Asoba Ona Documentation
Home Insights Contact Us

Agentic Use

Leverage Ona’s Power Tools through AI Agent Orchestration

Overview

The Ona platform can be intelligently accessed through AsobaCode CLI and AI agent workflows, enabling sophisticated analysis and automation for energy systems. This approach allows you to combine Ona’s forecasting and analysis capabilities with advanced AI reasoning for complex, multi-step energy assessments.

Architecture Overview

graph TB
    subgraph "AsobaCode CLI & Agent Orchestration"
        CLI[AsobaCode CLI Terminal]
        AO[Agent Orchestrator]
        NLP[Natural Language Processor]
        MR[Multi-Model Router]
    end
    
    subgraph "AI Models Layer"
        Claude4[Claude 4 Opus/Sonnet]
        Mistral[Mistral Policy Analysis Model]
        Bedrock[AWS Bedrock RAG/KB]
    end
    
    subgraph "Specialized Agents"
        DataAgent[Data Intake Agent]
        ForecastAgent[Forecasting Agent]
        BenchmarkAgent[Benchmark & Economics Agent]
        ComplianceAgent[Compliance Agent]
        AnalysisAgent[Analysis Agent]
    end
    
    subgraph "Ona Power Tools API (af-south-1)"
        IngestHist[ingestHistoricalLoadData]
        IngestNow[ingestNowcastLoadData]
        Interpolate[interpolateData]
        TrainFC[trainForecaster]
        ReturnFC[returnForecastingResults]
        Weather[weather API]
        Dispatch[electricityDispatch]
        Economics[project_economics]
        MarketFC[marketPriceForecast]
    end
    
    CLI --> AO
    AO --> NLP
    NLP --> MR
    MR --> Claude4
    MR --> Mistral
    MR --> Bedrock
    
    Claude4 --> DataAgent
    Claude4 --> ForecastAgent
    Claude4 --> BenchmarkAgent
    Claude4 --> ComplianceAgent
    Claude4 --> AnalysisAgent
    
    DataAgent --> IngestHist
    DataAgent --> IngestNow
    DataAgent --> Interpolate
    DataAgent --> Weather
    
    ForecastAgent --> TrainFC
    ForecastAgent --> ReturnFC
    
    BenchmarkAgent --> Dispatch
    BenchmarkAgent --> Economics
    BenchmarkAgent --> MarketFC

Getting Started with Agentic Workflows

1. Installation & Setup

# Clone the AsobaCode repository
git clone https://github.com/AsobaCloud/asoba-code.git
cd asoba-code

# Install dependencies
pip install -e .

# Configure for Ona API region
export AWS_DEFAULT_REGION=af-south-1
export ONA_API_KEY=your_api_key_here

# Launch interactive mode
asoba-code --interactive

2. Basic Agent-Powered Analysis

The simplest way to leverage agentic capabilities is through natural language commands:

asoba-code --interactive

Example prompts:

> "Analyze the uploaded solar generation data, train a forecasting model, and provide P50/P90 performance curves with confidence intervals"

> "Process historical load data, perform weather normalization, and generate a comprehensive risk assessment report"

> "Ingest the CSV data, clean it using interpolation, train a forecaster, and compare results against industry benchmarks"

Core Agent Workflows

Data Processing Agent

Purpose: Automated data ingestion, cleaning, and validation

Ona API Integration:

Example Workflow:

# Through natural language instruction to agent:
"Upload this 24-month solar generation CSV, clean any gaps using interpolation, 
and provide a data quality assessment report"

Agent Actions:

  1. Calls ingestHistoricalLoadData with uploaded CSV
  2. Analyzes data quality and identifies gaps
  3. Uses interpolateData to standardize intervals
  4. Applies weather API for normalization context
  5. Generates comprehensive data readiness report

Forecasting Agent

Purpose: Intelligent model training and prediction generation

Ona API Integration:

Example Workflow:

# Agent instruction:
"Train a forecasting model on this plant data and generate P50/P90 curves 
with seasonal variability analysis"

Agent Actions:

  1. Analyzes historical patterns and seasonality
  2. Calls trainForecaster with optimized parameters
  3. Retrieves results via returnForecastingResults
  4. Performs statistical analysis of confidence bands
  5. Generates visualizations and performance metrics

Economics & Benchmark Agent

Purpose: Revenue optimization and market analysis

Ona API Integration:

Example Workflow:

# Agent instruction:
"Analyze this solar project's economic performance, optimize dispatch strategy, 
and provide market risk assessment"

Agent Actions:

  1. Processes project financial parameters
  2. Calls electricityDispatch for revenue optimization
  3. Uses project_economics for comprehensive financial modeling
  4. Integrates marketPriceForecast for risk analysis
  5. Delivers actionable economic insights

Advanced Use Cases

1. Multi-Step Risk Assessment

> "Perform comprehensive insurance risk assessment on this solar dataroom. 
Include underwriting analysis, premium calculation, and policy recommendations."

Agent Workflow:

  1. Document Review: Parse technical specifications and maintenance records
  2. Data Processing: Ingest historical performance data
  3. Forecasting: Generate performance predictions with confidence intervals
  4. Economic Analysis: Model revenue scenarios and risk factors
  5. Risk Scoring: Calculate integrated risk metrics
  6. Report Generation: Compile comprehensive assessment with audit trail

2. Performance Benchmarking

> "Compare this wind farm's performance against industry benchmarks and 
identify optimization opportunities"

Agent Workflow:

  1. Data Ingestion: Process plant operational data
  2. Normalization: Apply weather corrections
  3. Benchmarking: Compare against industry standards
  4. Gap Analysis: Identify performance deviations
  5. Optimization: Recommend operational improvements

3. Real-Time Monitoring Setup

> "Set up continuous monitoring for this portfolio with automated alerts 
for performance deviations"

Agent Workflow:

  1. Stream Configuration: Set up nowcast data ingestion
  2. Baseline Establishment: Create performance thresholds
  3. Alert Rules: Configure deviation triggers
  4. Dashboard Setup: Real-time visualization
  5. Reporting: Automated anomaly reports

Agent Communication Patterns

Sequential Workflows

Agents can chain API calls for complex analysis:

DataAgent → ForecastAgent → EconomicsAgent → ReportAgent

Parallel Processing

Multiple agents can work simultaneously:

DataAgent ∥ WeatherAgent ∥ MarketAgent → AnalysisAgent

Feedback Loops

Agents can iterate and refine based on results:

ForecastAgent → ValidationAgent → (if unsatisfactory) → ForecastAgent

Best Practices

1. Prompt Engineering

Effective prompts include:

Example:

"Analyze the uploaded 18-month solar irradiance and generation data. 
Ensure data quality >95%, train a forecasting model with P50 MAE <5%, 
and generate a technical report with confidence intervals and seasonal analysis."

2. Data Preparation

Best practices:

3. Workflow Optimization

Strategies:

Integration Examples

Python Integration

import asoba_code

# Initialize agent orchestrator
orchestrator = asoba_code.AgentOrchestrator(
    api_key="your_ona_api_key",
    region="af-south-1"
)

# Execute agentic workflow
result = orchestrator.execute(
    prompt="Analyze solar performance and generate forecasts",
    data_sources=["solar_generation.csv", "irradiance_data.csv"],
    output_format="comprehensive_report"
)

CLI Scripting

#!/bin/bash
# Automated analysis pipeline

asoba-code analyze \
  --data "historical_data.csv" \
  --workflow "forecasting" \
  --output "forecast_report.json" \
  --agents "data,forecast,economics"

Monitoring & Debugging

Agent Execution Tracking

All agent workflows provide detailed execution logs:

{
  "workflow_id": "wf_12345",
  "agents_used": ["DataAgent", "ForecastAgent"],
  "api_calls": [
    {
      "endpoint": "ingestHistoricalLoadData",
      "timestamp": "2024-01-15T10:30:00Z",
      "status": "success",
      "s3_key": "processed/data_12345.parquet"
    }
  ],
  "execution_time": "45.2s",
  "quality_metrics": {
    "data_completeness": 0.98,
    "forecast_accuracy": 0.95
  }
}

Error Handling

Agents automatically handle common issues:

Advanced Configuration

Custom Agent Development

Create specialized agents for specific use cases:

class CustomRiskAgent(BaseAgent):
    """Specialized agent for insurance risk assessment"""
    
    def __init__(self):
        super().__init__()
        self.ona_apis = ["ingestHistoricalLoadData", "trainForecaster", "weather"]
    
    async def assess_risk(self, dataroom_path: str) -> RiskAssessment:
        # Custom risk assessment logic
        pass

Multi-Region Deployment

Configure agents for different geographic regions:

# agent_config.yaml
regions:
  af-south-1:
    endpoints: ["ingest", "forecast", "weather"]
    priority: primary
  us-east-1:
    endpoints: ["rag", "policy_analysis"]
    priority: secondary

Support & Resources

Documentation

Community

Training Resources

Ready to get started? Begin with our Quickstart Guide or dive into API Reference for detailed endpoint documentation.

Get Help & Stay Updated

Contact Support

We're constantly improving and want you to be a part of shaping the future of energy policy access and decision-making. If you encounter issues or have suggestions, please reach out to our dedicated support team.

Email Support Join Discord

Subscribe to Updates

* indicates required

© 2025 Asoba Corporation. All rights reserved.