Deploying from Source on Windows Help

[imamiao]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

I compiled and deployed Dify from source on Windows. It’s basically working now, but there are still a few minor issues that need to be resolved. The Dify version is 1.11.2, and the Dify-plugin-daemon version is 0.5.1.

1. I’ve handled many Windows path–related issues in dify-plugin-daemon, and now plugins can be downloaded, compiled, and run properly. However, in the dify-plugin database, the declaration field in the plugin_declarations table has values, but the models field is [], which results in 0 models being displayed. I also deployed a set in Docker and copied the models content from there into the corresponding fields, which fixed the 0 models issue. Could this problem still be related to Windows paths? (I’m not sure whether I should ask this in dify-plugin-daemon or here.)

2. When processing knowledge base files, I get the error: Provider langgenius/openai/openai does not exist. Upon checking the database, I found that the provider_orders, provider_models, provider_model_settings, and provider_model_credentials tables in the dify database are empty, whereas the Docker version contains data.
   logs:
   2026-01-06 09:40:28,124 ERROR [indexing_runner.py:55] consume document failed
   Traceback (most recent call last):
   File "C:\dify\dify-1.11.2\api\core\indexing_runner.py", line 97, in run
   documents = self._transform(
   ^^^^^^^^^^^^^^^^
   File "C:\dify\dify-1.11.2\api\core\indexing_runner.py", line 755, in _transform
   embedding_model_instance = self.model_manager.get_model_instance(
   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   File "C:\dify\dify-1.11.2\api\core\model_manager.py", line 492, in get_model_instance
   provider_model_bundle = self._provider_manager.get_provider_model_bundle(
   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   File "C:\dify\dify-1.11.2\api\core\provider_manager.py", line 275, in get_provider_model_bundle
   :return:
   ^^^^^
   ValueError: Provider langgenius/openai/openai does not exist.

Could someone help me with this, give me some advice, or point me in the right direction? I would really appreciate it.

2. Additional context or comments

No response

[dosubot[bot]]

Hi @imamiao! I'm Dosu and I’m helping the dify team.

For the empty models field in plugin_declarations, this is not caused by Windows path handling in Dify itself. The models list is populated from the plugin daemon's manifest response, so if it's empty on Windows but not Docker, the issue is likely with the plugin daemon or how the plugin package is built or executed on Windows. Double-check the plugin daemon logs and ensure the manifest includes models as expected. You might want to compare the manifest output between your Windows and Docker setups to spot differences. Reference

For the missing provider data (and the resulting "Provider langgenius/openai/openai does not exist" error), Dify only populates provider tables automatically if the edition is set to CLOUD and the relevant environment variables (like HOSTED_OPENAI_TRIAL_ENABLED, HOSTED_OPENAI_API_KEY) are configured. If these aren't set or the edition isn't CLOUD, the provider tables will remain empty, causing your error. This logic is not OS-specific, so check your environment variables and edition setting on Windows. You may need to manually set these variables or switch the edition to CLOUD if you want automatic provider population. Reference

Also, make sure your database migrations have run successfully and your database engine is initialized properly. A recent fix ensures the SQLAlchemy engine is built eagerly at startup, which could help with database initialization issues. Reference

Checklist for troubleshooting:

1. Verify the plugin daemon manifest includes models on Windows (compare with Docker).
2. Confirm your environment variables for hosted providers are set and valid.
3. Set EDITION=CLOUD if you want automatic provider/model population.
4. Ensure all database migrations have run and the engine initializes without errors.

If you need to manually populate provider tables, you can copy the relevant data from your Docker setup as a workaround, but ideally, the above steps should resolve the root cause.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}