Feat/peer metadata display#21
Merged
Merged
Conversation
Backend (model_service): pass through hostname, version, status, labels,
and convenience pulls (launched_by, slurm_job_id, worker_group_id,
framework, started_at) for each DNT peer. Also surface metrics-only
follower peers (no service, but with worker_group_id) so multi-node
replicas can be reconstructed in aggregation.
Frontend (ModelList): group raw peers by worker_group_id to count
replicas distinctly from peers/nodes. Headline now reads
"Available Models X, Replicas Y" when the two diverge.
Frontend (ModelCard): clicking the card now expands inline instead of
opening OpenWebUI. The expansion shows:
- Open in OpenWebUI button (the prior click behaviour)
- Per-replica monospace block with model, launched_by, slurm_job_id,
started_at, framework, version, head + follower hostnames
- Topology header (e.g. "2 nodes × 4x GH200")
- Per-replica extra-labels block for anything else OCF carries
Fixtures: snapshot of live prod /dnt/table + a script that synthesises
the post-v0.0.6 shape (hostname/version/status/labels) by adding a
multi-node replica demo (shared worker_group_id, one head + one
metrics-only follower) so the new code paths have a realistic test.
Settings: ocf_head_addr → otela_head_addr, ocf_fixture_path → otela_fixture_path. DEPLOY NOTE: this changes the env var names from OCF_HEAD_ADDR / OCF_FIXTURE_PATH to OTELA_HEAD_ADDR / OTELA_FIXTURE_PATH. Ops side must update before/with this deploy or /v1/models* will hit an empty endpoint. Comments, README, k8s manifests, and the in-repo guides now refer to "OpenTela" rather than "OCF". External image tag (ghcr.io/researchcomputer/ocf:*) and the in-binary mount path (/root/.ocfcore/keys) are untouched — both are dictated by upstream and would need a coordinated rename there first. Fixture mode: when OTELA_FIXTURE_PATH is set, /v1/models* reads that JSON file instead of HTTP-getting OTELA_HEAD_ADDR/v1/dnt/table. Used for iterating on the new model-card expansion UI against the synthesised post-v0.0.6 payload before the binary actually ships.
Use pydantic AliasChoices so OCF_HEAD_ADDR / OCF_FIXTURE_PATH still populate the renamed settings. Deployments can migrate on their own schedule without a synchronized cut-over. When both legacy and canonical names are set, the canonical OTELA_* wins — a partial migration shouldn't silently keep the legacy value in force.
…-ai/serving-api into feat/peer-metadata-display
Upgraded DNT fixture now includes framework_args (the second monospace block in the card expansion), expires_at (SLURM time limit applied to started_at), slurm_reservation (mixed in for some launches), and varied started_at values spread across several hours so the UI shows a realistic mix of ages. Note: framework_args isn't in the opentela --label set we shipped yet — this fixture preempts a planned follow-up patch there. Until that ships, real prod data won't carry framework_args; serving-api just hides the row. ModelCard expansion: filter out empty rows so the legacy / pre-v0.0.6 case shows just what's known (peer ids + model) instead of a wall of "?" placeholders. When no labels exist at all, render a small amber hint pointing at the v0.0.6 requirement instead of silently rendering an empty card.
Same as `make run` but forces OTELA_FIXTURE_PATH at the synthesised upgraded fixture, so the model card UI shows the v0.0.6-shape payload (hostname, version, labels, multi-node demo) without depending on live prod state or whatever's in the developer's .env. Use `make run` to hit live prod, `make dummy-run` to iterate on the UI.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.