I'd be happy to help you check the full prompt passed to the LLM for debugging and monitoring! You can do this by enabling verbose prompt logging or using specific API endpoints. Here’s how: • Enable Verbose Prompt Logging: If you're working with

llama.cpp

, you can turn on the

verbose_prompt

flag in the

main.cpp

file or simply run the binary with the

--verbose-prompt

flag. This will log the entire prompt along with tokenization details and token IDs, giving you a clear picture. • Use the /apply-template API Endpoint: When

llama.cpp

is running in server mode, you can call the

/apply-template

endpoint to see the full prompt before generation. This is especially handy for debugging chat templates and tool formatting. • LiveKit Agents: Although full prompt logging isn’t directly available here, user input and context messages are captured in OpenTelemetry traces and event logs. If you need more details, you can extend the logging accordingly. If you'd like, I can guide you through detailed steps for enabling verbose logging or using the API based on your specific setup. Just let me know! Sources: VoicePipelineAgent | LiveKit Docs | tools/main/main.cpp | livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/llm.py