[GH-ISSUE #21493] issue: [Bug] Vision images no longer forwarded to external OpenAI – v0.8.2 #19500

New Issue

GiteaMirror · 2026-04-20T01:58:23-05:00

GiteaMirror commented

2026-04-20 01:58:23 -05:00

Originally created by @LennartJKlein on GitHub (Feb 16, 2026).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/21493

Check Existing Issues

I have searched for any existing and/or related issues.
I have searched for any existing and/or related discussions.
I have also searched in the CLOSED issues AND CLOSED discussions and found no related items (your issue might already be addressed on the development branch!).
I am using the latest version of Open WebUI.

Installation Method

Docker

Open WebUI Version

v0.8.2

Ollama Version (if applicable)

No response

Operating System

Linux-based (linux/amd64) container

Browser (if applicable)

No response

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

When using an OpenAI vision-capable model and uploading an image in chat, Open WebUI should forward the image to the OpenAI /completions request as either:

a valid image_url.url (fully qualified fetchable URL), or
a data:image/...;base64,... inline URL

so the model receives image pixels and can answer questions about the image.

Actual Behavior

The image upload succeeds and is stored as a file (visible in the message metadata), but the outgoing /completions payload contains only an internal UUID in the image_url.url field.

Example from outgoing payload:

{
  "type": "image_url",
  "image_url": {
    "url": "38cd7bec-6497-4a31-8f28-931efe72544f"
  }
}

This UUID is not resolvable by OpenAI, so the model replies as if it cannot access/see the image.

Steps to Reproduce

Start with a clean machine (or new VM) running YOUR OS.
Install Docker YOUR DOCKER VERSION and start the Docker service.
Run Open WebUI v0.8.2 using the command shown above (no custom patches).
- Ensure WEBUI_URL is set (I tested both:
  - set via environment variables, and
  - set via Admin UI setting)
Open YOUR BROWSER + VERSION in incognito/private mode.
Navigate to http(s)://YOUR_WEBUI_HOST/
Create a user / log in.
Configure Settings → Connections → OpenAI:
- Base URL: https://api.openai.com/v1 (or your custom OpenAI-compatible gateway)
- API key: set
Select a vision-capable model (e.g., gpt-5.2) in the chat model selector.
In a new chat:
- Upload an image (PNG)
- Send prompt: “What do you see here?”
Observe the model response indicates it cannot view the image.
Open browser DevTools → Network:

Find the request to /completions (or relevant Open WebUI backend endpoint that forwards it)
Inspect request payload and confirm the image part is sent as:
- image_url.url = "<UUID>" (not a full URL and not base64)

Expected result at step 10: model describes the uploaded image.
Actual result at step 10: model states it cannot access/see image content.

Logs & Screenshots

This behavior of Open WebUI v0.8.2 is by design. So it is more of a degrading issue than a bug.

Additional Information

This appears to correlate with 0.8.x moving image uploads to server-side file storage. In my setup, the stored file’s UUID is being forwarded directly in image_url.url rather than being expanded into a usable URL or inlined as base64.
If there is an intended configuration flag to control “inline base64 vs hosted URL” for external providers, I could not find it; happy to test if pointed to the correct setting.

Originally created by @LennartJKlein on GitHub (Feb 16, 2026). Original GitHub issue: https://github.com/open-webui/open-webui/issues/21493 ### Check Existing Issues - [x] I have searched for any existing and/or related issues. - [x] I have searched for any existing and/or related discussions. - [x] I have also searched in the CLOSED issues AND CLOSED discussions and found no related items (your issue might already be addressed on the development branch!). - [x] I am using the latest version of Open WebUI. ### Installation Method Docker ### Open WebUI Version v0.8.2 ### Ollama Version (if applicable) _No response_ ### Operating System Linux-based (linux/amd64) container ### Browser (if applicable) _No response_ ### 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 When using an OpenAI vision-capable model and uploading an image in chat, Open WebUI should forward the image to the OpenAI `/completions` request as either: - a valid `image_url.url` (fully qualified fetchable URL), or - a `data:image/...;base64,...` inline URL so the model receives image pixels and can answer questions about the image. ### Actual Behavior The image upload succeeds and is stored as a file (visible in the message metadata), but the outgoing `/completions` payload contains only an internal UUID in the `image_url.url` field. Example from outgoing payload: ```json { "type": "image_url", "image_url": { "url": "38cd7bec-6497-4a31-8f28-931efe72544f" } } ``` This UUID is not resolvable by OpenAI, so the model replies as if it cannot access/see the image. ### Steps to Reproduce 1. Start with a clean machine (or new VM) running **YOUR OS**. 2. Install Docker **YOUR DOCKER VERSION** and start the Docker service. 3. Run Open WebUI v0.8.2 using the command shown above (no custom patches). * Ensure `WEBUI_URL` is set (I tested both: * set via environment variables, and * set via Admin UI setting) 4. Open **YOUR BROWSER + VERSION** in incognito/private mode. 5. Navigate to **http(s)://YOUR_WEBUI_HOST/** 6. Create a user / log in. 7. Configure **Settings → Connections → OpenAI**: * Base URL: `https://api.openai.com/v1` (or your custom OpenAI-compatible gateway) * API key: set 8. Select a vision-capable model (e.g., `gpt-5.2`) in the chat model selector. 9. In a new chat: * Upload an image (PNG) * Send prompt: “What do you see here?” 10. Observe the model response indicates it cannot view the image. 11. Open browser DevTools → Network: * Find the request to `/completions` (or relevant Open WebUI backend endpoint that forwards it) * Inspect request payload and confirm the image part is sent as: * `image_url.url = "<UUID>"` (not a full URL and not base64) **Expected result at step 10:** model describes the uploaded image. **Actual result at step 10:** model states it cannot access/see image content. ### Logs & Screenshots This behavior of Open WebUI v0.8.2 is by design. So it is more of a degrading issue than a bug. ### Additional Information * This appears to correlate with 0.8.x moving image uploads to server-side file storage. In my setup, the stored file’s UUID is being forwarded directly in `image_url.url` rather than being expanded into a usable URL or inlined as base64. * If there is an intended configuration flag to control “inline base64 vs hosted URL” for external providers, I could not find it; happy to test if pointed to the correct setting.

GiteaMirror added the bug label 2026-04-20 01:58:23 -05:00

GiteaMirror closed this issue

2026-04-20 01:58:23 -05:00

GiteaMirror commented

2026-04-20 01:58:25 -05:00

@Classic298 commented on GitHub (Feb 16, 2026):

fixed in dev - confirmation wanted (also: the third duplicate)

@Classic298 commented on GitHub (Feb 16, 2026): fixed in dev - confirmation wanted (also: the third duplicate)

GiteaMirror referenced this issue

2026-04-20 02:11:22 -05:00

[GH-ISSUE #22251] issue: .split error causes no responses in chat, stop sequences and possibly MCP servers involved (v0.8.8) #19673

GiteaMirror referenced this issue