Files
aitbc/skills/aitbc-ai-operations.md
aitbc1 ef24a7b834
All checks were successful
Deploy to Testnet / deploy-testnet (push) Successful in 1m5s
Node Failover Simulation / failover-test (push) Successful in 15s
Multi-Node Stress Testing / stress-test (push) Successful in 7s
Cross-Node Transaction Testing / transaction-test (push) Successful in 6s
fix: move skills back to project root for Hermes loading
Skills must be in skills/ (project root) for Hermes to load them via skill_view().
Previous move to docs/skills/ broke skill loading.
2026-05-08 22:33:14 +02:00

3.1 KiB

name, description, category
name description category
aitbc-ai-operations AI job operations for AITBC including job submission, monitoring, results retrieval, resource allocation testing, and AI job monitoring mlops

AITBC AI Operations Skill

Trigger Conditions

Activate when user requests AI operations: job submission, status monitoring, results retrieval, resource allocation testing, or AI job monitoring.

Purpose

Submit, monitor, and optimize AITBC AI jobs with deterministic performance tracking and resource management.

Prerequisites

  • AITBC CLI accessible at /opt/aitbc/aitbc-cli
  • AI services operational (Ollama, coordinator, exchange)
  • Wallet with sufficient balance for job payments
  • Default test wallet: "genesis" (password from /var/lib/aitbc/keystore/.genesis_password)
  • Resource allocation system functional

Operations

Submit AI Job

./aitbc-cli ai-ops submit \
  --wallet <wallet_name> \
  --type <job_type> \
  --prompt <prompt> \
  --payment <payment> \
  --password <password> \
  --rpc-url http://localhost:8006

Check AI Job Status

./aitbc-cli ai-ops status --job-id <job_id> --rpc-url http://localhost:8006

Job Types

  • inference - AI model inference
  • training - AI model training
  • multimodal - Multi-modal AI processing
  • ollama - Ollama-based operations
  • streaming - Streaming AI responses
  • monitoring - AI system monitoring

Resource Allocation

# Allocate resources for AI job
./aitbc-cli resource allocate \
  --agent-id <agent_id> \
  --gpu <gpu_count> \
  --memory <memory_mb> \
  --duration <seconds>

Check Resource Status

./aitbc-cli resource status

List Available Resources

./aitbc-cli resource list

Common Pitfalls

  1. Insufficient Wallet Balance: Check wallet balance before job submission
  2. Invalid Job Type: Ensure job type is valid (inference, training, multimodal, ollama, streaming, monitoring)
  3. Service Unavailability: Verify Ollama, coordinator, and exchange services are running
  4. Resource Allocation Failures: Check GPU and memory availability
  5. Job Timeout: Monitor job progress and adjust timeout if needed
  6. Payment Issues: Ensure payment amount covers job costs
  7. Invalid Job ID: Verify job ID when checking status

Verification Checklist

  • Job submission returns valid job ID
  • Job status shows correct state (submitted, processing, completed, failed)
  • Resource allocation succeeds
  • Job completes within expected timeframe
  • Results retrieved successfully
  • Payment processed correctly

GPU Provider Testing

# Test GPU inference
python3 cli/unified_cli.py ollama gpu-test \
  --wallet genesis \
  --model llama2 \
  --prompt "test prompt" \
  --marketplace-url http://aitbc1:8102

CLI Tool Preference

  • Primary CLI: /opt/aitbc/aitbc-cli is the single CLI entry point
  • Module: cli/unified_cli.py is a module within the CLI tool for marketplace and GPU operations
  • Note: For GPU provider operations, prefer python3 cli/unified_cli.py (verified working with 7 bugs fixed)