Skip to main content

OpenRouterAPIIntegration

What it is​

  • A small integration client for the OpenRouter API.
  • Wraps several OpenRouter endpoints (models, providers, credits, analytics, beta responses).
  • Includes:
    • Request handling with bearer auth
    • A 1-day filesystem cache for API requests
    • Optional JSON persistence of model lists to object storage (via StorageUtils)

Public API​

OpenRouterAPIIntegrationConfiguration​

Dataclass configuration for the integration.

  • Fields
    • api_key: str — OpenRouter API key (Bearer token)
    • object_storage: ObjectStorageService — backing storage used by StorageUtils for JSON saving
    • base_url: str = "https://openrouter.ai/api/v1" — API base URL
    • datastore_path: str = "openrouter" — base path used when saving JSON (e.g., models)

OpenRouterAPIIntegration​

Integration client.

  • create_response(input_prompt: str, tools: Optional[list[Dict]] = None, model: str = "openai/gpt-4.1-mini", temperature: float = 0.7, top_p: float = 0.9) -> Dict

    • POST /responses
    • Sends a beta “responses” payload with a single user message.
    • Accepts optional tools list.

  • get_user_activity(date: Optional[str] = None) -> Dict

    • GET /activity
    • Query param: date (YYYY-MM-DD, last 30 days)

  • get_remaining_credits() -> Dict

    • GET /credits

  • get_total_models_count() -> Dict

    • GET /models/count

  • list_models(params: Optional[Dict] = None, save_json: bool = True) -> List

    • GET /models
    • Returns the unwrapped list from response["data"] (or []).
    • If save_json=True, saves:
      • All models to: {datastore_path}/models/_all/models.json
      • Models split by owner prefix (before / in model id) to: {datastore_path}/models/{owner}/models.json

  • get_model_parameters(author: str, slug: str) -> Dict

    • GET /parameters with author and slug query params

  • list_providers() -> Dict

    • GET /providers

  • list_api_keys() -> Dict

    • GET /keys

  • get_current_api_key() -> Dict

    • GET /key

as_tools(configuration: OpenRouterAPIIntegrationConfiguration)​

Builds LangChain tools from this integration.

  • Returns a list of langchain_core.tools.StructuredTool:
    • openrouter_list_models → calls integration.list_models()
    • openrouter_list_providers → calls integration.list_providers()

Configuration/Dependencies​

  • HTTP: requests
  • Core integration types: naas_abi_core.integration.integration
    • Uses IntegrationConnectionError on request failures.
  • Caching:
    • Uses CacheFactory.CacheFS_find_storage(subpath="openrouter")
    • _make_request is cached for 1 day with cache key based on method/endpoint/params.
  • Object storage:
    • object_storage: ObjectStorageService is required by configuration.
    • Used by StorageUtils.save_json(...) when list_models(save_json=True).

Usage​

from naas_abi_marketplace.applications.openrouter.integrations.OpenRouterAPIIntegration import (
    OpenRouterAPIIntegration,
    OpenRouterAPIIntegrationConfiguration,
)

# Provide a concrete ObjectStorageService implementation from your environment
object_storage = ...  # ObjectStorageService

config = OpenRouterAPIIntegrationConfiguration(
    api_key="YOUR_OPENROUTER_API_KEY",
    object_storage=object_storage,
)

client = OpenRouterAPIIntegration(config)

# Models
models = client.list_models(save_json=False)
print(len(models))

# Providers
providers = client.list_providers()
print(providers)

# Credits
credits = client.get_remaining_credits()
print(credits)

# Beta responses
resp = client.create_response("Say hello in one sentence.")
print(resp)

Caveats​

  • _make_request caches responses for 1 day; repeated calls with the same method/endpoint/params may return cached data.
  • list_models(save_json=True) writes JSON via StorageUtils to paths derived from datastore_path; ensure object_storage is correctly configured.
  • Cache key uses only method, endpoint, and params (not request body); POST calls with different payloads but same endpoint may collide in cache.