tracellm 0.2.1

Local-first observability for LLM applications

tracellm

Lightweight tracing for LLM applications. One decorator — every API interaction logged locally, queryable from your terminal.

No backend. No signup. Nothing leaves your machine.

Install

pip install tracellm

Usage

from tracellm import trace
import groq

client = groq.Groq(api_key="your-key")

@trace
def ask(model, messages):
    return client.chat.completions.create(model=model, messages=messages)

ask(
    model="llama-3.1-8b-instant",
    messages=[{"role": "user", "content": "Explain black holes in one line"}]
)
...

Output on query:

 --- Trace ---
  Model: llama-3.1-8b-instant
  Prompt: Explain black holes in one line
  Response: A black hole is a region where gravity...
  Tokens: 43
  Latency: 0.847
  Status: success
  Timestamp: 2026-04-03 19:46:27
-------------
...

That's it. Every call is traced automatically.

Query traces from terminal

python -m tracellm.cli --Status failed
python -m tracellm.cli --Latency 2.0
python -m tracellm.cli --Model llama-3.1-8b-instant
python -m tracellm.cli --Status failed --Latency 1.5
...

What gets captured

• Model, prompt, response
• Tokens used, latency, finish reason
• Error type and message on failures
• Timestamp for every call

Limitations

Storage is append-only JSON lines. Latency query supports >= for latency, exact match for everything else. Early days.

Roadmap

• Binary storage for faster querying at scale
• Cost calculation per model
• Terminal dashboard