Debugging

Capture raw prompts, responses, and troubleshoot issues.

Quick Debug with Presets

import { HookPresets } from 'llmist';

// Full monitoring
.withHooks(HookPresets.monitoring({ verbose: true }))

// Specific debugging
.withHooks(HookPresets.logging({ verbose: true }))
.withHooks(HookPresets.errorLogging())
.withHooks(HookPresets.timing())

Symptom	Recommended Preset
Agent not responding	monitoring()
Unexpected behavior	logging({ verbose: true })
High costs	tokenTracking()
Errors/crashes	errorLogging()
Slow performance	timing()

Capture Raw Requests

.withHooks({
  observers: {
    onLLMCallStart: async (ctx) => {
      console.log('Model:', ctx.options.model);
      console.log('Messages:', ctx.options.messages);
    },
    onLLMCallComplete: async (ctx) => {
      console.log('Response:', ctx.rawResponse);
      console.log('Tokens:', ctx.usage);
    },
  },
})

Debug Gadget Execution

.withHooks({
  observers: {
    onGadgetExecutionStart: async (ctx) => {
      console.log(`[${ctx.gadgetName}] Starting`);
      console.log('Parameters:', ctx.parameters);
    },
    onGadgetExecutionComplete: async (ctx) => {
      console.log(`[${ctx.gadgetName}] ${ctx.executionTimeMs}ms`);
      if (ctx.error) console.error('Error:', ctx.error);
    },
  },
})

Custom Logger

For detailed logger configuration, see Logging Reference.

import { createLogger } from 'llmist';

const logger = createLogger({ minLevel: 'debug' });
.withLogger(logger)

CLI Debugging

bun

bunx @llmist/cli complete "test" --model haiku 2>&1 | tee debug.log
bunx @llmist/cli agent "test" --gadget ./tool.ts 2>&1

npm

npx @llmist/cli complete "test" --model haiku 2>&1 | tee debug.log
npx @llmist/cli agent "test" --gadget ./tool.ts 2>&1

See Also

• Hooks Guide - All hook types
• Troubleshooting - Common issues