A testing and debugging infrastructure for the autothread system that enables rapid iteration in Val Town's deploy-to-prod environment.

1. Set environment variables:

   ADMIN_TOKEN=your-secret-token
   ENABLE_TEST_API=true
   

2. Test in sandbox mode (default - no side effects):

   curl -X POST https://your-val.val.run/api/autothread/debug/run \
     -H "Authorization: Bearer $ADMIN_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{"mode": "plan"}'

• Default: mode=plan, namespace=sandbox
• live + prod requires confirm: "LIVE_PROD"
• Resetting prod is blocked
• Caps are enforced server-side (max 3 iterations, 5 threads)

Trigger an autothread run with custom configuration.

{
  "mode": "plan|dry_run|live",
  "namespace": "sandbox|prod",
  "enableAI": false,
  "allowlist": ["channel_id_1", "channel_id_2"],
  "maxChannels": 1,
  "maxThreads": 2,
  "iterations": 1,
  "durationMs": 10000,
  "confirm": "LIVE_PROD"  // Required for live+prod
}

Response includes:

• result.plannedActions[] - What would/did happen
• result.events[] - Detailed event log
• result.threadsCreated, result.errorCount

Returns channels, recent processed messages, and today's stats.

List recent runs with their status.

Get detailed events for a specific run.

Reset all sandbox state.

{"namespace": "sandbox", "confirm": "RESET_SANDBOX"}

Reset a channel's last_message_id cursor.

{"namespace": "sandbox", "channelId": "123", "lastMessageId": null}

Clear processed messages with filters.

{"namespace": "sandbox", "channelId": "123", "status": "dry_run"}

Test Discord API connectivity. Returns all guild channels.

Fetch messages from a channel.

Test gate evaluation on a message.

{"content": "Hello world, this is a test message", "isBot": false}

or

{"channelId": "123", "messageId": "456"}

Test deterministic thread naming.

{"content": "This is a long message that will be truncated for the thread name"}

Test AI thread naming (requires OPENAI_API_KEY).

{
  "content": "Can someone help me with Rust lifetimes?",
  "context": [
    {"author": "alice", "content": "I'm stuck on this borrow checker error"},
    {"author": "bob", "content": "What does the error say?"}
  ]
}

# See what the system would do
curl -X POST .../debug/run -d '{"mode":"plan"}' -H "Authorization: Bearer $TOKEN"

# Check planned actions
# Look at result.plannedActions

# Run with DB writes but no Discord
curl -X POST .../debug/run -d '{"mode":"dry_run"}' -H "..."

# Inspect state
curl .../debug/state -H "..."

# Reset and try again
curl -X POST .../debug/reset -d '{"confirm":"RESET_SANDBOX"}' -H "..."

# Get channel IDs
curl .../debug/discord/channels -H "..."

# Run on specific channel
curl -X POST .../debug/run -d '{"allowlist":["123456789"]}' -H "..."

# Why was this message skipped?
curl -X POST .../debug/eval-message \
  -d '{"channelId":"123","messageId":"456"}' -H "..."

# Test with synthetic content
curl -X POST .../debug/eval-message \
  -d '{"content":"!help","isBot":false}' -H "..."
# Will show: "Starts with command prefix (!, /, .)"

curl -X POST .../debug/generate-ai-name \
  -d '{"content":"Question about async/await","context":[{"content":"Previous discussion"}]}' \
  -H "..."

backend/autothread/
├── index.ts      # Re-exports
├── types.ts      # Type definitions, config defaults
├── store.ts      # Database operations with namespace support
├── logic.ts      # Core processing logic (gates, AI, threading)
└── runner.ts     # High-level runner used by cron and debug

The cron job (autothread.cron.ts) and debug endpoint (autothread-debug.http.ts) both use the same underlying runAutothread() function, ensuring consistent behavior.