Skip to content

docs: clarify mock-epp/llmd-ext-proc image scope in operations.md - #51

Open
jordigilh wants to merge 1 commit into
praxis-proxy:mainfrom
jordigilh:docs/operations-md-mock-epp-scope
Open

docs: clarify mock-epp/llmd-ext-proc image scope in operations.md#51
jordigilh wants to merge 1 commit into
praxis-proxy:mainfrom
jordigilh:docs/operations-md-mock-epp-scope

Conversation

@jordigilh

Copy link
Copy Markdown
Contributor

Context

While root-causing ai#716 (the SWIM-partition failover fixture's missing images), I found that this repo's own docs overstate the requirement and would send the next contributor down the same investigation.

What's wrong

The "Required local images" table in docs/architecture/operations.md reads as if all provider/consumer gateways need locally-built praxis-ai:llmd-ext-proc and praxis-ai-mock-epp images. In practice:

  • Only the generic xtask env harness (validate-all, verify-swim-mesh-three-node, verify-failover-under-lost-peer) uses those defaults.
  • The named demos (grid-glb-demo, grid-combined-site, grid-llmd-pool-metrics) override them via GRID_XTASK_GATEWAY_IMAGE/GRID_XTASK_MOCK_EPP_IMAGE (see xtask/src/env/image_overrides.rs) with published ghcr.io/praxis-proxy/grid-ai-rollup images and never touch an AI repository checkout.
  • Neither Containerfile.composed nor a mock llm-d EPP server exists in the AI repository today, so build-gateway-images --ai-repo can't currently produce either image — this only blocks the generic-harness path (verify-failover-under-lost-peer), not the three named demos.

Change

Added a clarifying note after the table, cross-linking ai#716 (tracking the gap) and ai#334 (the in-flight, other-team-owned ext_proc work it's soft-blocked on, deliberately paused per an Aug 10 AI Gateway release-timeline decision rather than abandoned).

Test plan

Docs-only change, no code touched.

The "Required local images" table read as if all provider/consumer
gateways need locally-built `praxis-ai:llmd-ext-proc` and
`praxis-ai-mock-epp` images. In practice only the generic `xtask env`
harness (validate-all, verify-swim-mesh-three-node,
verify-failover-under-lost-peer) uses those defaults; the named demos
(grid-glb-demo, grid-combined-site, grid-llmd-pool-metrics) override
them with published ghcr.io/praxis-proxy/grid-ai-rollup images via
GRID_XTASK_GATEWAY_IMAGE/GRID_XTASK_MOCK_EPP_IMAGE and never touch an
AI repository checkout.

Also note that Containerfile.composed and a mock llm-d EPP server
don't exist in the AI repository yet, so build-gateway-images
--ai-repo can't produce either image today. That gap only blocks the
generic-harness path (verify-failover-under-lost-peer); it doesn't
affect the three named demos above. Cross-links ai#716 (tracking the
gap) and ai#334 (the in-flight, other-team-owned ext_proc work it's
soft-blocked on).

Signed-off-by: Jordi Gil <jgil@redhat.com>
@nerdalert

Copy link
Copy Markdown
Member

@jordigilh thanks! 🥇 Just needs a rebase.

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