issue: Nth Document Provided not Being Used by Model after first document. #5394

New Issue

2025-11-11T16:19:49-06:00

GiteaMirror commented

2025-11-11 16:19:49 -06:00

Originally created by @Ducky6944 on GitHub (May 29, 2025).

Check Existing Issues

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

Installation Method

Docker

Open WebUI Version

v0.6.12

Ollama Version (if applicable)

No response

Operating System

Debian 12

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

Answer should reference current document.

Actual Behavior

Answer references original document.

Steps to Reproduce

Context: All model related settings are default.

Start a new chat with any model
add document to prompt #https://www.youtube.com/watch?v=yCnkkljii_k
give instruction summarize
Get expected Output

1-3. Repeat 1-3 with a new URL (eg: https://www.youtube.com/watch?v=3sTsMVAMF1g).
4. Model uses original document.

You can look in the chat context and see both documents.

Logs & Screenshots

Correct Answer Given

New Document Provided and correct data pulled

Model Uses Original Document (clicking on the document in the second question still shows the correct information.)

Chat has access to both documents

Additional Information

I'm not sure what logs here are relevant here because I know there's a lot of moving parts and I didn't want to spam the issue with useless data. I did include the required logs. If you need anything let me know.

I have tried changing the context length among other things, but currently everything I suspect would be relevant is default.

I did see one similar issue in discussion, but there was no answer.

There were occassional failures to get the "youtube" document.

Logs

logs.txt

Originally created by @Ducky6944 on GitHub (May 29, 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 Docker ### Open WebUI Version v0.6.12 ### Ollama Version (if applicable) _No response_ ### Operating System Debian 12 ### 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 Answer should reference current document. ### Actual Behavior Answer references original document. ### Steps to Reproduce Context: All model related settings are default. 1. Start a new chat with any model 2. add document to prompt `#https://www.youtube.com/watch?v=yCnkkljii_k` 3. give instruction `summarize` 4. Get expected Output 1-3. Repeat 1-3 with a new URL (eg: https://www.youtube.com/watch?v=3sTsMVAMF1g). 4. Model uses original document. You can look in the chat context and see both documents. ### Logs & Screenshots <img width="1568" alt="Image" src="https://github.com/user-attachments/assets/a651ff12-daa7-4e7b-a949-6a5588a00e41" /> Correct Answer Given <img width="1419" alt="Image" src="https://github.com/user-attachments/assets/af2182c2-0c02-4b57-aa5d-920c8194430b" /> New Document Provided and correct data pulled <img width="1506" alt="Image" src="https://github.com/user-attachments/assets/b070eb85-9346-43e9-be2a-704d6e1dae58" /> Model Uses Original Document (clicking on the document in the second question still shows the correct information.) <img width="1513" alt="Image" src="https://github.com/user-attachments/assets/cccb8624-6bba-4ac0-80ca-545a865a0486" /> Chat has access to both documents <img width="1651" alt="Image" src="https://github.com/user-attachments/assets/1f306d12-089a-4138-b88b-39b4ecaa91a7" /> ### Additional Information I'm not sure what logs here are relevant here because I know there's a lot of moving parts and I didn't want to spam the issue with useless data. I did include the required logs. If you need anything let me know. I have tried changing the context length among other things, but currently everything I suspect would be relevant is default. I did see one similar issue in discussion, but there was no answer. There were occassional failures to get the "youtube" document. Logs [logs.txt](https://github.com/user-attachments/files/20511647/details.txt)

GiteaMirror added the bug label 2025-11-11 16:19:49 -06:00

GiteaMirror closed this issue

2025-11-11 16:19:50 -06:00

GiteaMirror commented

2025-11-11 16:19:50 -06:00