issue: RAG with gemma-3-12b-instruct causes infinite loop and HTTPException "list object has no attribute 'get'" error #6321

New Issue

GiteaMirror · 2025-11-11T16:51:05-06:00

GiteaMirror commented

2025-11-11 16:51:05 -06:00

Originally created by @dev-jonathan on GitHub (Sep 4, 2025).

Check Existing Issues

I have searched the existing issues and discussions.
I am using the latest version of Open WebUI.

Installation Method

Git Clone

Open WebUI Version

v0.6.26 most recent commit

Ollama Version (if applicable)

No response

Operating System

Windows

Browser (if applicable)

Brave

Confirmation

I have read and followed all instructions in README.md.
I am using the latest version of both Open WebUI and Ollama.
I have included the browser console logs.
I have included the Docker container logs.
I have provided every relevant configuration, setting, and environment variable used in my setup.
I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
Start with the initial platform/version/OS and dependencies used,
Specify exact install/launch/configure commands,
List URLs visited, user input (incl. example values/emails/passwords if needed),
Describe all options and toggles enabled or changed,
Include any files or environmental changes,
Identify the expected and actual result at each stage,
Ensure any reasonably skilled user can follow and hit the same issue.

Expected Behavior

Expected Behavior
Using the gemma-3-12b-instruct model with RAG enabled in the latest OpenWebUI (local setup), the system should process the request and return a response without getting stuck.

Actual Behavior

Actual Behavior
The system enters an infinite loop and no response is returned. The error in the logs shows:
fastapi.exceptions.HTTPException: 400: 'list' object has no attribute 'get'
with a message indicating "Developer instruction is not enabled for models/gemma-3-12b-it".

obs: Without RAG, the gemma-3-12b-instruct model works fine. Also, RAG works correctly with the gemini-2.0-flash model, so the issue seems to be specific to the gemma-3-12b-instruct model combined with RAG.

Steps to Reproduce

Use the latest OpenWebUI version locally with Conda and Node.js.

Set the model to gemma-3-12b-instruct and activate RAG.

Use embeddings from qwen3-embedding running locally on LM Studio and Google AI Studio free API as backend.

Send a request in a chat and observe the infinite loop and error in the logs.

Logs & Screenshots

Additional Information

No response

Originally created by @dev-jonathan on GitHub (Sep 4, 2025). ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Git Clone ### Open WebUI Version v0.6.26 most recent commit ### Ollama Version (if applicable) _No response_ ### Operating System Windows ### Browser (if applicable) Brave ### Confirmation - [x] I have read and followed all instructions in `README.md`. - [x] I am using the latest version of **both** Open WebUI and Ollama. - [x] I have included the browser console logs. - [x] I have included the Docker container logs. - [x] I have **provided every relevant configuration, setting, and environment variable used in my setup.** - [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc). - [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps: - Start with the initial platform/version/OS and dependencies used, - Specify exact install/launch/configure commands, - List URLs visited, user input (incl. example values/emails/passwords if needed), - Describe all options and toggles enabled or changed, - Include any files or environmental changes, - Identify the expected and actual result at each stage, - Ensure any reasonably skilled user can follow and hit the same issue. ### Expected Behavior Expected Behavior Using the gemma-3-12b-instruct model with RAG enabled in the latest OpenWebUI (local setup), the system should process the request and return a response without getting stuck. ### Actual Behavior Actual Behavior The system enters an infinite loop and no response is returned. The error in the logs shows: fastapi.exceptions.HTTPException: 400: 'list' object has no attribute 'get' with a message indicating "Developer instruction is not enabled for models/gemma-3-12b-it". obs: Without RAG, the gemma-3-12b-instruct model works fine. Also, RAG works correctly with the gemini-2.0-flash model, so the issue seems to be specific to the gemma-3-12b-instruct model combined with RAG. ### Steps to Reproduce Steps to Reproduce Use the latest OpenWebUI version locally with Conda and Node.js. Set the model to gemma-3-12b-instruct and activate RAG. Use embeddings from qwen3-embedding running locally on LM Studio and Google AI Studio free API as backend. Send a request in a chat and observe the infinite loop and error in the logs. ### Logs & Screenshots <img width="1208" height="824" alt="Image" src="https://github.com/user-attachments/assets/89f31574-c3ea-4f38-bad6-9804c884082e" /> ### Additional Information _No response_

GiteaMirror added the bug label 2025-11-11 16:51:05 -06:00

GiteaMirror closed this issue

2025-11-11 16:51:06 -06:00

GiteaMirror commented

2025-11-11 16:51:07 -06:00

@tjbck commented on GitHub (Sep 5, 2025):

Is the inference provider openai compatible? the format shown above does not follow the spec.

@tjbck commented on GitHub (Sep 5, 2025): Is the inference provider openai compatible? the format shown above does not follow the spec.

GiteaMirror commented

2025-11-11 16:51:07 -06:00

@tjbck commented on GitHub (Sep 5, 2025):

1bc9711afd