trace-field-notes / README.md
JacobLinCool's picture
feat: improve qwen zerogpu ux
7c8120d verified
|
Raw
History Blame
2.28 kB
metadata
title: Trace Field Notes
colorFrom: green
colorTo: gray
sdk: gradio
sdk_version: 5.50.0
app_file: app.py
pinned: false
license: mit
hf_oauth: true
hf_oauth_scopes:
  - inference-api
hf_oauth_expiration_minutes: 480

Trace Field Notes

Trace Field Notes turns coding-agent session logs into qualitative field reports.

Upload a Codex, Claude Code, or Pi Agent JSONL trace. The app ignores raw tool telemetry by default and analyzes only the agent's visible narrative messages: what it planned, where it got stuck, how it detoured, how it recovered, and how it claimed completion.

Built for the Build Small Hackathon as a Gradio app. The default engine is the quick Qwen3.5 9B model-assisted path on ZeroGPU, with a verified deterministic codebook analyzer as the always-available recovery path. The app also exposes nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 through Hugging Face Inference Providers when the user signs in with Hugging Face OAuth.

Run Locally

python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python app.py

Test

python3.11 -m unittest discover -s tests

Analysis Engines

  • Quick small-model assist: Qwen3.5 9B: default model-assisted memo.
  • NVIDIA Nemotron 3 Nano 30B-A3B assist: uses Nemotron through the signed-in user's inference-api OAuth scope.
  • Deterministic field notes: local, no model dependency.

If a selected model is unavailable or the user is not signed in, the report records the reason in model notes and returns the deterministic analysis instead of failing the whole Space.

The Gradio endpoint is decorated with @spaces.GPU so the app can run on Hugging Face ZeroGPU hardware. The deterministic path still works without model weights; ZeroGPU only supplies the runtime contract and queueing surface.

Agent Session Locations

# Codex
ls ~/.codex/sessions

# Claude Code
ls ~/.claude/projects

# Pi Agent
ls ~/.pi/agent/sessions

Privacy

Agent traces can contain prompts, tool inputs, command outputs, local file paths, screenshots, secrets, private source code, and personal data. Review and redact before uploading or sharing publicly. The app defaults to basic regex redaction and exports only a redacted narrative text file.