ArtificialAnalysisWorkflow

What it is

• A workflow that fetches model data from the Artificial Analysis API, optionally filters the results to only providers that have corresponding local agent modules, and saves the result as a timestamped JSON file under a fixed storage path.

Public API

• ArtificialAnalysisWorkflowConfiguration (WorkflowConfiguration, dataclass)
  • api_key: str: Artificial Analysis API key (required).
  • base_url: str = "https://artificialanalysis.ai/api/v2": API base URL.
• ArtificialAnalysisWorkflowParameters (WorkflowParameters)
  • endpoint: str = "llms": Endpoint to fetch (e.g., "llms" or media endpoints like "text-to-image").
  • include_categories: bool = False: Adds include_categories=true query param for media endpoints.
  • validate_agents_only: bool = True: If True, filters models to those with corresponding local agent modules.
• ArtificialAnalysisWorkflow (Workflow)
  • __init__(configuration: ArtificialAnalysisWorkflowConfiguration): Stores configuration.
  • run_workflow(parameters: ArtificialAnalysisWorkflowParameters) -> dict: Fetches API data, optionally filters, writes JSON to disk, returns summary dict.
  • as_tools() -> list[BaseTool]: Exposes a LangChain StructuredTool named artificial_analysis_data_fetch.
  • as_api(...) -> None: Present but intentionally does nothing (returns None).

Internal helpers (not public API): _fetch_models_data, _get_modules_with_agents, _filter_data_for_valid_modules, _extract_module_name_from_model.

Configuration/Dependencies

• Network/API
  • Uses requests.get with header x-api-key: <api_key>.
  • URL logic:
    • If endpoint == "llms": .../data/llms/models
    • Else: .../data/media/{endpoint}
• Filesystem
  • Scans for modules with agents under: src/core/modules/<module>/agents/*Agent.py
  • Saves output JSON under:
    • storage/datastore/core/modules/abi/ArtificialAnalysisWorkflow/<timestamp>_<endpoint>_data.json
• Key libraries
  • requests, pydantic, langchain_core.tools, naas_abi_core.workflow

Usage

from naas_abi.workflows.ArtificialAnalysisWorkflow import (
    ArtificialAnalysisWorkflow,
    ArtificialAnalysisWorkflowConfiguration,
    ArtificialAnalysisWorkflowParameters,
)
 
wf = ArtificialAnalysisWorkflow(
    ArtificialAnalysisWorkflowConfiguration(api_key="YOUR_ARTIFICIAL_ANALYSIS_API_KEY")
)
 
result = wf.run_workflow(
    ArtificialAnalysisWorkflowParameters(
        endpoint="llms",
        include_categories=False,
        validate_agents_only=True,
    )
)
 
print(result["status"], result["output_file"])

Caveats

• Filtering via validate_agents_only=True depends on the local repository layout existing at src/core/modules and containing agents/*Agent.py files; if the path is missing, no modules are considered valid.
• as_api(...) is a no-op and does not register any routes.
• Provider-to-module matching uses a fixed mapping (e.g., "openai" -> "chatgpt", "anthropic" -> "claude") and fallbacks based on provider/model name heuristics; mismatches may cause models to be filtered out.