(env) {
+  const uri = env.ECS_CONTAINER_METADATA_URI_V4;
+  if (!uri) {
+    return { limits: null };
+  }
+  const url = `${uri}/task`;
+  try {
+    const response = await fetch(url, { signal: AbortSignal.timeout(ECS_METADATA_TIMEOUT_MS) });
+    if (!response.ok) {
+      return {
+        limits: null,
+        warning: `ECS task metadata fetch (${url}) returned HTTP ${response.status}`
+      };
+    }
+    const payload = await response.json();
+    const limits = parseEcsTaskLimits(payload);
+    if (limits === null) {
+      return {
+        limits: null,
+        warning: `ECS task metadata fetch (${url}) returned an unexpected payload`
+      };
+    }
+    return { limits };
+  } catch (error2) {
+    const message = error2 instanceof Error ? error2.message : String(error2);
+    return { limits: null, warning: `ECS task metadata fetch (${url}) failed: ${message}` };
+  }
+}
+
+// src/lib/resource-usage.ts
+function readCpuSample() {
+  let busyMs = 0;
+  let idleMs = 0;
+  for (const cpu of cpus()) {
+    busyMs += cpu.times.user + cpu.times.nice + cpu.times.sys + cpu.times.irq;
+    idleMs += cpu.times.idle;
+  }
+  return { busyMs, idleMs };
+}
+function cpuPercentBetween(previous, current) {
+  const deltaBusy = current.busyMs - previous.busyMs;
+  const deltaIdle = current.idleMs - previous.idleMs;
+  const total = deltaBusy + deltaIdle;
+  if (total === 0) return null;
+  return Math.round((deltaBusy / total * 100 + Number.EPSILON) * 100) / 100;
+}
+function clamp(value, min, max) {
+  return Math.min(Math.max(value, min), max);
+}
+function round2(value) {
+  return Math.round((value + Number.EPSILON) * 100) / 100;
+}
+function readDisk(homeDir) {
+  try {
+    const stats = statfsSync2(homeDir);
+    return {
+      totalBytes: stats.bsize * stats.blocks,
+      freeBytes: stats.bsize * stats.bavail
+    };
+  } catch (error2) {
+    const message = error2 instanceof Error ? error2.message : String(error2);
+    return {
+      totalBytes: null,
+      freeBytes: null,
+      warning: `Could not read disk usage for ${homeDir}: ${message}`
+    };
+  }
+}
+function createResourceUsageCollector(homeDir) {
+  let previous = readCpuSample();
+  return async () => {
+    const current = readCpuSample();
+    const hostCpuPercent = cpuPercentBetween(previous, current);
+    const hostCpuCount = cpus().length;
+    previous = current;
+    const disk = readDisk(homeDir);
+    const opencodeDbBytes = statSessionDbBytes(homeDir);
+    const { limits, warning: ecsWarning } = await readEcsTaskLimits(process.env);
+    const warnings = [];
+    if (disk.warning) warnings.push(disk.warning);
+    if (ecsWarning) warnings.push(ecsWarning);
+    let cpuPercent = hostCpuPercent;
+    let cpuCount = hostCpuCount;
+    let memoryTotalBytes = totalmem();
+    let memoryAvailableBytes = freemem();
+    if (limits !== null) {
+      cpuCount = limits.cpuCount;
+      memoryTotalBytes = limits.memoryTotalBytes;
+      memoryAvailableBytes = clamp(
+        limits.memoryTotalBytes - (totalmem() - freemem()),
+        0,
+        limits.memoryTotalBytes
+      );
+      cpuPercent = hostCpuPercent === null ? null : clamp(round2(hostCpuPercent * hostCpuCount / limits.cpuCount), 0, 100);
+    }
+    return {
+      usage: {
+        cpuPercent,
+        cpuCount,
+        memoryTotalBytes,
+        memoryAvailableBytes,
+        diskTotalBytes: disk.totalBytes,
+        diskFreeBytes: disk.freeBytes,
+        opencodeDbBytes
+      },
+      warnings
+    };
+  };
 }
 
 // src/lib/channels/driver.ts
@@ -4851,6 +5049,17 @@ var ChannelDriver = class _ChannelDriver {
           message: `OpenCode session ${bound} for conversation ${conv.id.slice(0, 8)} no longer exists (deleted or DB reset) \u2014 creating a fresh session and rebinding.`,
           conversation_id: conv.id
         });
+        const watcher = this.watchers.get(bound);
+        if (watcher) {
+          for (const [evidentMessageId, inFlight] of [...watcher.inFlight]) {
+            this.recordReleasedOpencodeId(evidentMessageId, bound, inFlight.opencodeMessageId);
+            this.removeInFlight(watcher, evidentMessageId);
+            void this.postSignal(watcher.conv.id, evidentMessageId, "watcher_recovered", {
+              recovery: "session_gone_released"
+            });
+          }
+          this.watchers.delete(bound);
+        }
         this.sessions.delete(conv.id);
         return { sessionId: await this.createAndBindSession(conv.id), created: true };
       }
@@ -7735,9 +7944,6 @@ function scheduleSessionCleanup(state, driver, options) {
   );
   state.sessionCleanupTimers.push(interval, firstSweep);
 }
-function claudeUsageFailureStreakSuffix(consecutiveFailures) {
-  return consecutiveFailures > 1 ? ` (${consecutiveFailures} consecutive failures)` : "";
-}
 function scheduleClaudeUsageReporting(state, options) {
   const { mode, warnings } = resolveClaudeUsageReportingMode(
     options.claudeUsageReporting,
@@ -7819,7 +8025,7 @@ function scheduleClaudeUsageReporting(state, options) {
         logActivity(state, {
           type: "info",
           level: claudeUsageFailureLogLevel(consecutiveFailures),
-          message: `Failed to report Claude usage: ${result.error}${claudeUsageFailureStreakSuffix(consecutiveFailures)}`
+          message: `Failed to report Claude usage: ${result.error}${failureStreakSuffix(consecutiveFailures)}`
         });
       }
       scheduleNextTick();
@@ -7854,7 +8060,7 @@ function scheduleClaudeUsageReporting(state, options) {
         logActivity(state, {
           type: "info",
           level: claudeUsageFailureLogLevel(consecutiveFailures),
-          message: `Claude usage reporting failed: ${message}${claudeUsageFailureStreakSuffix(consecutiveFailures)}`
+          message: `Claude usage reporting failed: ${message}${failureStreakSuffix(consecutiveFailures)}`
         });
         scheduleNextTick();
       }
@@ -7863,6 +8069,90 @@ function scheduleClaudeUsageReporting(state, options) {
   armProbe();
   return rearm;
 }
+var RESOURCE_USAGE_BASE_REPORT_DELAY_MS = 10 * 6e4;
+var RESOURCE_USAGE_REPORT_DELAY_JITTER_FRACTION = 0.2;
+var RESOURCE_USAGE_FAILURE_REESCALATION_TICKS = 6;
+function scheduleResourceUsageReporting(state, options) {
+  const { enabled, warnings } = resolveResourceUsageReportingEnabled(
+    options.resourceUsageReporting,
+    process.env
+  );
+  for (const warning2 of warnings) {
+    logActivity(state, {
+      type: "info",
+      level: "warn",
+      message: `Resource usage reporting: ${warning2}`
+    });
+  }
+  if (!enabled) {
+    logActivity(state, {
+      type: "info",
+      level: "debug",
+      message: "Resource usage reporting is off (--no-resource-usage-reporting)"
+    });
+    return;
+  }
+  const collect = createResourceUsageCollector(homedir3());
+  let consecutiveFailures = 0;
+  const tick = async () => {
+    try {
+      const { usage, warnings: collectWarnings } = await collect();
+      for (const warning2 of collectWarnings) {
+        logActivity(state, {
+          type: "info",
+          level: "debug",
+          message: `Resource usage collection: ${warning2}`
+        });
+      }
+      const result = await reportResourceUsage(state.agentId, state.authHeader, usage);
+      if (result.ok) {
+        if (consecutiveFailures > 0) {
+          logActivity(state, {
+            type: "info",
+            level: "info",
+            message: "Resource usage reporting recovered"
+          });
+        }
+        consecutiveFailures = 0;
+        logActivity(state, {
+          type: "info",
+          level: "debug",
+          message: "Reported resource usage to Evident"
+        });
+      } else {
+        consecutiveFailures++;
+        logActivity(state, {
+          type: "info",
+          level: reportFailureLogLevel(
+            consecutiveFailures,
+            RESOURCE_USAGE_FAILURE_REESCALATION_TICKS
+          ),
+          message: `Failed to report resource usage: ${result.error}${failureStreakSuffix(consecutiveFailures)}`
+        });
+      }
+    } catch (error2) {
+      consecutiveFailures++;
+      const message = error2 instanceof Error ? error2.message : String(error2);
+      logActivity(state, {
+        type: "info",
+        level: reportFailureLogLevel(
+          consecutiveFailures,
+          RESOURCE_USAGE_FAILURE_REESCALATION_TICKS
+        ),
+        message: `Resource usage reporting failed: ${message}${failureStreakSuffix(consecutiveFailures)}`
+      });
+    } finally {
+      state.resourceUsageTimer = setTimeout(
+        () => void tick(),
+        jitteredDelayMs(
+          RESOURCE_USAGE_BASE_REPORT_DELAY_MS,
+          RESOURCE_USAGE_REPORT_DELAY_JITTER_FRACTION
+        )
+      );
+    }
+  };
+  state.resourceUsageTimer = setTimeout(() => void tick(), firstReportDelayMs());
+}
 async function notifyOffline(state) {
   if (!state.agentId || !state.authHeader) return;
   if (!state.connected) {
@@ -7903,6 +8193,10 @@ async function cleanup(state, opts = {}) {
     state.claudeUsageTimer = null;
   }
   state.claudeUsageRearm = null;
+  if (state.resourceUsageTimer) {
+    clearTimeout(state.resourceUsageTimer);
+    state.resourceUsageTimer = null;
+  }
   if (opts.graceful && state.channelDriver) {
     state.channelDriver.stop();
     log2(state, "Draining in-flight channel work before shutdown...");
@@ -7985,6 +8279,7 @@ async function run(options) {
     sessionCleanupTimers: [],
     claudeUsageTimer: null,
     claudeUsageRearm: null,
+    resourceUsageTimer: null,
     authHeader: ""
   };
   setTelemetryAuthProvider(() => ({ authHeader: state.authHeader, agentId: state.agentId }));
@@ -8384,6 +8679,7 @@ async function run(options) {
     }
     scheduleSessionCleanup(state, channelDriver, options);
     state.claudeUsageRearm = scheduleClaudeUsageReporting(state, options);
+    scheduleResourceUsageReporting(state, options);
     if (!interactive || state.json) {
       log2(state, "Driving channel messages...");
     }
@@ -8464,6 +8760,9 @@ program.command("run").description("Connect to Evident and process messages").op
 ).option(
   "--claude-usage-reporting <mode>",
   "Report Claude subscription usage to Evident: auto | on | off (default: auto). Env: EVIDENT_CLAUDE_USAGE_REPORTING"
+).option(
+  "--no-resource-usage-reporting",
+  "Don't report this machine's CPU and memory usage to Evident (reporting is on by default). Env: EVIDENT_RESOURCE_USAGE_REPORTING=off"
 ).option(
   "--enable-file-sync-to <dir>",
   "Allow Evident to write files into this directory (repeatable). Omit to disable file sync entirely.",
@@ -8496,6 +8795,9 @@ program.command("run").description("Connect to Evident and process messages").op
       // Raw string — the resolver in run.ts single-sources parsing
       // (resolveClaudeUsageReportingMode).
       claudeUsageReporting: options.claudeUsageReporting,
+      // Raw value — resolution is single-sourced in run.ts's
+      // resolveResourceUsageReportingEnabled.
+      resourceUsageReporting: options.resourceUsageReporting,
       // Raw values — expansion/validation is single-sourced in run.ts's
       // resolveFileSyncDirectories.
       enableFileSyncTo: options.enableFileSyncTo,