Skip to content

fix(provider): Alibaba Token Plan provider - #3787

Merged
amitksingh1490 merged 3 commits into
mainfrom
feat/alibaba-token-plan
Jul 30, 2026
Merged

fix(provider): Alibaba Token Plan provider#3787
amitksingh1490 merged 3 commits into
mainfrom
feat/alibaba-token-plan

Conversation

@amitksingh1490

@amitksingh1490 amitksingh1490 commented Jul 30, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Summary

Alibaba Token Plan

Adds support for the Alibaba Token Plan provider (models.dev/providers/alibaba-token-plan), mirroring how opencode supports it.

  • OpenAI-compatible endpoint: https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1
  • Auth via ALIBABA_TOKEN_PLAN_API_KEY (same env var as opencode)
  • New ProviderId::ALIBABA_TOKEN_PLAN built-in constant with FromStr roundtrip support
  • Models hardcoded in provider.json (endpoint does not expose capability metadata), covering the text/chat models: Qwen3.8 Max Preview, Qwen3.7 Max/Plus, Qwen3.6 Plus/Flash, DeepSeek V4 Pro/Flash, DeepSeek V3.2, Kimi K2.5/2.6/2.7 Code, GLM-5/5.1/5.2, MiniMax-M2.5 (image/video generation models excluded)

Neuralwatt missing models

Fixes missing Neuralwatt models reported by a user ("what happened to flex models from neuralwatt? they have em on opencode"):

  • Flex tier: glm-5.2-flex, glm-5.2-short-flex, glm-5.2-short-fast-flex, kimi-k2.6-flex, kimi-k2.7-code-flex
  • Also adds kimi-k2.5 and kimi-k2.5-fast, which were missing entirely

Test plan

  • Added test_alibaba_token_plan_config in provider_repo.rs
  • Extended test_neuralwatt_config to assert flex models are present
  • Added display-name, built-in-list, and FromStr roundtrip tests in forge_domain/src/provider.rs
  • cargo test -p forge_domain -p forge_repo provider tests pass; provider.json validated as JSON

Co-Authored-By: ForgeCode noreply@forgecode.dev
@amitksingh1490 @forge-code-agent
feat(provider): add Alibaba Token Plan provider with hardcoded models
a865e5a 
Adds support for Alibaba Cloud Model Studio Token Plan
(https://models.dev/providers/alibaba-token-plan/), an OpenAI-compatible
endpoint at token-plan.ap-southeast-1.maas.aliyuncs.com. Models are
hardcoded in provider.json (mirroring opencode's alibaba-token-plan
provider) since the endpoint does not expose capability metadata.

Co-Authored-By: ForgeCode <noreply@forgecode.dev>
@github-actions github-actions Bot added type: feature Brand new functionality, features, pages, workflows, endpoints, etc. type: provider Updates provider.json configuration. labels Jul 30, 2026
@amitksingh1490 amitksingh1490 changed the title feat(provider): add Alibaba Token Plan provider Jul 30, 2026
@amitksingh1490 amitksingh1490 removed the type: feature Brand new functionality, features, pages, workflows, endpoints, etc. label Jul 30, 2026
@amitksingh1490 @forge-code-agent
feat(provider): add missing Neuralwatt flex and Kimi K2.5 models
a48bc3c 
Adds the flex-tier variants (glm-5.2-flex, glm-5.2-short-flex,
glm-5.2-short-fast-flex, kimi-k2.6-flex, kimi-k2.7-code-flex) and the
missing Kimi K2.5 / K2.5 Fast models to the Neuralwatt provider,
matching the model list available on opencode.

Co-Authored-By: ForgeCode <noreply@forgecode.dev>
@amitksingh1490 amitksingh1490 changed the title fux(provider): Alibaba Token Plan provider Jul 30, 2026
@amitksingh1490 amitksingh1490 changed the title feat(provider): add Alibaba Token Plan provider and missing Neuralwatt models Jul 30, 2026
@amitksingh1490 amitksingh1490 added the type: fix Iterations on existing features or infrastructure. label Jul 30, 2026
@hattapauzi

hattapauzi commented Jul 30, 2026 •
edited
Loading

Copy link
Copy Markdown

Summary

Alibaba Token Plan

Adds support for the Alibaba Token Plan provider (models.dev/providers/alibaba-token-plan), mirroring how opencode supports it.

  • OpenAI-compatible endpoint: https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1
  • Auth via ALIBABA_TOKEN_PLAN_API_KEY (same env var as opencode)
  • New ProviderId::ALIBABA_TOKEN_PLAN built-in constant with FromStr roundtrip support
  • Models hardcoded in provider.json (endpoint does not expose capability metadata), covering the text/chat models: Qwen3.8 Max Preview, Qwen3.7 Max/Plus, Qwen3.6 Plus/Flash, DeepSeek V4 Pro/Flash, DeepSeek V3.2, Kimi K2.5/2.6/2.7 Code, GLM-5/5.1/5.2, MiniMax-M2.5 (image/video generation models excluded)

Neuralwatt missing models

Fixes missing Neuralwatt models reported by a user ("what happened to flex models from neuralwatt? they have em on opencode"):

  • Flex tier: glm-5.2-flex, glm-5.2-short-flex, glm-5.2-short-fast-flex, kimi-k2.6-flex, kimi-k2.7-code-flex
  • Also adds kimi-k2.5 and kimi-k2.5-fast, which were missing entirely

Test plan

  • Added test_alibaba_token_plan_config in provider_repo.rs
  • Extended test_neuralwatt_config to assert flex models are present
  • Added display-name, built-in-list, and FromStr roundtrip tests in forge_domain/src/provider.rs
  • cargo test -p forge_domain -p forge_repo provider tests pass; provider.json validated as JSON

Co-Authored-By: ForgeCode noreply@forgecode.dev

I have tested this branch with my own API key, and it is working, I only tested it with Qwen 3.8 Preview model though...
Here is the dump file for a reference!

2026-07-30_11-23-36-dump.json
@amitksingh1490
amitksingh1490 merged commit c2cd6fd into main Jul 30, 2026
8 checks passed
@amitksingh1490
amitksingh1490 deleted the feat/alibaba-token-plan branch July 30, 2026 04:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

type: fix Iterations on existing features or infrastructure. type: provider Updates provider.json configuration.

3 participants

@amitksingh1490 @hattapauzi @ssddOnTop