Skip to content

chore: add info for working with intrinsics to AGENTS.md#768

Open
psschwei wants to merge 2 commits intogenerative-computing:mainfrom
psschwei:intrinsics-agents-md
Open

chore: add info for working with intrinsics to AGENTS.md#768
psschwei wants to merge 2 commits intogenerative-computing:mainfrom
psschwei:intrinsics-agents-md

Conversation

@psschwei
Copy link
Copy Markdown
Member

@psschwei psschwei commented Mar 31, 2026

Misc PR

Type of PR

  • Bug Fix
  • New Feature
  • Documentation
  • Other

Description

  • Link to Issue: Fixes

Adds information for agents when working with intrinsics in this repo

Testing

  • Tests added to the respective file if code was changed
  • New code has 100% coverage if code as added
  • Ensure existing tests and github automation passes (a maintainer will kick off the github automation when the rest of the PR is populated)

Signed-off-by: Paul S. Schweigert <paul@paulschweigert.com>
@psschwei psschwei requested a review from a team as a code owner March 31, 2026 21:41
@github-actions
Copy link
Copy Markdown
Contributor

The PR description has been updated. Please fill out the template for your PR to be reviewed.

@psschwei
Copy link
Copy Markdown
Member Author

cc @generative-computing/mellea-intrinsics

Copy link
Copy Markdown
Contributor

@jakelorocco jakelorocco left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is a good thing to add. I have a few vibe-based thoughts. Feel free to disagree with them.

AGENTS.md Outdated
Comment on lines +150 to +154
## 13. Working with Intrinsics

Intrinsics are specialized LoRA adapters hosted on Hugging Face that add task-specific capabilities (RAG evaluation, safety checks, calibration, etc.) to Granite models.

**⚠️ Before writing code that uses an intrinsic, fetch its README from Hugging Face.** Each README contains the authoritative specification for input format, output format, intended use, and examples. Do not guess these details.
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I worry that this piece of things might be too generic to regular intrinsics and not utilizing them in Mellea. The following is just based off vibes:

  1. I would maybe preface with the how to use in Mellea stuff (the project resources below but maybe with more details here).
  2. I would phrase this current working with intrinsics section as something like "how to add new intrinsics".

Since we utilize the "granite-common" formatters for intrinsics, the agent shouldn't actually need to know how to modify the context / messages, the library should do that for us. It just needs to know whether to call the high-level / convenience wrappers or the regular usage.

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Those are fair points. Referring to the model cards was something the intrinsics team suggested, but I believe that was mostly around intended usage and examples (in a previous PR we had used them in a counterproductive way). Let me revise a bit

Signed-off-by: Paul S. Schweigert <paul@paulschweigert.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants