Praison AI

Attribution CLI

Track and trace agent execution from the command line

The praisonai-ts CLI exposes attribution and trace identifiers for debugging.

Diagram

CLI

Agent

Trace Headers

Quick Start

  1. Simple Usage
    bash
    praisonai-ts chat "Hello" --verbose
  2. With Configuration
    bash
    praisonai-ts chat "Hello" --session my-session --json

Track agent execution with attribution headers and tracing using the PraisonAI CLI.

Commands

View Trace Information

bash
# Run with tracing enabled
praisonai-ts llm trace "Hello world" --model openai/gpt-4o-mini

# Show attribution headers in output
praisonai-ts llm trace "Hello" --verbose

Set Custom Attribution

bash
# Custom agent ID
praisonai-ts chat "Hello" --agent-id my-custom-agent

# Custom session ID
praisonai-ts chat "Hello" --session my-session-123

# Custom run ID
praisonai-ts chat "Hello" --run-id run-abc-123

# All custom attribution
praisonai-ts chat "Hello" \
  --agent-id my-agent \
  --session my-session \
  --run-id my-run \
  --trace-id my-trace

Multi-Agent Tracing

bash
# Run workflow with tracing
praisonai-ts workflow run ./workflow.yaml --trace

# Show agent attribution for each step
praisonai-ts workflow run ./workflow.yaml --trace --verbose

Options

OptionDescription
--agent-idCustom agent identifier
--sessionSession ID for conversation continuity
--run-idCustom run identifier
--trace-idTrace ID for distributed tracing
--traceEnable tracing output
--verboseShow detailed attribution info
--jsonOutput attribution as JSON

Examples

Basic Tracing

bash
$ praisonai-ts llm trace "What is 2+2?" --model openai/gpt-4o-mini

✓ Response: 2 + 2 = 4

Attribution:
  Agent ID: agent_abc123
  Run ID: run_xyz789
  Session ID: session_def456
  Backend: ai-sdk
  Duration: 234ms

JSON Attribution Output

bash
$ praisonai-ts llm trace "Hello" --json
json
{
  "success": true,
  "data": {
    "response": "Hello! How can I help you today?",
    "attribution": {
      "agentId": "agent_abc123",
      "runId": "run_xyz789",
      "sessionId": "session_def456",
      "traceId": null,
      "backend": "ai-sdk",
      "provider": "openai",
      "model": "gpt-4o-mini"
    }
  },
  "meta": {
    "duration_ms": 234,
    "tokens": {
      "prompt": 12,
      "completion": 8,
      "total": 20
    }
  }
}

Workflow Tracing

bash
$ praisonai-ts workflow run ./pipeline.yaml --trace --verbose

=== Workflow: data-pipeline ===

Step 1: extract
  Agent ID: extractor
  Run ID: run_step1_abc
  Status: ✓ Complete (1.2s)

Step 2: transform
  Agent ID: transformer
  Run ID: run_step2_def
  Status: ✓ Complete (0.8s)

Step 3: load
  Agent ID: loader
  Run ID: run_step3_ghi
  Status: ✓ Complete (0.5s)

=== Workflow Complete ===
Total Duration: 2.5s

Session Continuity

bash
# First message - creates session
$ praisonai-ts chat "My name is John" --session user-123
Session: user-123
Response: Nice to meet you, John!

# Second message - continues session
$ praisonai-ts chat "What is my name?" --session user-123
Session: user-123
Response: Your name is John.

Custom Attribution for Debugging

bash
# Tag requests for debugging
$ praisonai-ts chat "Process this data" \
  --agent-id data-processor \
  --run-id debug-run-001 \
  --trace-id incident-12345 \
  --verbose

Attribution Headers:
  X-Agent-Id: data-processor
  X-Run-Id: debug-run-001
  X-Trace-Id: incident-12345
  X-Session-Id: session_auto_xyz

Response: ...

Parallel Agent Tracing

bash
$ praisonai-ts workflow run ./parallel-agents.yaml --trace

=== Parallel Execution ===

[agent-1] Starting... (Run: run_p1_abc)
[agent-2] Starting... (Run: run_p2_def)
[agent-3] Starting... (Run: run_p3_ghi)

[agent-2] Complete (0.9s)
[agent-1] Complete (1.1s)
[agent-3] Complete (1.3s)

=== All Agents Complete ===

Environment Variables

bash
# Enable tracing globally
export PRAISONAI_TRACE=true

# Set default agent ID prefix
export PRAISONAI_AGENT_PREFIX=myapp

# Enable verbose attribution logging
export PRAISONAI_VERBOSE=true

Scripting Examples

Capture Attribution

bash
#!/bin/bash
# capture-trace.sh

result=$(praisonai-ts llm trace "$1" --json)

agent_id=$(echo "$result" | jq -r '.data.attribution.agentId')
run_id=$(echo "$result" | jq -r '.data.attribution.runId')
duration=$(echo "$result" | jq -r '.meta.duration_ms')

echo "Agent: $agent_id"
echo "Run: $run_id"
echo "Duration: ${duration}ms"

Log Attribution to File

bash
#!/bin/bash
# log-traces.sh

LOG_FILE="traces.log"

praisonai-ts chat "$1" --json | jq '{
  timestamp: now | todate,
  agent: .data.attribution.agentId,
  run: .data.attribution.runId,
  duration: .meta.duration_ms
}' >> "$LOG_FILE"

Monitor Multi-Agent Workflow

bash
#!/bin/bash
# monitor-workflow.sh

praisonai-ts workflow run ./workflow.yaml --trace --json | while read -r line; do
  agent=$(echo "$line" | jq -r '.attribution.agentId // empty')
  status=$(echo "$line" | jq -r '.status // empty')

  if [ -n "$agent" ]; then
    echo "[$(date +%H:%M:%S)] $agent: $status"
  fi
done

Integration with Observability Tools

Export to OpenTelemetry

bash
# Set OTLP endpoint
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

# Run with telemetry
praisonai-ts chat "Hello" --telemetry

Export to Logging Service

bash
# Pipe JSON output to logging service
praisonai-ts llm trace "Hello" --json | \
  curl -X POST -H "Content-Type: application/json" \
  -d @- https://logs.example.com/ingest

Exit Codes

CodeDescription
0Success
1General error
2Invalid arguments

Best Practices

Use consistent session IDs

for conversation continuity

Log run IDs

for debugging and auditing

Enable tracing in production

for observability

Use meaningful agent IDs

for easier debugging

Export traces

to observability platforms

Related

Attribution

SDK documentation

 Telemetry CLI

Telemetry CLI

 Agent CLI

Single-agent CLI