CLI Reference
Paprika provides four CLI commands for managing and inspecting runs.
paprika runs list
List recent execution records.
Usage:
bash
paprika runs list [--limit N] [--trace-dir PATH]Options:
--limit N(default: 20) — maximum runs to display--trace-dir PATH— override default trace directory (~/.paprika/traces/)
Output:
plaintext
Run ID Agent Started Status Steps
───────────────────────────────────────────────────────────────────────────────────────
abc123def456 researcher 2024-01-15 14:32:10 success 3
xyz789abc123 researcher 2024-01-15 14:31:05 success 5Fields:
Run ID— Unique identifier for the runAgent— Agent name (from@runtime.agent(name=...))Started— Execution start timeStatus—success,error, orpolicy_violationSteps— Total number of steps executed
Runs are sorted by most recent first.
---
paprika runs inspect
Display full details of a single execution record.
Usage:
bash
paprika runs inspect [--verbose] [--trace-dir PATH] Options:
<run-id>(required) — run ID (full UUID or prefix match)--verbose(optional) — show full input/output JSON--trace-dir PATH— override default trace directory
Output:
plaintext
Record ID: abc123def456
Agent: researcher
Started: 2024-01-15 14:32:10 UTC
Ended: 2024-01-15 14:32:10 UTC
Duration: 125ms
Status: success
Total tokens: 142
Total steps: 3
Step 0: llm_call (gpt-4o)
Provider: openai
Model: gpt-4o
Input hash: a1b2c3d4e5f6g7h8
Tokens: 20
Duration: 50ms
Step 1: tool_call (search)
Tool: search
Input hash: i9j0k1l2m3n4o5p6
Duration: 35ms
Step 2: llm_call (gpt-4o)
Provider: openai
Model: gpt-4o
Input hash: q7r8s9t0u1v2w3x4
Tokens: 122
Duration: 40msWith --verbose, full JSON input and output are displayed.
---
paprika runs diff
Compare two execution records step-by-step.
Usage:
bash
paprika runs diff [--trace-dir PATH] Options:
<run-id-a>(required) — first run ID<run-id-b>(required) — second run ID--trace-dir PATH— override default trace directory
Output:
plaintext
Comparing run-id-a vs run-id-b
Step 0 (llm_call): MATCH
Hash: a1b2c3d4e5f6g7h8
Step 1 (tool_call): MATCH
Hash: i9j0k1l2m3n4o5p6
Step 2 (llm_call): MISMATCH
Expected hash: q7r8s9t0u1v2w3x4
Actual hash: y9z0a1b2c3d4e5f6Interpretation:
MATCH— Same input hash (identical behavior)MISMATCH— Different input hashes (behavioral divergence)
Use this to identify where two runs diverged. Common case: comparing original vs replayed runs to catch regressions.
---
paprika ui
Start the local browser UI for trace inspection.
Usage:
bash
paprika ui [--port PORT] [--trace-dir PATH] [--no-open]Options:
--port PORT(default: 8787) — port to bind to--trace-dir PATH— override default trace directory--no-open— do not auto-open browser
Output:
plaintext
Starting Paprika UI...
Listening on http://127.0.0.1:8787/
Opening browser...The UI opens at http://127.0.0.1:8787/ and displays:
- Runs List (
/) — table of recent runs - Run Detail (
/runs/:run_id) — execution timeline with step details
Requirements:
paprika[ui]extra:pip install paprika[ui]- Requires FastAPI and uvicorn
Security:
- Binds to
127.0.0.1only (local network only, not accessible from the internet) - No authentication (intended for local debugging only)
---
Environment Variables
PAPRIKA_TRACE_DIR — override default trace directory.
bash
PAPRIKA_TRACE_DIR=/tmp/paprika paprika runs list---
Next Steps
- Understand execution records: Execution Records
- Configure trace location and policies: Configuration
- Explore the browser UI: UI