joelclaw（joelclaw 搜索）

joelclaw — CLI & Event Bus

The joelclaw CLI is the primary operator interface to the entire joelclaw system: event bus (Inngest), gateway, observability (OTEL), agent loops, subscriptions, and more. Built with @effect/cli, returns HATEOAS JSON envelopes.

If the CLI crashes, that's the highest priority fix.

Binary: ~/.bun/bin/joelclaw Source: ~/Code/joelhooks/joelclaw/packages/cli/ Build: bun build packages/cli/src/cli.ts --compile --outfile ~/.bun/bin/joelclaw

Architecture

┌─ Colima VM (VZ framework, aarch64) ────────────────────────────────┐
│  Talos v1.12.4 → k8s v1.35.0 (single node, namespace: joelclaw)  │
│                                                                     │
│  inngest-0             StatefulSet   ports 8288 (API), 8289 (dash) │
│  redis-0               StatefulSet   port 6379                     │
│  typesense-0           StatefulSet   port 8108                     │
│  restate-0             StatefulSet   ports 8080, 9070, 9071        │
│  system-bus-worker     Deployment    port 3111 (110+ functions)    │
│  restate-worker        Deployment    port 9080                     │
│  dkron-0               StatefulSet   port 8080                     │
│  minio-0               StatefulSet   ports 9000, 9001              │
│  docs-api              Deployment    port 3838                     │
│  livekit-server        Deployment    ports 7880, 7881              │
│  bluesky-pds           Deployment    port 3000                     │
│                                                                     │
│  ⚠️ Inngest service named inngest-svc (not inngest)                │
│     k8s auto-injects INNGEST_PORT env collision otherwise          │
└─────────────────────────────────────────────────────────────────────┘
        ↕ NodePort on localhost

Gateway daemon (always-on pi session, Redis event bridge)
NAS "three-body" (ASUSTOR, 10GbE NFS, 64TB RAID5 + NVMe cache)
Vault ~/Vault (Obsidian, PARA method — ADRs, system log, contacts)

Inngest event key: 37aa349b89692d657d276a40e0e47a15 k8s manifests: ~/Code/joelhooks/joelclaw/k8s/

CLI Command Reference

Health & Status

joelclaw status                            # Health: server + worker + k8s pods
joelclaw jobs status                       # Unified runtime workload snapshot (queue + Restate + Dkron + Inngest)
joelclaw inngest status                    # Inngest server details
joelclaw functions                         # List all 110+ registered functions
joelclaw refresh                           # Force re-register with Inngest server

Send Events

joelclaw send "event/name" --data '{"key":"value"}'
joelclaw send "pipeline/video.download" --data '{"url":"https://youtube.com/watch?v=XXX"}'
joelclaw send "agent/story.start" --data '{"prdPath":"/abs/path/prd.json","storyId":"S-1"}'

View Runs

joelclaw runs                              # Recent 10
joelclaw runs --count 20 --hours 24        # More runs, wider window
joelclaw runs --status FAILED              # Just failures
joelclaw run <RUN_ID>                      # Step trace + errors for one run

View Events

joelclaw events                            # Last 4 hours
joelclaw events --prefix memory/ --hours 24
joelclaw events --prefix agent/ --hours 24
joelclaw events --count 50 --hours 48

Logs

joelclaw logs                              # Worker stdout (default 30 lines)
joelclaw logs errors                       # Worker stderr (stack traces)
joelclaw logs server                       # Inngest k8s pod logs
joelclaw logs server -n 50 --grep error    # Filtered server errors
joelclaw logs worker --grep "observe"      # Grep worker logs

Structured Log Writes

joelclaw log write --action configure --tool cli --detail "updated capability adapter config" --reason "ADR-0169 phase 1"

log writes canonical structured entries (slog backend). logs remains runtime log read/analyze.

Secrets

joelclaw secrets status
joelclaw secrets lease <name> --ttl 15m
joelclaw secrets revoke <lease-id>
joelclaw secrets revoke --all
joelclaw secrets audit --tail 50
joelclaw secrets env --dry-run

Notify

joelclaw notify send "Worker restarted and healthy" --priority normal
joelclaw notify send "Immediate action required" --priority urgent --telegram-only

Capability Adapter Paths (ADR-0169 phase 4)

mail, otel, recall, and subscribe now run through the CLI capability registry/adapter runtime while preserving their existing command UX and JSON envelopes.

Deploy

joelclaw deploy worker                              # dry-run deploy plan
joelclaw deploy worker --restart --execute         # execute worker sync deployment
joelclaw deploy worker --restart --execute --force # force with active runs (disruptive)

Heal

joelclaw heal list
joelclaw heal run RUN_FAILED --phase fix --context '{"run-id":"01ABC"}'           # dry-run
joelclaw heal run RUN_FAILED --phase fix --context '{"run-id":"01ABC"}' --execute # execute

Gateway

joelclaw gateway status                    # Gateway health + session info
joelclaw gateway events                    # Recent gateway events
joelclaw gateway test                      # Send test message through gateway
joelclaw gateway restart                   # Restart gateway daemon
joelclaw gateway stream                    # Live stream gateway events

Workload Rig

joelclaw workload plan "intent" --stages-from stages.json --write-plan plan.json
joelclaw workload run plan.json --stage <stage-id> --tool pi|codex|claude
joelclaw workload dispatch plan.json --to BlueFox --from MaroonReef --send-mail
joelclaw workload sandboxes list
joelclaw workload sandboxes cleanup --expired --dry-run
joelclaw workload sandboxes janitor

Runtime truth:

• workload run admits into Redis, then Restate executes the stage through dagOrchestrator → dagWorker
• dagWorker handlers: shell, infer, microvm
• infer now runs pi inside the restate-worker k8s pod with mounted auth, skills, and identity
• --stages-from is the explicit stage-DAG front door; dependency gates and cycle detection are real

Async runtime monitoring in pi

The loaded pi extension at packages/pi-extensions/inngest-monitor/index.ts now exposes runtime_jobs_monitor alongside inngest_send / inngest_runs.

Use it when you want a pi session to keep watch on the ADR-0217 runtime substrate while you do other work:

• runtime_jobs_monitor {"action":"start","interval":5,"report":true} — start background polling of joelclaw jobs status
• runtime_jobs_monitor {"action":"status"} — inspect the latest runtime snapshot
• runtime_jobs_monitor {"action":"stop"} — stop the poller and send a final follow-up summary

The widget shows runtime health first (queue / Restate / Dkron / Inngest), then any followed Inngest runs underneath. Severity changes and meaningful workload-state changes emit OTEL and hidden follow-up messages for asynchronous report-back, so a healthy-but-now-held backlog does not stay silent.

Observability (OTEL)

joelclaw otel list --hours 1               # Recent telemetry events
joelclaw otel search "error" --hours 24    # Search OTEL events
joelclaw otel stats --hours 24             # Aggregate stats
joelclaw otel emit "action.name" --source codex --component agent-loop --success  # Emit event

System Knowledge (ADR-0199)

joelclaw knowledge search "query"          # Search system knowledge
joelclaw knowledge search "query" --type adr  # Filter by type (adr|skill|lesson|pattern|retro|failed_target)
joelclaw knowledge sync                    # Re-index ADRs + skills from filesystem
joelclaw knowledge clear-failed <target>   # Clear resolved failed targets

Brain/codebase patterns (browsable by agents):

ls ~/Vault/system/brain/codebase/          # List established patterns
cat ~/Vault/system/brain/codebase/<name>.md  # Read a specific pattern

Subscriptions (ADR-0127)

joelclaw subscribe list                    # All feed subscriptions
joelclaw subscribe add <url> [--name N]    # Add a feed
joelclaw subscribe remove <url>            # Remove a feed
joelclaw subscribe check [--url URL]       # Check feeds for new items
joelclaw subscribe summary                 # Summary of recent items

Agent Runtime Validation (ADR-0180)

Use this exact smoke test when validating roster dispatch end-to-end:

joelclaw agent list
joelclaw agent run coder "reply with OK" --timeout 20
joelclaw event <event-id>

Expected signal:

• agent list includes builtin coder, designer, ops
• event shows one Agent Task Run with status: COMPLETED
• run output contains {"status":"completed",...}

Failure handling:

• Unknown agent roster entry: coder means runtime drift, not prompt failure.

- Deploy latest system-bus-worker - Restart host worker process - Re-run the same 3-step smoke

• If Inngest API is unreachable (localhost:8288), recover local control-plane first (Colima/Talos), then retry validation.

Semantic Recall

joelclaw recall "query about past context"  # Search semantic memory

Discovery

joelclaw discover "https://example.com" --context "why this is interesting"

Agent Loops

# Start a loop
joelclaw loop start --project ~/Code/joelhooks/joelclaw \
  --goal "Implement feature X" \
  --context ~/Vault/docs/decisions/0XXX.md \
  --max-retries 2

# Start with existing PRD
joelclaw loop start --project PATH --prd prd.json --max-retries 2

# Monitor
joelclaw loop status <LOOP_ID>
joelclaw loop status <LOOP_ID> -c          # Compact: one line per story
joelclaw loop status <LOOP_ID> -v          # Verbose: criteria, output paths
joelclaw watch <LOOP_ID>                   # Live: polls 15s, exits on completion
joelclaw watch                             # Auto-detects active loop

# Management
joelclaw loop list                         # All loops in Redis
joelclaw loop cancel <LOOP_ID>             # Stop + cleanup
joelclaw loop nuke dead                    # Remove completed loops from Redis

Other Commands

joelclaw sleep [on|off|status]             # Sleep mode for gateway
joelclaw note <text>                       # Quick note to Vault
joelclaw vault read <ref>                  # Resolve/read ADR/project/path refs
joelclaw vault search <query>              # Search vault markdown
joelclaw vault adr list                    # ADR inventory (optionally by status)
joelclaw vault adr collisions              # ADR number collision report
joelclaw vault adr audit                   # ADR health + collision + index checks
joelclaw vault adr rank                            # ADR NRC+novelty ranking
joelclaw skills audit [--deep]             # On-demand skill garden report
joelclaw search <query>                    # Full-text search
joelclaw email [scan|triage]               # Email operations
joelclaw x [post|mentions]                 # X/Twitter operations
joelclaw nas [status|health]               # NAS operations
joelclaw diagnose <topic>                  # System diagnosis
joelclaw langfuse [traces|costs]           # Langfuse analytics
joelclaw deploy worker [--restart] [--execute]
joelclaw heal [list|run]
joelclaw inngest sync-worker [--restart]   # Worker lifecycle

For vault-heavy or ADR-gardening tasks, use the dedicated vault skill.

Output Modes

Most commands support --compact/-c for plain text. Use compact for monitoring. JSON (default) returns HATEOAS envelopes with next_actions.

Story Pipeline (ADR-0155)

3-stage pipeline: implement → prove → judge. Each story runs through the stages with Inngest durability.

# Fire a single story
joelclaw send agent/story.start -d '{
  "prdPath": "/Users/joel/Code/joelhooks/joelclaw/prd.json",
  "storyId": "CFP-2"
}'
# ⚠️ ALWAYS use absolute path for prdPath — worker CWD is packages/system-bus/

PRD format (Zod-validated):

{
  "name": "Project Name",
  "context": {},
  "stories": [
    {
      "id": "STORY-1",
      "title": "What to build",
      "description": "Details",
      "priority": 1,
      "acceptance": ["criterion 1", "criterion 2"],
      "files": ["path/to/relevant/file.ts"]
    }
  ]
}

Critical:

• context must be {} or object — NEVER null or string
• Every story needs priority (number)
• NEVER set retries: 0 on Inngest functions — breaks restart safety (ADR-0156)

Event Types

Pipelines

Memory

Agent Loops

Gateway & Channels

Subscriptions & Discovery

System & Scheduled

Notifications

Debugging Failed Runs

joelclaw runs --status FAILED              # 1. Find the failure
joelclaw run <RUN_ID>                      # 2. Step trace + inline errors
joelclaw logs errors                       # 3. Worker stderr
joelclaw logs server --grep error          # 4. Inngest server errors
joelclaw otel search "error" --hours 1     # 5. OTEL telemetry

Common Failure Patterns

Stale RUNNING forensics (SDK unreachable ghosts)

When joelclaw runs --status RUNNING shows old health jobs that never clear:

1. Use the operator command first

- Preview: joelclaw inngest sweep-stale-runs - Apply (backup + transaction): joelclaw inngest sweep-stale-runs --apply

1. Validate the symptom class

- joelclaw run <run-id> - Look for trace/finalization errors containing Unable to reach SDK URL or EOF writing request to SDK.

1. Treat list vs detail disagreements as a known mask issue

- runs list can show stale metadata. - run detail + trace/history is the source of truth.

1. Raw runtime DB edits are last resort only

- Inngest state is in k8s StatefulSet PVC: inngest-0:/data/main.db. - Backup first: kubectl -n joelclaw exec inngest-0 -- sqlite3 /data/main.db '.backup /data/main.db.pre-sweep-<ts>.sqlite'.

1. Terminalize stale runs with full contract, not partial edits

- Insert missing history.type='FunctionCancelled' for stale runs. - Ensure function_finishes row exists. - Then set trace_runs.status=500 (cancelled) for stale candidates.

1. Verify after mutation

- joelclaw run <run-id> should resolve terminal state. - joelclaw runs --status RUNNING should only show genuinely active runs.

Never mutate main.db without a point-in-time backup.

Deploying Worker Changes

Use the publish script — it handles build, push, k8s apply, and rollout:

~/Code/joelhooks/joelclaw/k8s/publish-system-bus-worker.sh

See the sync-system-bus skill for the full deploy workflow.

Key Paths

Building the CLI

cd ~/Code/joelhooks/joelclaw
bun build packages/cli/src/cli.ts --compile --outfile ~/.bun/bin/joelclaw

Test after every change:

joelclaw status
joelclaw send --help
joelclaw runs --count 1

CLI commands are in packages/cli/src/commands/, one file per command. Follow the cli-design skill. Heavy deps must be lazy-loaded — top-level import crashes are unacceptable.