Repository navigation
Conversation
NeonModel (neon/<neon-short-id>) in the TypeScript and Python SDKs for the Codex and OpenCode harnesses, harness inference for neon/ models, and Neon groups in the CLI picker. GPT models go to Codex; GPT OSS, Claude and the other chat-only models go to OpenCode. Box does not run Claude Code with Neon. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Copilot review overview
🟡 Changes recommended
Public documentation conflicts with the implemented Grok and GPT OSS routing behavior.
Review effort: Balanced
Findings: 4
Open (4)
What changed in this PR
Adds Neon AI Gateway model support across the TypeScript SDK, Python SDK, and CLI.
Changes:
- Adds and exports 36 Neon model identifiers.
- Infers Codex or OpenCode based on Neon endpoint support.
- Adds CLI model groups, tests, parity documentation, and release notes.
| File | Description |
|---|---|
packages/sdk/src/types.ts |
Defines Neon models and harness compatibility. |
packages/sdk/src/index.ts |
Exports NeonModel. |
packages/sdk/src/client.ts |
Adds Neon provider inference. |
packages/sdk/src/__tests__/types.test.ts |
Tests Neon identifiers. |
packages/sdk/src/__tests__/infer-runner.test.ts |
Tests Neon inference. |
packages/python-sdk/upstash_box/types.py |
Mirrors Neon model definitions. |
packages/python-sdk/upstash_box/_common.py |
Mirrors provider inference. |
packages/python-sdk/upstash_box/__init__.py |
Exports NeonModel. |
packages/python-sdk/tests/_async/test_models.py |
Tests Python identifiers. |
packages/python-sdk/tests/_async/test_infer_provider.py |
Tests Python inference. |
packages/python-sdk/PARITY.md |
Records SDK parity. |
packages/python-sdk/CHANGELOG.md |
Documents Python changes. |
packages/cli/src/models.ts |
Adds Neon picker groups. |
packages/cli/src/__tests__/models.test.ts |
Tests picker availability. |
.changeset/neon-ai-gateway-models.md |
Adds package release notes. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
GPT OSS is the exception to neon/gpt-* inferring Codex, and Grok 4.6 runs on both harnesses, so it is not OpenCode-only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…om endpoint support - The create wizard no longer offers Upstash-managed or a raw key for neon/ models, which the backend rejects; it sets the stored credential. - Inference comments (TS and Python) and the test name describe default routing separately from endpoint support: Grok 4.6 defaults to OpenCode but Neon also serves it on the Responses API. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CahidArda
approved these changes
Oct 6, 2026
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
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.
Summary
Follows upstash/box-backend#277 (Neon AI Gateway provider) and #285 (Neon harness rules), both merged and live on prod. This adds the Neon models to the SDKs and CLI.
NeonModelhas 36 constants for Neon's short ids underneon/, for exampleNeonModel.GPT_5_5forneon/gpt-5-5andNeonModel.Claude_Haiku_4_5forneon/claude-haiku-4-5. The Python SDK has the same list (GPT_5_5,CLAUDE_HAIKU_4_5, …). The Codex and OpenCode harness types accept them.inferDefaultProviderand Python'sinfer_default_providermapneon/gpt-*to Codex, exceptgpt-oss-*. Every other Neon model maps to OpenCode, including GPT OSS, Claude, Gemini, Llama, Qwen, Kimi, GLM and Grok.gpt-5-5-pro,gpt-5-3-codex) are Codex only. Chat-only models, including all Claude models, are OpenCode only, because Neon doesn't serve Claude on its Responses API. Claude Code gets no Neon group.BoxApiKey.StoredKey, and the backend refuses raw keys and the managed key for Neon.Tests
neon/gpt-5-nanoas a plain run and a streamed run.neon/gpt-oss-20b,neon/gemini-3-5-flash-liteandneon/claude-haiku-4-5, plain and streamed.All test boxes were deleted.
🤖 Generated with Claude Code