fix(lightspeed): added empty state for unconfigured LLM

[ciiay]

Hey, I just made a Pull Request!

For RHIDP-12816

✔️ Checklist

• A changeset describing the change and affected packages. (more info)
• Added or Updated documentation
• Tests for new functionality and regression tests for bug fixes
• Screenshots attached (for UI changes)

Test steps:

1. In run.yaml comment out your provider config in the inference array, for example:

providers:
  ...
  inference:
    - config:
        api_token: ${env.VLLM_API_KEY:=}
        base_url: ${env.VLLM_URL:=}
        max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
        network:
          tls:
            verify: ${env.VLLM_TLS_VERIFY:=true}
      provider_id: ${env.ENABLE_VLLM:+vllm}
      provider_type: remote::vllm
    # - config:
    #     base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
    #   provider_id: ${env.ENABLE_OLLAMA:+ollama}
    #   provider_type: remote::vllm
    ...

2. Remove local persistence (matches whatever paths you have in storage.backend in your run.yaml)

rm -f /tmp/kvstore.db /tmp/sql_store.db

3. Restart Llama Stack and LCS

Screenshot(updated on 4/16):

[rhdh-qodo-merge]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

1. Error blocks usable chat ☑ 🐞 ☼

Description

LightspeedChatContainerInner always renders the error empty-state when useAllModels().isError is
true, even if useAllModels().data already contains LLM models (modelsItems non-empty). This can
unnecessarily hide the chat UI (and force a retry) in scenarios where model data is still usable
(e.g., stale/cached data or partial failures).

Code

workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatContainer.tsx[R176-182]

+  if (modelsError) {
+    return <ModelsLoadErrorEmptyState onRetry={() => refetchModels()} />;
+  }
+
+  if (modelsItems.length === 0) {
+    return <LcoreNotConfiguredEmptyState />;
+  }

Evidence

The container derives modelsItems from the query data (models) and then checks modelsError before
checking whether modelsItems is empty; because data and error state are not mutually exclusive in
the hook contract, this ordering can select the error screen even when modelsItems has entries.

workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatContainer.tsx[56-86]
workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatContainer.tsx[172-182]
workspaces/lightspeed/plugins/lightspeed/src/hooks/useAllModels.ts[25-35]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`LightspeedChatContainerInner` shows `ModelsLoadErrorEmptyState` whenever `isError` is true, even if there is still model data available (i.e., `modelsItems.length > 0`). This can block the chat UI unnecessarily.

### Issue Context
`modelsItems` is computed from `useAllModels().data` and may be non-empty while `useAllModels().isError` is true (the hook’s return type does not guarantee these are mutually exclusive).

### Fix Focus Areas
- workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatContainer.tsx[172-182]

### Suggested change
Adjust the conditional to only show the error empty-state when there is no usable model data, e.g.:
- `if (modelsError && modelsItems.length === 0) return <ModelsLoadErrorEmptyState ... />;`

Optionally, if `modelsItems.length > 0` and `modelsError` is true, keep rendering the chat and surface a non-blocking warning/retry control.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

ⓘ The new review experience is currently in Beta. Learn more

[rhdh-gh-app]

Missing Changesets

The following package(s) are changed by this PR but do not have a changeset:

• @red-hat-developer-hub/backstage-plugin-lightspeed

See CONTRIBUTING.md for more information about how to add changesets.

Changed Packages

Package Name	Package Path	Changeset Bump	Current Version
@red-hat-developer-hub/backstage-plugin-lightspeed	workspaces/lightspeed/plugins/lightspeed	none	v2.0.0

[rhdh-qodo-merge]

Review Summary by Qodo

Add empty and error states for Lightspeed LLM configuration

✨ Enhancement

Walkthroughs

Description

• Added empty state UI for unconfigured LLM models
• Added error state UI for failed model loading
• Added loading state UI while fetching models
• Extended translations across 6 languages
• Added comprehensive test coverage for new states

Diagram

flowchart LR
  A["Models API"] -->|loading| B["Loading State"]
  A -->|error| C["Error State with Retry"]
  A -->|empty list| D["Not Configured State"]
  A -->|success| E["Chat Interface"]

Loading

File Changes

1. workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatModelsState.tsx ✨ Enhancement +196/-0

New component for model state UI displays

workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatModelsState.tsx

---

2. workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatContainer.tsx ✨ Enhancement +23/-1

Integrated model state handling and displays

workspaces/lightspeed/plugins/lightspeed/src/components/LightspeedChatContainer.tsx

---

3. workspaces/lightspeed/plugins/lightspeed/src/translations/ref.ts 📝 Documentation +11/-0

Added English translations for model states

workspaces/lightspeed/plugins/lightspeed/src/translations/ref.ts

---

View more (6)

4. workspaces/lightspeed/plugins/lightspeed/src/translations/de.ts 📝 Documentation +9/-0

Added German translations for model states

workspaces/lightspeed/plugins/lightspeed/src/translations/de.ts

---

5. workspaces/lightspeed/plugins/lightspeed/src/translations/es.ts 📝 Documentation +9/-0

Added Spanish translations for model states

workspaces/lightspeed/plugins/lightspeed/src/translations/es.ts

---

6. workspaces/lightspeed/plugins/lightspeed/src/translations/fr.ts 📝 Documentation +9/-0

Added French translations for model states

workspaces/lightspeed/plugins/lightspeed/src/translations/fr.ts

---

7. workspaces/lightspeed/plugins/lightspeed/src/translations/it.ts 📝 Documentation +9/-0

Added Italian translations for model states

workspaces/lightspeed/plugins/lightspeed/src/translations/it.ts

---

8. workspaces/lightspeed/plugins/lightspeed/src/translations/ja.ts 📝 Documentation +9/-0

Added Japanese translations for model states

workspaces/lightspeed/plugins/lightspeed/src/translations/ja.ts

---

9. workspaces/lightspeed/plugins/lightspeed/src/components/__tests__/LightspeedPage.test.tsx 🧪 Tests +105/-3

Added tests for model loading and error states

workspaces/lightspeed/plugins/lightspeed/src/components/tests/LightspeedPage.test.tsx

---

[ciiay]

Hi @ShiranHi , any input to the empty state display?

[aprilma419]

Hi, I’m no longer working with Red Hat. Please remove me from your list and find @shiran for any UX questions. Thank you April Ma Sr. Interaction Designer, User Experience Design Red Hat <https://www.redhat.com/> <https://www.redhat.com/>

…

On Thu, Apr 16, 2026 at 02:43 Yi C ***@***.***> wrote: *ciiay* left a comment (redhat-developer/rhdh-plugins#2781) <#2781 (comment)> Hi @aprilma419 <https://github.com/aprilma419> , any input to the empty state display? — Reply to this email directly, view it on GitHub <#2781 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKS7NJ3HMAFE6RWU6CEMQML4V7J6HAVCNFSM6AAAAACX2SBGWGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DENJUGU3TENBRGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[ShiranHi]

@ciiay looks good to me, I have some small comments:

1. Let's change the text to:
   Title: Connect an LLM to get started
   Text: To use Lightspeed's AI capabilities, you'll need to register an OpenAI-compatible LLM with your Llama Stack deployment.
2. Change the first button label to be "Configure Llama Stack"
3. Change the second button label to be "Lightspeed Backend Setup" should be a link and not a secondary button
4. Remove the top X button, users can close this window from the floating button

[ciiay]

Hi @ShiranHi , thanks for the input. I have updated the screenshot in the PR description. Can you take a look and see if it looks good now?

[ShiranHi]

Hi @ShiranHi , thanks for the input. I have updated the screenshot in the PR description. Can you take a look and see if it looks good now?

Thank you! Looks good to me!

[karthikjeeyar]

This makes the persisting & retaining user mode selection feature only work on overlay and docked mode? I am trying to understand why we need to skip the full-screen mode here, most of the users may use full screen mode, with this change to get to fullscreen then user have to change it manually everytime via mode selector now?

[Jdubrick]

Thanks for this! Couple of things I have in mind / questions:

1. In the blurb about using Lightspeed and connecting to an LLM to get started, I think we should maybe be also mentioning that they should be contacting an administrator to get it setup? A normal user of Lightspeed won't have the access required to set up the secrets required
2. I like the link to Llama Stack, but I think if we leave a link there it should be to our rhdh documentation instead of the llama stack github
3. I noticed in the test steps it instructs to set the inference settings to an empty array [], what version of Lightspeed Core / Llama Stack did you test this against? My concern is that it will never truly be empty as we always will have our RAG transformer in the inference, even if there is no LLM:

I.e.

- provider_id: sentence-transformers
      provider_type: inline::sentence-transformers
      config: {}

The content in main of https://github.com/redhat-ai-dev/lightspeed-configs/tree/main can spin up a local LCORE instance with our latest config set

[ciiay]

Hi @Jdubrick , thanks for the review and feedback.

I noticed in the test steps it instructs to set the inference settings to an empty array [], what version of Lightspeed Core / Llama Stack did you test this against? My concern is that it will never truly be empty as we always will have our RAG transformer in the inference, even if there is no LLM:

Good catch 👍 You’re right about the inference config. I tested against latest main of lightspeed-stack at the time I worked on this PR. I only described an empty array for convenience. I’ve updated the Test steps in the PR description so other reviewers aren’t confused.

I agree with the other two points. I’ll get input from our UX team and update the wording and links accordingly.

[ciiay]

Hi @ShiranHi , any suggestions to these points? Shall we use Installing and configuring Red Hat Developer Lightspeed for Red Hat Developer Hub for the second point?

1. In the blurb about using Lightspeed and connecting to an LLM to get started, I think we should maybe be also mentioning that they should be contacting an administrator to get it setup? A normal user of Lightspeed won't have the access required to set up the secrets required
2. I like the link to Llama Stack, but I think if we leave a link there it should be to our rhdh documentation instead of the llama stack github

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud