Repository navigation
feat: suggest current MiniMax models through OpenRouter - #1458
Open
octo-patch wants to merge 2 commits into
Open
octo-patch wants to merge 2 commits into
octo-patch wants to merge 2 commits into
Conversation
octo-patch
force-pushed
the
octo/20261021-provider-add-reczz28Ox4432xHU
branch
from
October 20, 2026 21:06
4efab34 to
175e5d8
Compare
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Reason: Make the current MiniMax text models discoverable through the Dot's existing OpenRouter configuration and metering contract.
What changes
The Brain model field now suggests MiniMax M3.1 Flash (
minimax/minimax-m3.1-flash:free) before MiniMax M3 (minimax/minimax-m3). The existing default is unchanged. Saved IDs, custom IDs, older MiniMax IDs, summary models and YAML round trips remain supported.These selections use the stored OpenRouter key. Model limits and request costs continue to come from OpenRouter; no direct API limits, prices or reasoning defaults are imposed on gateway requests. The provider, config projection and secret contract remain the upstream implementation.
OpenRouter route evidence: Flash endpoints and M3 endpoints.
Related issues
None.
How it was tested
apps/web/test/dot-form.test.ts,apps/web/test/yaml.test.ts, andpackages/shared/test/config.test.ts.tests/core/providers,tests/dots/test_provider.py,tests/dots/test_protocol.py,tests/dots/test_projection.py, andtests/dots/test_spend.py.git diff --checkpassed.Validation used existing read-only caches, with isolated credential-free environments: Node 24, TypeScript 7.0.2, Vitest 5.0.4, Python 3.11 and OpenAI SDK 3.35.1. This is not the exact CI environment (Python 3.12 and the image lock's SDK 3.24.0). Full workspace typecheck, complete pre-push suites, build, VM/browser smokes and paid/live inference were not run: the shared host has about 10 GiB free and heavy setup is paused. Focused checks are not a replacement for those gates.
Checklist
git config core.hooksPath .githooks).docs/updated if what a person sees changed.