feat: add openai-compatible endpoint support for google-vertex provider by leehack · Pull Request #10303 · anomalyco/opencode

leehack · 2026-01-23T23:22:37Z

Summary

Adds a new google-vertex-openapi provider that enables using third-party models (like GLM-4.7) on Google Vertex AI through the OpenAPI-compatible endpoint.

Fixes #10304

Changes

Provider Implementation (src/provider/provider.ts):

Replaced gcloud CLI subprocess approach with Application Default Credentials (ADC) using google-auth-library
- No subprocess overhead (performance improvement)
- Automatic token refresh (handled by library)
- Supports multiple credential sources (gcloud ADC, service account, etc.)
Added getModel function for consistency with other providers

Test Infrastructure (test/provider/provider.test.ts):

Mocked google-auth-library for testing support
Configuration:
- Environment variables: GOOGLE_CLOUD_PROJECT, GCP_PROJECT, GCLOUD_PROJECT (project)
- Environment variables: GOOGLE_CLOUD_LOCATION, VERTEX_LOCATION (location)
- Config options: provider.options.project, provider.options.location
Authentication:
- Works with gcloud auth application-default login (easiest for development)
- Supports GOOGLE_APPLICATION_CREDENTIALS for service accounts (production)

Verification

Typecheck: Passed
Tests: All passing (pre-existing failures unaffected)
Manual testing: Can configure provider and make requests to Vertex AI OpenAPI endpoint

Usage

{
  "provider": {
    "google-vertex-openapi": {
      "name": "Google Vertex AI (OpenAPI)",
      "npm": "@ai-sdk/openai-compatible",
      "api": "https://aiplatform.googleapis.com/v1beta1",
      "options": {
        "project": "my-project-id",
        "location": "global"
      },
      "models": {
        "glm-4.7": {
          "name": "GLM-4.7",
          "tool_call": true,
          "reasoning": true,
          "limit": {"context": 128000, "output": 8192}
        }
      }
    }
  }
}

Setup

# Using gcloud ADC (recommended for development)
gcloud auth application-default login
export GOOGLE_CLOUD_PROJECT=my-project-id

# OR using service account (recommended for production)
export GOOGLE_APPLICATION_CREDENTIALS=~/.config/google/vertex-ai-service-account.json
export GOOGLE_CLOUD_PROJECT=my-project-id

github-actions · 2026-01-23T23:22:50Z

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

Open an issue describing the bug/feature (if one doesn't exist)
Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

github-actions · 2026-01-23T23:23:15Z

The following comment was made by an LLM, it may be inaccurate:

Potential Duplicate Found:

PR feat: add Google Vertex AI service account JSON authentication #6287: "feat: add Google Vertex AI service account JSON authentication"
- feat: add Google Vertex AI service account JSON authentication #6287
- This PR is also related to Google Vertex AI authentication. It may address similar functionality for service account authentication, which overlaps with the ADC authentication approach in PR feat: add openai-compatible endpoint support for google-vertex provider #10303.

The other Google Vertex-related PRs (#9568, #6860, #5422, #5166) appear to be focused on different aspects (model resolution, token stats, cache configuration, naming conventions) rather than authentication implementation.

…port

NimaAra · 2026-01-29T22:36:13Z

Can we get this merged please?

bluet · 2026-01-30T08:25:36Z

Hey @leehack, great work on this! I was working on a similar solution (now closed #10742) and your approach of extending the existing google-vertex provider with protocol: "openapi" is cleaner than creating a separate provider.

I've also been coordinating with @rekram1-node on models.dev PR #716 for the model definitions. He mentioned consolidating all google-vertex models under a single provider, which aligns perfectly with your implementation here.

Happy to help test this PR if needed - I have a GCP project with GLM-4.7 access via Vertex AI Model Garden.

One small observation: The baseURL uses v1beta1 - just confirming this works with the OpenAPI-compatible endpoint? In my testing, I was using v1 for partner models. Let me know if you'd like me to verify.

leehack · 2026-01-30T10:25:11Z

Hey @leehack, great work on this! I was working on a similar solution (now closed #10742) and your approach of extending the existing google-vertex provider with protocol: "openapi" is cleaner than creating a separate provider.

I've also been coordinating with @rekram1-node on models.dev PR #716 for the model definitions. He mentioned consolidating all google-vertex models under a single provider, which aligns perfectly with your implementation here.

Happy to help test this PR if needed - I have a GCP project with GLM-4.7 access via Vertex AI Model Garden.

One small observation: The baseURL uses v1beta1 - just confirming this works with the OpenAPI-compatible endpoint? In my testing, I was using v1 for partner models. Let me know if you'd like me to verify.

Awesome!
Yes, I'm testing with the glm 4.7 model with the Vertex AI model Garden. I've only tested glm 4.7 with the v1beta1 endpoint. Feel free to test and let me know if you think we should better use the v1 endpoint.

bluet · 2026-02-02T18:19:49Z

Test Results: v1 vs v1beta1 Endpoints

I tested both API versions with GLM-4.7. Both work!

Results

Endpoint	Status
`/v1/.../openapi/chat/completions`	✅ Works
`/v1beta1/.../openapi/chat/completions`	✅ Works

Conclusion

Your current implementation using v1beta1 is fine - both versions work identically for the OpenAPI-compatible endpoint. No changes needed.

The correct model ID format is <publisher>/<model-id>, e.g., zai-org/glm-4.7-maas.

Documentation Reference

GLM 4.7 Model Card
Model ID: glm-4.7-maas
Publisher: zai-org
Region: global only

bluet · 2026-02-02T18:26:22Z

Follow-up: Recommendation to use `/v1/` instead of `/v1beta1/`

After researching Google's API versioning policy, I'd suggest using /v1/ instead of /v1beta1/ for better production stability.

Why `/v1/` is preferable:

Aspect	`/v1/` (Stable)	`/v1beta1/` (Beta)
SLA	✅ Covered	❌ No guarantees
Breaking Changes	❌ Not allowed	⚠️ Allowed after deprecation
Deprecation Notice	Min 1 year	~180 days (not guaranteed)
Production Use	✅ Recommended	⚠️ Test environments

Reference: Google's Official Policy

From AIP-181 and Google Cloud Launch Stages:

Beta: No SLAs or technical support obligations unless specified
Stable (v1): Fully-supported, covered by SLA, no breaking changes

Suggested Change

In the loader's baseURL construction:

- `https://${location}-aiplatform.googleapis.com/v1beta1/projects/...`
+ `https://${location}-aiplatform.googleapis.com/v1/projects/...`

Both work identically for the OpenAPI-compatible endpoint (I tested both), but /v1/ provides stronger stability guarantees for production use.

bluet

according to google vertex docs, /v1beta1/ url path is for beta testing purpose. better to use /v1/ for stability.

packages/opencode/src/provider/provider.ts

Co-authored-by: BlueT - Matthew Lien - 練喆明 <BlueT@BlueT.org>

leehack · 2026-02-03T13:42:07Z

according to google vertex docs, /v1beta1/ url path is for beta testing purpose. better to use /v1/ for stability.

Thanks @bluet for the test and suggestion. I updated the url as you suggested.

…ni in production while allowing custom proxies and localhost testing

leehack · 2026-02-03T15:15:05Z

Fixed the test failures and ensured compatibility with proxy servers.

NimaAra · 2026-02-05T20:25:55Z

Is this ever going to be merged?

leehack · 2026-02-06T17:40:03Z

Is this ever going to be merged?

I'm also waiting; I'm currently running my own local build just for this. lol

ddesmond · 2026-02-10T06:57:43Z

same here +1!

KuzonCode · 2026-02-11T23:25:12Z

Would also appreciate this change!

rekram1-node · 2026-02-11T23:58:54Z

So this introduces a new "protocol" field which seems like it is used to set the npm package.

But we already have fields to set the npm package any model can have it's own npm package setting regardless of the provider default

klinki · 2026-02-12T09:23:16Z

I would also love this to get merged and new version with this released. It's blocking me from using Google Vertex API as well.

leehack · 2026-02-12T12:48:20Z

So this introduces a new "protocol" field which seems like it is used to set the npm package.

But we already have fields to set the npm package any model can have it's own npm package setting regardless of the provider default

Yep. we can use provider.npm instead. I removed protocol field.

Here is the example for a model:

"google-vertex": {
	"options": {
		"project": "gcp-project-id",
		"location": "global"
	},
	"models": {
		"zai-org/glm-4.7-maas": {
			"name": "GLM-4.7",
			"provider": { "npm": "@ai-sdk/openai-compatible", "api": "" },
			"limit": { "context": 204800, "output": 131072 },
			"cost": { "input": 0.6, "output": 2.2 },
			"interleaved": {
				"field": "reasoning_content"
			},
			"attachment": false,
			"reasoning": true,
			"family": "glm",
			"temperature": true,
			"tool_call": true,
			"modalities": {
				"input": [ "text" ],
				"output": [ "text" ]
			},
			"release_date": "2025-12-22",
		}
	}
},

leehack · 2026-02-12T14:19:32Z

@rekram1-node cleaned up the PR. removed all useless codes, mocks, tests and comments. rebased dev branch. Hope this is good now.

ddesmond · 2026-02-13T03:46:08Z

merge train hype cho cho cho! thumbs up and thanks for this work! :) 🚂

packages/opencode/src/provider/provider.ts

rekram1-node · 2026-02-16T08:31:32Z

This is also gonna be combined w/ changes in models.dev we will have things like this:

"google-vertex": {
      "options": {},
      "models": {
        "zai-org/glm-5-maas": {
          "provider": {
            "npm": "@ai-sdk/openai-compatible",
            "api": "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_CLOUD_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi",
          },
        },
      },
    },

Note that this is existing we just have to introduce this env var resolution pattern, and this ofc could be cleaned up a bit but the idea here is that models dont have to be tied to the provider base url, models can have their own package and own api = "base url"

* dev: (1474 commits) fixing merge issue fix: google vertex var priority (anomalyco#13816) chore: generate feat: add openai-compatible endpoint support for google-vertex provider (anomalyco#10303) fix(desktop): normalize Linux Wayland/X11 backend and decoration policy (anomalyco#13143) feat(opencode): Add Venice support in temperature, topP, topK and smallOption (anomalyco#13553) desktop: use process-wrap instead of manual job object (anomalyco#13431) chore: generate fix(website): correct zh-CN translation of proprietary terms in zen.mdx (anomalyco#13734) feat(opencode): add `cljfmt` formatter support for Clojure files (anomalyco#13426) release: v1.2.5 feat(app): localize "free usage exceeded" error & "Add credits" clickable link (anomalyco#13652) fix(desktop): issue viewing new files opened from the file tree (anomalyco#13689) fix(app): only navigate prompt history at input boundaries (anomalyco#13690) fix(app): keybind [shift+tab] (anomalyco#13695) docs: add Ukrainian README translation (anomalyco#13697) fix(desktop): focus window after update/relaunch (anomalyco#13701) feat: Add GeistMono Nerd Font to available mono font options (anomalyco#13720) fix: ensure sqlite migration logs to stderr instead of stdout (anomalyco#13691) release: v1.2.4 ...

NimaAra · 2026-02-16T21:22:21Z

I am trying to use this in v1.2.6 with the following config:

"google-vertex": {
      "models": {
        "zai-org/glm-4.7-maas": {
            "name": "GLM-4.7",
            "provider": { 
                "npm": "@ai-sdk/openai-compatible",
                "api": "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi"
            },
            "limit": { "context": 204800, "output": 131072 },
            "cost": { "input": 0.6, "output": 2.2 },
            "interleaved": {
                "field": "reasoning_content"
            },
            "attachment": false,
            "reasoning": true,
            "family": "glm",
            "temperature": true,
            "tool_call": true,
            "modalities": {
                "input": [ "text" ],
                "output": [ "text" ]
            },
            "release_date": "2025-12-22",
        },

But getting the error:

Error: invalid_scope: Invalid OAuth scope or ID token audience provided.

Gemini methods under same google-vertex provider work fine (via the GOOGLE_APPLICATION_CREDENTIALS pointing to the service.account.json)

I can manually try to do this which works:

"google-vertex-custom": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Google Vertex Custom",
      "options": {
        "baseURL": "https://${GOOGLE_VERTEX_ENDPOINT}/v1/projects/${GOOGLE_VERTEX_PROJECT}/locations/${GOOGLE_VERTEX_LOCATION}/endpoints/openapi",
        "apiKey": "{file:./google-vertex-ai/api.key}"
      },

But I will have to manually refresh the key since gcloud apis are valid for 1 hour only.

What am I missing?

leehack · 2026-02-16T21:39:03Z

@NimaAra I haven't tested version 1.2.6 yet, but I have tested on the dev branch. I copy-pasted your configuration, and it worked.

I suspect that you may not have properly logged in using gcloud.

NimaAra · 2026-02-16T21:44:57Z

@leehack Do you also use GOOGLE_APPLICATION_CREDENTIALS pointing to your service.account.json or do you use another method?

leehack · 2026-02-16T21:49:02Z

@NimaAra For my configuration, I am not utilizing the JSON key. Instead, it should authenticate using the account I used when logging in with gcloud. I don't see any reason why the GAC is not functioning.

NimaAra · 2026-02-16T22:20:31Z

Got it, had to run gcloud auth application-default login I was running gcloud auth login instead.

Works great. Once again, thank you for the PR.

rekram1-node · 2026-02-16T23:53:32Z

Yes big thanks to @leehack this was a failure on our part. The google vertex ux through opencode is poor. We will improve.

github-actions bot added the needs:issue label Jan 23, 2026

github-actions bot removed the needs:issue label Jan 23, 2026

This was referenced Jan 26, 2026

feat(provider): add google-vertex-openai for Vertex AI partner models #10742

Closed

feat: added Google OAuth #11116

Closed

leehack added 3 commits January 29, 2026 09:51

fix(provider): add google-vertex-openapi provider for GLM-4 model sup…

f8d1fc3

…port

refactor: use ADC with google-auth-library instead of gcloud subprocess

c4e4b33

Consolidate the provider into google-vertex

5aba6e0

leehack force-pushed the fix/vertex-glm-openapi branch from 377d80e to 5aba6e0 Compare January 29, 2026 18:46

NimaAra mentioned this pull request Jan 29, 2026

Unable to access GLM 4.7 on Google Vertex AI #9909

Closed

thdxr force-pushed the dev branch from cbab81f to 2d3c7a0 Compare January 30, 2026 04:49

opencode-agent bot force-pushed the dev branch from 00637c0 to 71e0ba2 Compare January 30, 2026 14:32

thdxr force-pushed the dev branch 4 times, most recently from f1ae801 to 08fa7f7 Compare January 30, 2026 14:37

bluet reviewed Feb 2, 2026

View reviewed changes

packages/opencode/src/provider/provider.ts Outdated Show resolved Hide resolved

packages/opencode/src/provider/provider.ts Outdated Show resolved Hide resolved

Apply suggestions from code review

7b520a0

Co-authored-by: BlueT - Matthew Lien - 練喆明 <BlueT@BlueT.org>

leehack and others added 2 commits February 3, 2026 08:43

Merge branch 'dev' into fix/vertex-glm-openapi

3d0268d

fix(provider): surgically sanitize Google SDK options to restore Gemi…

2bd9b14

…ni in production while allowing custom proxies and localhost testing

leehack added 3 commits February 12, 2026 07:38

remove protocol field

53948d4

Merge branch 'dev' into fix/vertex-glm-openapi

1b5007c

fix typecheck

58a729c

leehack added 2 commits February 12, 2026 07:55

fix: update tests and mocks for protocol removal and dependency fixes

65dcb36

chore: restore src/provider/models.ts from dev

7536184

leehack changed the title ~~feat: add google-vertex-openapi provider with ADC authentication~~ feat: add openai-compatible endpoint support for google-vertex provider Feb 12, 2026

leehack added 4 commits February 12, 2026 08:19

test: revert broad provider test changes AND add focused tests

8758bf4

add tests

483e4eb

clean up

da72b60

test clean up

690266a

bluet mentioned this pull request Feb 14, 2026

feat: add google-vertex-openai provider for Vertex AI partner models anomalyco/models.dev#716

Open

Merge branch 'dev' into fix/vertex-glm-openapi

39441f9

rekram1-node reviewed Feb 16, 2026

View reviewed changes

packages/opencode/src/provider/provider.ts Outdated Show resolved Hide resolved

some tweaks

4507995

rekram1-node merged commit f7708ef into anomalyco:dev Feb 16, 2026
6 checks passed

Conversation

leehack commented Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Verification

Usage

Setup

Uh oh!

github-actions bot commented Jan 23, 2026

Uh oh!

github-actions bot commented Jan 23, 2026

Uh oh!

NimaAra commented Jan 29, 2026

Uh oh!

bluet commented Jan 30, 2026

Uh oh!

leehack commented Jan 30, 2026

Uh oh!

bluet commented Feb 2, 2026

Test Results: v1 vs v1beta1 Endpoints

Results

Conclusion

Documentation Reference

Uh oh!

bluet commented Feb 2, 2026

Follow-up: Recommendation to use /v1/ instead of /v1beta1/

Why /v1/ is preferable:

Reference: Google's Official Policy

Suggested Change

Uh oh!

bluet left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

leehack commented Feb 3, 2026

Uh oh!

leehack commented Feb 3, 2026

Uh oh!

NimaAra commented Feb 5, 2026

Uh oh!

leehack commented Feb 6, 2026

Uh oh!

ddesmond commented Feb 10, 2026

Uh oh!

KuzonCode commented Feb 11, 2026

Uh oh!

rekram1-node commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

klinki commented Feb 12, 2026

Uh oh!

leehack commented Feb 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

leehack commented Feb 12, 2026

Uh oh!

ddesmond commented Feb 13, 2026

Uh oh!

Uh oh!

rekram1-node commented Feb 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

NimaAra commented Feb 16, 2026

Uh oh!

leehack commented Feb 16, 2026

Uh oh!

NimaAra commented Feb 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

leehack commented Feb 16, 2026

Uh oh!

NimaAra commented Feb 16, 2026

Uh oh!

rekram1-node commented Feb 16, 2026

Uh oh!

Reviewers

Assignees

leehack commented Jan 23, 2026 •

edited

Loading

Follow-up: Recommendation to use `/v1/` instead of `/v1beta1/`

Why `/v1/` is preferable:

rekram1-node commented Feb 11, 2026 •

edited

Loading

leehack commented Feb 12, 2026 •

edited

Loading

rekram1-node commented Feb 16, 2026 •

edited

Loading

NimaAra commented Feb 16, 2026 •

edited

Loading