Remove the meta-llama integration

[julian-risch]

Remove the meta-llama integration

We are discontinuing the meta-llama integration (meta-llama-haystack).

Reason: Meta is winding down the Llama API (public preview) on July 6, 2026. After that date the service shuts down and API requests return a sunset response with redirect guidance. Llama models remain available; only the hosted Llama API is being retired.

Users can keep using Meta Llama models via other integrations.

Removal checklist

haystack-core-integrations repo

• Delete the integration code directory integrations/meta_llama/
• Delete the CI workflow .github/workflows/meta_llama.yml
• Remove the Test / meta_llama entry from .github/workflows/CI_coverage_comment.yml
• Remove the integration:meta-llama block from .github/labeler.yml
• Mark the meta-llama-haystack integration as archived in the Inventory section of the repo README.md

haystack repo

• Add a note to the docs from https://docs.haystack.deepset.ai/

haystack-integrations repo

• Add a note to the integration tile from https://github.com/deepset-ai/haystack-integrations

We can then also mark the pypi package as deprecated.

[A-Amini]

Hi @julian-risch ! I’d like to work on this issue. Could you please assign it to me?

[Armaansaini20]

hi @julian-risch ! if the issue is still open i would love to work on it.

[mittalpk]

Hi, checking in on this — looks like the checklist items scoped to this repo are already done via #3552 (merged 2026-07-07: integration directory deleted, CI workflow removed, labeler entry removed, README archived). The two remaining unchecked items (docs note in haystack, integration tile note in haystack-integrations) live in different repos, not this one. Should this issue be closed here, or is there something still outstanding in haystack-core-integrations I'm missing?

[tonydzi]

The haystack-integrations checklist item is now covered: deepset-ai/haystack-integrations#566 adds a 🚧 discontinuation note to the meta_llama.md tile (house style of the google-ai.md deprecation notice, with live alternatives for running Llama models). That leaves only the docs.haystack.deepset.ai note, which looks like it needs deepset-side access.