CLI Reference

LongTracer installs a longtracer command for verification and trace inspection.

longtracer check — Verify from the command line

longtracer check "The Eiffel Tower is in Berlin." "The Eiffel Tower is in Paris."

Output:

✗ FAIL  trust=0.50  hallucinations=1
  0/1 claims supported, 1 hallucination(s) detected.

  ✗ The Eiffel Tower is in Berlin. [HALLUCINATION]
    ↳ source: The Eiffel Tower is in Paris.

Options

Option	Description
--json	Output results as JSON
--threshold 0.7	Set verification threshold (default: 0.5)

JSON output

longtracer check "response" "source" --json

{
  "verdict": "FAIL",
  "trust_score": 0.5,
  "summary": "0/1 claims supported, 1 hallucination(s) detected.",
  "hallucination_count": 1,
  "claims": [...]
}

---

Commands

longtracer view

List recent traces:

longtracer view
longtracer view --limit 20
longtracer view --project chatbot-prod

View a specific trace

longtracer view --id <trace_id>

View the most recent trace

longtracer view --last

Export to JSON

longtracer view --export <trace_id>
# Saves to trace_<id>.json

Export to HTML report

longtracer view --html <trace_id>
# Saves to trace_<id>.html — open in any browser

---

Options

Option	Description
--id <trace_id>	View a specific trace by ID
--last	View the most recent trace
--project <name>	Filter traces by project name
--limit N	Max number of traces to list (default: 10)
--export <trace_id>	Export trace to JSON file
--html <trace_id>	Export trace to self-contained HTML report

---

HTML Report

The HTML report is a self-contained single file with:

• Summary cards: verdict, trust score, total duration, claim count
• Waterfall timeline of all pipeline spans
• Per-claim verification table with scores and source evidence
• No external dependencies — works offline

longtracer view --html abc123
open trace_abc123.html

---

Environment Variables

The CLI respects the same environment variables as the SDK:

TRACE_CACHE_BACKEND=sqlite longtracer view
TRACE_PROJECT=chatbot-prod longtracer view