Skip to content

feat(models): add Nebius model mappings - #3187

Merged
smakosh merged 3 commits into
theopenco:mainfrom
RATCHAW:RATCHAW/add-nebius-models
Jul 22, 2026
Merged

smakosh merged 3 commits into
theopenco:mainfrom
RATCHAW:RATCHAW/add-nebius-models

Conversation

@RATCHAW

@RATCHAW RATCHAW commented Jul 22, 2026 •

Copy link
Copy Markdown
Contributor

Summary

  • add Nebius mappings for current DeepSeek, MiniMax, Kimi, Hermes, Nemotron, Cosmos, GLM, Gemma, and Qwen models
  • add OpenBMB catalogue support with MiniCPM-V 4.5 and expose OpenBMB in the open-source family filter
  • align existing Nebius pricing, context limits, capabilities, stability, and endpoint availability with live provider behavior
  • resolve embedding base URLs through the shared provider configuration so Nebius-hosted embedding models use the correct endpoint

Why

The model catalogue did not reflect Nebius's current hosted lineup or several behaviors verified against its live endpoints. Embeddings also used a local base-URL allowlist, which omitted provider defaults already maintained by the shared actions package.

Impact

Nebius models can be selected and routed with accurate capability metadata, Qwen3 Embedding 8B can use the Nebius endpoint, and removed or unreliable deployments are kept out of normal routing.

Validation

  • pnpm format
  • pnpm build (17/17 tasks passed)
  • pnpm test:unit (2,865 passed, 2 skipped; one unrelated timezone-sensitive calendar-month assertion failed under Africa/Casablanca)
  • TZ=UTC pnpm exec vitest run packages/db/src/api-key-period-limit.spec.ts --no-file-parallelism (5/5 passed, confirming the unrelated failure is timezone-dependent)

Summary by CodeRabbit

  • New Features
    • Added support for additional Nebius-hosted models, including OpenBMB, NVIDIA, Nous Research, Moonshot, Z.ai, DeepSeek, and Alibaba.
    • Added the Qwen3 Embedding 8B model for embedding requests.
    • Added OpenBMB models to the Open Source category filter.
  • Updates
    • Expanded embedding model compatibility and improved embedding endpoint/provider selection.
    • Updated pricing, context limits, and capabilities for multiple existing models; marked some deployments as no longer available.

@coderabbitai

coderabbitai Bot commented Jul 22, 2026 •

Copy link
Copy Markdown
Contributor

Review Change Stack

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 6b163a0d-7153-4a21-8ed2-5c0bfc8469cf

📥 Commits

Reviewing files that changed from the base of the PR and between 509ed94 and bbb0025.

📒 Files selected for processing (4)
  • apps/gateway/src/embeddings/embeddings.ts
  • packages/models/src/models/moonshot.ts
  • packages/models/src/models/nvidia.ts
  • packages/models/src/models/zai.ts
🚧 Files skipped from review as they are similar to previous changes (4)
  • packages/models/src/models/zai.ts
  • packages/models/src/models/moonshot.ts
  • packages/models/src/models/nvidia.ts
  • apps/gateway/src/embeddings/embeddings.ts

Walkthrough

The PR updates embedding upstream URL resolution and expands the model catalog with OpenBMB registration plus new or revised Nebius provider configurations across multiple model families.

Changes

Embedding routing

Layer / File(s) Summary
Provider base-URL resolution
apps/gateway/src/embeddings/embeddings.ts
The embeddings route uses getProviderDefaultBaseUrl(providerId) instead of an inline provider URL map and broadens the dimensions schema description.

Model catalog

Layer / File(s) Summary
OpenBMB model registration
packages/models/src/models/openbmb.ts, packages/models/src/models.ts, packages/shared/src/components/models-directory/model-category-filters.ts
Adds the minicpm-v-4.5 model, includes it in the consolidated registry, and classifies the openbmb family as open source.
Nebius model definitions
packages/models/src/models/alibaba.ts, packages/models/src/models/deepseek.ts, packages/models/src/models/google.ts, packages/models/src/models/minimax.ts, packages/models/src/models/moonshot.ts, packages/models/src/models/nousresearch.ts, packages/models/src/models/nvidia.ts, packages/models/src/models/zai.ts
Adds Nebius-backed models and provider entries, deactivates unavailable endpoints, and updates pricing, context, quantization, and capability metadata.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related PRs

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Description Check ✅ Passed Check skipped - CodeRabbit’s high-level summary is enabled.
Title check ✅ Passed The title clearly summarizes the main change: adding Nebius model mappings across the model catalog.
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@smakosh
smakosh added this pull request to the merge queue Jul 22, 2026
Merged via the queue into theopenco:main with commit a8157bd Jul 22, 2026
11 checks passed
pull Bot pushed a commit to soitun/llmgateway that referenced this pull request Jul 23, 2026
## Summary

Three related marketing-surface updates:

**SCX.ai "Up to 4x faster" badge on model cards**
(https://llmgateway.io/providers/scx-ai)
- New optional `modelCardBadge` field on `ProviderDefinition`
(data-driven — any provider can carry one), set to `"Up to 4x faster"`
on `scx-ai`.
- Rendered in the shared model card's provider header as an amber Zap
badge, next to the provider name. Plumbed through the shared
`ApiProvider` type and the provider detail page's catalogue conversion.

**Theme-aware SCX logo**
- The SCX wordmark SVG hardcoded `fill="#262626"`, making it
near-invisible in dark mode. All four fills now use `currentColor`, so
it renders black in light mode and white in dark mode (matching the
pattern used by the other monochrome provider icons).

**Changelog: July roundup**
(`/changelog/upgrade-rollover-new-providers`)
- New entry (id 67) covering everything user-facing since the Jul 16
Reset Passes entry: DevPass upgrade rollover + upgrade-timing choice
(theopenco#3147), SCX.ai (theopenco#2958) and Gonka24 (theopenco#3189) providers, Nebius mappings
(theopenco#3187), Gemini 3.6 Flash / 3.5 Flash Lite (theopenco#3164), Gemini TTS on two
providers (theopenco#3143), the Empryo coding agent (theopenco#3161), and the new
request-timeouts docs page (theopenco#3190).
- OG image generated with gpt-image-2 in the house circuit-board style.

## Testing

- Full `pnpm build` passes (validates the changelog frontmatter against
the content-collections schema and typechecks the badge plumbing).
- Verified live on the dev server: badge renders on every SCX model card
in both themes, and the logo is black-on-light / white-on-dark
(screenshots shared in session).

https://claude.ai/code/session_01UGktbNs7a35FU1k2ooKfaB

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **New Features**
* Added provider badges to model cards, including an “Up to 4x faster”
badge for SCX.ai.
* Added new inference providers and expanded model availability,
including Gemini models and Gemini TTS.
* Upgrade credits can roll over, with immediate or next-renewal upgrade
options.
* Added Empryo coding-agent usage attribution and request-timeout
documentation.
* **Style**
  * Improved provider icon coloring for better theme compatibility.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: Luca Steeb <contact@luca-steeb.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