provider: add SCX.ai - #1760
Open
bhuvan2134686 wants to merge 3 commits into
Open
Conversation
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
@bhuvan2134686 take a moment to raise this PR against maintained fork https://github.com/glama-ai/lightport |
bhuvan2134686
force-pushed
the
provider/scx-ai
branch
from
August 6, 2026 02:03
6b513e7 to
a206343
Compare
Add scx-ai, an Australian AI platform serving models over an OpenAI-compatible API at https://api.scx.ai/v1 with bearer auth. The provider reuses chatCompleteParams and responseTransformers from open-ai-base. Streaming emits standard SSE with a trailing usage-only chunk and a \n\n delimiter, so neither a custom stream transform nor a split-pattern entry in utils.ts is needed. customHost is honoured for SCX's self-hosted container deployments. Excluded params are the ones SCX documents as ignored and omits from each model's supported_sampling_parameters in /v1/models: logprobs, top_logprobs, n and logit_bias. seed, presence_penalty and frequency_penalty are kept — SCX lists them as supported and seed was verified to produce deterministic output. Verified live against api.scx.ai through the local gateway: chat, tool calls and streaming with usage.
bhuvan2134686
force-pushed
the
provider/scx-ai
branch
from
August 6, 2026 02:08
a206343 to
9065d38
Compare
Add the six models SCX now serves publicly alongside GLM-5.2 and qwen3.8-max: GLM-5.2-Fast, MiniMax-M2.7, gpt-oss-120b, gemma-4-31B-it, Qwen3-32B and Llama-4-Maverick-17B-128E-Instruct. All six verified live against api.scx.ai (chat completion round-trip per model, 2026-08-27).
Author
|
Updated: extended the catalogue entries in |
Drop the five secondary models — SCX is limiting this listing to its three headline models. GLM-5.2-Fast stays, verified live with a chat-completion round-trip through api.scx.ai.
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
2 participants
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.
Adds SCX.ai as a provider. SCX.ai is an Australian AI platform serving models over an OpenAI-compatible API at
https://api.scx.ai/v1.Companion pricing PR: Portkey-AI/models#874
Implementation
The API is OpenAI-compatible with bearer auth, so the provider reuses
chatCompleteParamsandresponseTransformersfromopen-ai-basewith no custom transforms:src/providers/scx-ai/api.ts— base URL, auth header,/chat/completionsendpoint.customHostis honoured for SCX's self-hosted container deployments.src/providers/scx-ai/index.ts— params + response transformssrc/globals.ts—SCX_AIconstant, added toVALID_PROVIDERSsrc/providers/index.ts— registered in theProvidersmapsrc/data/{providers,models}.json— catalogue metadataNo custom stream transform is needed: SCX emits standard SSE with a trailing usage-only chunk and a
\n\ndelimiter, so the default split pattern insrc/utils.tsapplies unchanged (verified against the raw byte stream).Excluded params
logprobs,top_logprobs,nandlogit_bias— the params SCX documents as ignored and omits from each model'ssupported_sampling_parametersin/v1/models.seed,presence_penaltyandfrequency_penaltyare deliberately kept: SCX lists them as supported, andseedwas verified to produce deterministic output (identical seeds returned byte-identical completions, different seeds diverged).Verification
Ran the gateway locally (
npm run dev:node) and routed live traffic through it, confirming:provider: "scx-ai"tool_choiceauto/none/required/ named, andparallel_tool_calls)[DONE]and a usage chunknpm run format:checkpasses.npm run test:gatewayis unchanged frommain(8 suites fail to compile on a clean tree from pre-existing type errors insrc/providers/bytez/api.tsandtests/unit/.../preRequestValidatorService.test.ts; 47 tests pass with and without this change).