Skip to content

Dataset · semianalysisai

Claude Code session logs by SemiAnalysis

semianalysisai/cc-traces-weka-062126

Filtered traces of the Claude Code agent CLI without images and service classifier calls.

What's inside

The WekaTrace corpus is collected from Claude Code proxy traces (CLI version not lower than 2.1.139) and includes less than 1,000 text entries. The sample includes long agent sessions (from 20 requests to Anthropic, up to 10 parallel subagent groups), from which images and short technical queries like header generation were cut out in advance.

What you can do with it

  • SFT of code models on multi-step agent trajectories
  • Analyzing tool call patterns and subagent work in the CLI
Tasks
Text generation
Size
up to 1K rows
Downloads this month
102.6K
Likes
22

How to load it

from datasets import load_dataset

ds = load_dataset("semianalysisai/cc-traces-weka-062126")