This pull request introduces significant improvements to how Ollama endpoints are resolved and passed as global variables to Langflow and MCP services, ensuring more reliable container-to-container connectivity. The main theme is replacing static localhost transformations with dynamic endpoint resolution using the new FlowsService.resolve_ollama_url method. The changes also update several service constructors to accept a flows_service instance, enabling consistent endpoint resolution across the codebase.

Ollama endpoint resolution improvements:

• Added FlowsService.resolve_ollama_url, which dynamically probes candidate container hosts and caches the resolved Ollama endpoint for improved reliability when running in containerized environments.
• Updated transform_localhost_url in utils/container_utils.py to simplify its logic and remove container-specific arguments, delegating advanced resolution to FlowsService.
• The Ollama endpoint is now resolved using FlowsService.resolve_ollama_url in global variable updates, MCP server updates, and header construction, replacing the previous static transformation. [1] [2]

Service constructor and dependency updates:

• Modified constructors for ChatService and LangflowFileService to accept an optional flows_service parameter, enabling them to use the new dynamic endpoint resolution. [1] [2]
• Updated initialize_services in main.py to instantiate and pass flows_service to dependent services, ensuring consistent endpoint resolution throughout the application.

Global variable and credential propagation:

• Changed all relevant update and onboarding functions to accept and propagate flows_service, ensuring Ollama endpoint resolution is performed consistently whenever global variables or provider credentials are set. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

Configuration enhancements:

• Added a resolved_endpoint field to the OllamaConfig class to cache the resolved Ollama endpoint, reducing redundant network probing and improving performance.

These changes collectively make Ollama connectivity more robust and maintainable, especially in containerized deployments, by centralizing and automating endpoint resolution logic.