Unable to chat with an Agent

[LufinityLucas]

What happened?

I have enabled the Agent Builder and created an agent named Test. However, when chatting with it, I get error Something went wrong. Here's the specific error message we encountered: No model spec selected. Indeed no model spec is selected in the top menu:

If I select a model spec (I have two of them configured in the YAML file), the endpoint changes back to my Azure OpenAI. Selecting the Agent endpoint again and now a model spec is selected:

However, when chatting now I get error Something went wrong. Here's the specific error message we encountered: Model spec mismatch.

Relevant parts from my YAML file:

...
interface:
  endpointsMenu: true
  modelSelect: true
  parameters: false
  sidePanel: true
  presets: false
  prompts: true
  bookmarks: true
  multiConvo: false
...
modelSpecs:
  enforce: false
  prioritize: false
  list:
    - name: "ada-gpt"
      label: "Ada"
      preset:
        default: true
      ...
...

I don't know if I have a specific set of configuration settings causing this, or if it's actually a bug. I would suspect no model spec to be selected when using your own agent.

Steps to Reproduce

1. Configure the YAML file as per above
2. Add two specs to the modelSpecs section, where one is set to default
3. Go to the web UI and start a new chat.
4. Select Agent from the endpoint menu.
5. Select an agent (or create a new one using the Agent Builder and select it)
6. Write something to the bot. Observere the error.
7. Select a model spec, and change back to Agent in the endpoint menu.
8. Write something to the bot. Observe the error.

What browsers are you seeing the problem on?

Microsoft Edge

Relevant log output

No response

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

This could not be reproduced, see my video:
https://github.com/user-attachments/assets/1f9f7845-0bb4-4dde-812a-295c5e74db9a

This is relevant part of my config, matching what you wrote:

interface:
  endpointsMenu: true
  modelSelect: true
  parameters: false
  sidePanel: true
  presets: false
  prompts: true
  bookmarks: true
  multiConvo: false
modelSpecs:
  enforce: false
  prioritize: false
  list:
    - name: "Phi4"
      label: "Phi4"
      iconURL: https://cdn1.cloud.e-infra.cz/logos/meta-llama.png
      showIconInMenu: true
      showIconInHeader: true
      preset:
        endpoint: "Ollama"
        model: "phi4:latest"
        modelLabel: "Phi4"
        maxContextTokens: 15000
    - name: "Max"
      label: "Max"
      default: true
      description: "Smartest model"
      iconURL: "https://code.librechat.ai/_next/image?url=%2Ficon.png&w=32&q=75"
      preset:
        endpoint: "bedrock"
        modelLabel: "Max"
        model: "anthropic.claude-3-haiku-20240307-v1:0"
        greeting: |
          Hello there!

Try updating, as there were many fixes to modelSpecs in the last week:

https://www.librechat.ai/docs/local/docker#update-librechat

[LufinityLucas]

Thanks for chekcing it out!

I actually upgraded just before making the post. I noticed the agent thing in the change log and wanted to try it out. :) We have the same settings indeed, although I'm using Azure Open AI instead, but I don't see how that should make any difference.

So, to troublesoot it further, I enabled debugging and restarted all containers. And guess what... Today it works. :) None of my 100 restarts the other day worked, but today was my lucky day.