Bug in "Continue Response" #1830

New Issue

GiteaMirror · 2025-11-11T14:54:18-06:00

GiteaMirror commented

2025-11-11 14:54:18 -06:00

Originally created by @Tostino on GitHub (Aug 20, 2024).

Bug Report

Installation Method

Docker

Environment

Open WebUI Version: v0.3.8
vllm inference
Operating System: Pop_OS
Browser (if applicable): Firefox

Confirmation:

I have read and followed all the instructions provided in the README.md.
I am on 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 the exact steps to reproduce the bug in the "Steps to Reproduce" section below.

Expected Behavior:

When using the "continue response" button, the API call should pass add_generation_prompt=false to the OpenAI chat endpoint so it can be passed to the template. This should prevent the end of turn ID from being added, allowing the generation to continue where it left off.

Actual Behavior:

The API call for the "continue response" button does not pass add_generation_prompt=false. As a result, the end of turn ID is added, preventing the generation from continuing seamlessly from where it left off.

Description

Bug Summary:
The "continue response" functionality is not working correctly due to the absence of the add_generation_prompt=false parameter in the API call, which is supported by the custom chat template but not being utilized.

Reproduction Details

Steps to Reproduce:

Set up Open WebUI v0.3.8 using Docker with vllm inference.
Use the custom chat template provided in the vllm docker command.
Start a conversation with the AI model using the curl command provided.
Observe that the add_generation_prompt parameter is set to true.
When continuing the response, check if the API call includes add_generation_prompt=false.
Observe that the continuation doesn't seamlessly follow the previous response due to the missing parameter.

Logs and Screenshots

Browser Console Logs:
[Not provided]

Docker Container Logs:
[Not provided]

Screenshots/Screen Recordings (if applicable):
[Not provided]

Additional Information

Example curl command demonstrating the current behavior:

curl -X POST "http://127.0.0.1:8000/v1/chat/completions" \
     -H "Content-Type: application/json" \
     -d '{
           "model": "unsloth/Meta-Llama-3.1-8B-Instruct",
           "temperature": 0,
           "stream": false,
           "messages": [
             {
               "role": "user",
               "content": "tell me a common saying"
             },
             {
               "role": "assistant",
               "content": "Here is a common saying about apple. An apple a day, keeps"
             }
           ],
           "echo": true,
           "add_generation_prompt": false
         }'

vllm docker command with custom chat template:

docker run --runtime nvidia --gpus all \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HUGGING_FACE_HUB_TOKEN=<secret>" \
    -p 8000:8000 \
    --ipc=host \
    vllm/vllm-openai:latest \
    --model unsloth/Meta-Llama-3.1-8B-Instruct \
    --chat-template "{{- bos_token }}\n{%- if custom_tools is defined %}\n    {%- set tools = custom_tools %}\n{%- endif %}\n{%- if not tools_in_user_message is defined %}\n    {%- set tools_in_user_message = true %}\n{%- endif %}\n{%- if not date_string is defined %}\n    {%- set date_string = \"26 Jul 2024\" %}\n{%- endif %}\n{%- if not tools is defined %}\n    {%- set tools = none %}\n{%- endif %}\n\n{#- This block extracts the system message, so we can slot it into the right place. #}\n{%- if messages[0]['role'] == 'system' %}\n    {%- set system_message = messages[0]['content']|trim %}\n    {%- set messages = messages[1:] %}\n{%- else %}\n    {%- set system_message = \"\" %}\n{%- endif %}\n\n{#- System message + builtin tools #}\n{{- \"<|start_header_id|>system<|end_header_id|>\\n\\n\" }}\n{%- if builtin_tools is defined or tools is not none %}\n    {{- \"Environment: ipython\\n\" }}\n{%- endif %}\n{%- if builtin_tools is defined %}\n    {{- \"Tools: \" + builtin_tools | reject('equalto', 'code_interpreter') | join(\", \") + \"\\n\\n\"}}\n{%- endif %}\n{{- \"Cutting Knowledge Date: December 2023\\n\" }}\n{{- \"Today Date: \" + date_string + \"\\n\\n\" }}\n{%- if tools is not none and not tools_in_user_message %}\n    {{- \"You have access to the following functions. To call a function, please respond with JSON for a function call.\" }}\n    {{- 'Respond in the format {\"name\": function name, \"parameters\": dictionary of argument name and its value}.' }}\n    {{- \"Do not use variables.\\n\\n\" }}\n    {%- for t in tools %}\n        {{- t | tojson(indent=4) }}\n        {{- \"\\n\\n\" }}\n    {%- endfor %}\n{%- endif %}\n{{- system_message }}\n{{- \"<|eot_id|>\" }}\n\n{#- Custom tools are passed in a user message with some extra guidance #}\n{%- if tools_in_user_message and not tools is none %}\n    {#- Extract the first user message so we can plug it in here #}\n    {%- if messages | length != 0 %}\n        {%- set first_user_message = messages[0]['content']|trim %}\n        {%- set messages = messages[1:] %}\n    {%- else %}\n        {{- raise_exception(\"Cannot put tools in the first user message when there's no first user message!\") }}\n{%- endif %}\n    {{- '<|start_header_id|>user<|end_header_id|>\\n\\n' -}}\n    {{- \"Given the following functions, please respond with a JSON for a function call \" }}\n    {{- \"with its proper arguments that best answers the given prompt.\\n\\n\" }}\n    {{- 'Respond in the format {\"name\": function name, \"parameters\": dictionary of argument name and its value}.' }}\n    {{- \"Do not use variables.\\n\\n\" }}\n    {%- for t in tools %}\n        {{- t | tojson(indent=4) }}\n        {{- \"\\n\\n\" }}\n    {%- endfor %}\n    {{- first_user_message + \"<|eot_id|>\"}}\n{%- endif %}\n\n{%- for message in messages %}\n    {%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %}\n        {{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\\n\\n'+ message['content'] | trim + '<|eot_id|>' }}\n    {%- elif 'tool_calls' in message %}\n        {%- if not message.tool_calls|length == 1 %}\n            {{- raise_exception(\"This model only supports single tool-calls at once!\") }}\n        {%- endif %}\n        {%- set tool_call = message.tool_calls[0].function %}\n        {%- if builtin_tools is defined and tool_call.name in builtin_tools %}\n            {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' -}}\n            {{- \"<|python_tag|>\" + tool_call.name + \".call(\" }}\n            {%- for arg_name, arg_val in tool_call.arguments | items %}\n                {{- arg_name + '=\"' + arg_val + '\"' }}\n                {%- if not loop.last or add_generation_prompt %}\n                    {{- \", \" }}\n                {%- endif %}\n                {%- endfor %}\n            {{- \")\" }}\n        {%- else  %}\n            {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' -}}\n            {{- '{\"name\": \"' + tool_call.name + '\", ' }}\n            {{- '\"parameters\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- \"}\" }}\n        {%- endif %}\n        {%- if builtin_tools is defined %}\n            {#- This means we're in ipython mode #}\n            {{- \"<|eom_id|>\" }}\n        {%- else %}\n            {{- \"<|eot_id|>\" }}\n        {%- endif %}\n    {%- elif message.role == \"tool\" or message.role == \"ipython\" %}\n        {{- \"<|start_header_id|>ipython<|end_header_id|>\\n\\n\" }}\n        {%- if message.content is mapping or message.content is iterable %}\n            {{- message.content | tojson }}\n        {%- else %}\n            {{- message.content }}\n        {%- endif %}\n        {{- \"<|eot_id|>\" }}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' }}\n{%- endif %}\n"

The custom chat template supports the add_generation_prompt feature. The bug prevents this feature from being utilized correctly when continuing a response.

When looking at the vLLM logs, you will be able to see the templated query sent to the LLM, and see if the chat template was properly applied or not.

Originally created by @Tostino on GitHub (Aug 20, 2024). # Bug Report ## Installation Method Docker ## Environment - **Open WebUI Version:** v0.3.8 - vllm inference - **Operating System:** Pop_OS - **Browser (if applicable):** Firefox **Confirmation:** - [X] I have read and followed all the instructions provided in the README.md. - [ ] I am on the latest version of both Open WebUI and Ollama. - [ ] I have included the browser console logs. - [ ] I have included the Docker container logs. - [X] I have provided the exact steps to reproduce the bug in the "Steps to Reproduce" section below. ## Expected Behavior: When using the "continue response" button, the API call should pass `add_generation_prompt=false` to the OpenAI chat endpoint so it can be passed to the template. This should prevent the end of turn ID from being added, allowing the generation to continue where it left off. ## Actual Behavior: The API call for the "continue response" button does not pass `add_generation_prompt=false`. As a result, the end of turn ID is added, preventing the generation from continuing seamlessly from where it left off. ## Description **Bug Summary:** The "continue response" functionality is not working correctly due to the absence of the `add_generation_prompt=false` parameter in the API call, which is supported by the custom chat template but not being utilized. ## Reproduction Details **Steps to Reproduce:** 1. Set up Open WebUI v0.3.8 using Docker with vllm inference. 2. Use the custom chat template provided in the vllm docker command. 3. Start a conversation with the AI model using the curl command provided. 4. Observe that the `add_generation_prompt` parameter is set to `true`. 5. When continuing the response, check if the API call includes `add_generation_prompt=false`. 6. Observe that the continuation doesn't seamlessly follow the previous response due to the missing parameter. ## Logs and Screenshots **Browser Console Logs:** [Not provided] **Docker Container Logs:** [Not provided] **Screenshots/Screen Recordings (if applicable):** [Not provided] ## Additional Information 1. Example curl command demonstrating the current behavior: ``` curl -X POST "http://127.0.0.1:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ -d '{ "model": "unsloth/Meta-Llama-3.1-8B-Instruct", "temperature": 0, "stream": false, "messages": [ { "role": "user", "content": "tell me a common saying" }, { "role": "assistant", "content": "Here is a common saying about apple. An apple a day, keeps" } ], "echo": true, "add_generation_prompt": false }' ``` 2. vllm docker command with custom chat template: ``` docker run --runtime nvidia --gpus all \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HUGGING_FACE_HUB_TOKEN=<secret>" \ -p 8000:8000 \ --ipc=host \ vllm/vllm-openai:latest \ --model unsloth/Meta-Llama-3.1-8B-Instruct \ --chat-template "{{- bos_token }}\n{%- if custom_tools is defined %}\n {%- set tools = custom_tools %}\n{%- endif %}\n{%- if not tools_in_user_message is defined %}\n {%- set tools_in_user_message = true %}\n{%- endif %}\n{%- if not date_string is defined %}\n {%- set date_string = \"26 Jul 2024\" %}\n{%- endif %}\n{%- if not tools is defined %}\n {%- set tools = none %}\n{%- endif %}\n\n{#- This block extracts the system message, so we can slot it into the right place. #}\n{%- if messages[0]['role'] == 'system' %}\n {%- set system_message = messages[0]['content']|trim %}\n {%- set messages = messages[1:] %}\n{%- else %}\n {%- set system_message = \"\" %}\n{%- endif %}\n\n{#- System message + builtin tools #}\n{{- \"<|start_header_id|>system<|end_header_id|>\\n\\n\" }}\n{%- if builtin_tools is defined or tools is not none %}\n {{- \"Environment: ipython\\n\" }}\n{%- endif %}\n{%- if builtin_tools is defined %}\n {{- \"Tools: \" + builtin_tools | reject('equalto', 'code_interpreter') | join(\", \") + \"\\n\\n\"}}\n{%- endif %}\n{{- \"Cutting Knowledge Date: December 2023\\n\" }}\n{{- \"Today Date: \" + date_string + \"\\n\\n\" }}\n{%- if tools is not none and not tools_in_user_message %}\n {{- \"You have access to the following functions. To call a function, please respond with JSON for a function call.\" }}\n {{- 'Respond in the format {\"name\": function name, \"parameters\": dictionary of argument name and its value}.' }}\n {{- \"Do not use variables.\\n\\n\" }}\n {%- for t in tools %}\n {{- t | tojson(indent=4) }}\n {{- \"\\n\\n\" }}\n {%- endfor %}\n{%- endif %}\n{{- system_message }}\n{{- \"<|eot_id|>\" }}\n\n{#- Custom tools are passed in a user message with some extra guidance #}\n{%- if tools_in_user_message and not tools is none %}\n {#- Extract the first user message so we can plug it in here #}\n {%- if messages | length != 0 %}\n {%- set first_user_message = messages[0]['content']|trim %}\n {%- set messages = messages[1:] %}\n {%- else %}\n {{- raise_exception(\"Cannot put tools in the first user message when there's no first user message!\") }}\n{%- endif %}\n {{- '<|start_header_id|>user<|end_header_id|>\\n\\n' -}}\n {{- \"Given the following functions, please respond with a JSON for a function call \" }}\n {{- \"with its proper arguments that best answers the given prompt.\\n\\n\" }}\n {{- 'Respond in the format {\"name\": function name, \"parameters\": dictionary of argument name and its value}.' }}\n {{- \"Do not use variables.\\n\\n\" }}\n {%- for t in tools %}\n {{- t | tojson(indent=4) }}\n {{- \"\\n\\n\" }}\n {%- endfor %}\n {{- first_user_message + \"<|eot_id|>\"}}\n{%- endif %}\n\n{%- for message in messages %}\n {%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %}\n {{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\\n\\n'+ message['content'] | trim + '<|eot_id|>' }}\n {%- elif 'tool_calls' in message %}\n {%- if not message.tool_calls|length == 1 %}\n {{- raise_exception(\"This model only supports single tool-calls at once!\") }}\n {%- endif %}\n {%- set tool_call = message.tool_calls[0].function %}\n {%- if builtin_tools is defined and tool_call.name in builtin_tools %}\n {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' -}}\n {{- \"<|python_tag|>\" + tool_call.name + \".call(\" }}\n {%- for arg_name, arg_val in tool_call.arguments | items %}\n {{- arg_name + '=\"' + arg_val + '\"' }}\n {%- if not loop.last or add_generation_prompt %}\n {{- \", \" }}\n {%- endif %}\n {%- endfor %}\n {{- \")\" }}\n {%- else %}\n {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' -}}\n {{- '{\"name\": \"' + tool_call.name + '\", ' }}\n {{- '\"parameters\": ' }}\n {{- tool_call.arguments | tojson }}\n {{- \"}\" }}\n {%- endif %}\n {%- if builtin_tools is defined %}\n {#- This means we're in ipython mode #}\n {{- \"<|eom_id|>\" }}\n {%- else %}\n {{- \"<|eot_id|>\" }}\n {%- endif %}\n {%- elif message.role == \"tool\" or message.role == \"ipython\" %}\n {{- \"<|start_header_id|>ipython<|end_header_id|>\\n\\n\" }}\n {%- if message.content is mapping or message.content is iterable %}\n {{- message.content | tojson }}\n {%- else %}\n {{- message.content }}\n {%- endif %}\n {{- \"<|eot_id|>\" }}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' }}\n{%- endif %}\n" ``` The custom chat template supports the `add_generation_prompt` feature. The bug prevents this feature from being utilized correctly when continuing a response. When looking at the vLLM logs, you will be able to see the templated query sent to the LLM, and see if the chat template was properly applied or not.

GiteaMirror closed this issue

2025-11-11 14:54:18 -06:00

GiteaMirror referenced this issue

2025-11-11 17:31:20 -06:00

[PR #1830] [MERGED] feat: add tests for db migration on sqlite and postgres #7604

GiteaMirror referenced this issue

2026-04-20 03:11:42 -05:00

[PR #1830] [MERGED] feat: add tests for db migration on sqlite and postgres #20808

GiteaMirror referenced this issue

2026-04-25 10:18:26 -05:00

[PR #1830] [MERGED] feat: add tests for db migration on sqlite and postgres #36438

GiteaMirror referenced this issue