PR Custody

[LOW PRIORITY] feat: add script to populate agent_traces from existing S3 files

@AgentWrapperchecks n/achecks…feat/populate-agent-traces-script → next1 files · +484 −0updated 3mo ago

▸Description· 1 comment

Summary

Standalone script to backfill the agent_traces column from existing S3 event files.

This PR contains only the script - no DB migrations or code changes. Run manually after the agent_traces column exists.

Usage

# Audit mode (read-only, shows what would be done)
uv run infra/scripts/populate_agent_traces.py --env staging --mode audit

# With limit (for testing)
uv run infra/scripts/populate_agent_traces.py --env staging --mode audit --limit 10

# Write mode (updates DB)
uv run infra/scripts/populate_agent_traces.py --env staging --mode write

What It Does

Scans S3 bucket for existing event files
Parses paths to extract: workflow_id, run_number, exec_id, agent_name
Updates agent_traces JSONB column with file metadata

Staging Audit Results

Found 4693 event files
Parsed 4693 files across 69 workflows

🤖 Generated with Claude Code

@AgentWrapper4mo ago

Fixed in latest commit. Changed regex from ([^_]+)_events\.jsonl to (.+)_events\.jsonl to handle exec_ids that may contain underscores.

loading diff…