Add timestamps to llama-server logging

[nialllo]

I find it annoying to check whether a llama-server is active in my home lab.

The latest time stamp in the log isn't visible because it doesn't exist.

In python the logger defaults to time stamps for this reason.

Add time stamps to llama-server output.

Otherwise one must poke around in htop or task manager to try to infer activity.

[ggerganov]

This is already supported - add these to your env:

#export LLAMA_LOG_COLORS=1
export LLAMA_LOG_PREFIX=1
export LLAMA_LOG_TIMESTAMPS=1

[TFWol]

What about for more typical timestamps?
ex:
[2025-11-02 05:29:00 UTC] - slot update_slots: id 0 | task 886 | prompt done, n_tokens = 372, batch.n_tokens = 362

[The-LittleTeapot]

+ 1

[rbosneag]

+1

What about for more typical timestamps? ex: [2025-11-02 05:29:00 UTC] - slot update_slots: id 0 | task 886 | prompt done, n_tokens = 372, batch.n_tokens = 362

+1

[ThePowerOfMonkeys]

Agreed, having a properly formatted timestamp on logging is really important, especially if this is to be used as a production backend where we need to know what's going on at what time?

[gaeljw]

Agree as well that we should have an option to have the actual datetime rather than a timestamp since start

[shiftySenpai]

I got timestamps to work with this:

--log-timestamps 2>&1 | tee >(sed 's/^/['"$(date +'%Y-%m-%d %H:%M:%S')"'] /' >> logs/llama-server-gpt-oss-120b.log)

Adjust the timestamp notation and logging location and filename to meet your needs.

[ozturkoktay]

This is how I get the real-time logs:

--log-timestamps 2>&1 | ts "[%%Y-%%m-%%d %%H:%%M:%%S]" >> logs/llamacpp-server.log

[M0ns1gn0r]

The current timestamp format is unreadable.