Skip to content

feat: usability improvements for embedded intrinsic adapters #930

Closed

Description

@jakelorocco

cc80f51

We can add to the vllm plugin that enables switch models. We should work with the team to see if it makes sense to advertise model capabilities. This would make it much easier to coordinate between the client/server and auto-load the available adapters.

Activity

  1. added
    area/backendsProvider-specific work: Ollama, HF, LiteLLM, OpenAI, Bedrock, vLLM
    area/adapter-functionsGranite adapter functions: framework and adaptiers including RAG, Guardian, Core
    on May 20, 2026
  2. added
    p2Medium/low: minor bugs, niche features, polish, docs, tests, cleanup. Scoped, lower urgency.
    on Jun 18, 2026
  3. planetf1 commented on Oct 20, 2026

    @planetf1
    Contributor

    Closing as stale: quiet since April, and the vLLM capability-advertisement question hasn't moved. If it becomes relevant again (the llama.cpp serving work in #1736 is the likelier venue), happy to reopen or refile with specifics.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area/adapter-functionsGranite adapter functions: framework and adaptiers including RAG, Guardian, Corearea/backendsProvider-specific work: Ollama, HF, LiteLLM, OpenAI, Bedrock, vLLMenhancementNew feature or requestp2Medium/low: minor bugs, niche features, polish, docs, tests, cleanup. Scoped, lower urgency.

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions