[BUG] Error - Invalid API endpoint. Please check your URL configuration.

[wmfeng007]

Problem (one or two sentences)

Always and only show Error - Invalid API endpoint. Please check your URL configuration.

Context (who is affected and when)

i do not else, but certainly me.

Reproduction steps

1. windows VS Code.
2. Embedder Provider info are correct
3. Qdrant API ok
4. Qdrant installed in docker and curl http://localhost:6333 no problem.

Expected result

fix the bug

Actual result

Make sure i can use codebase indexing.

Variations tried (optional)

swtiched between every one. Only one Error - Invalid API endpoint. Please check your URL configuration showing up, no matter how i change the setting.

App Version

V3.36.16

API Provider (optional)

None

Model Used (optional)

Deepseek and Qwen

Roo Code Task Links (optional)

Not task related.

Relevant logs or errors (optional)

[roomote]

Thanks for reporting this issue. I have investigated the codebase and need some clarification to better understand the root cause.

The error message "Invalid API endpoint. Please check your URL configuration" is displayed when the embedder service returns an HTTP 404 error during configuration validation. This error is related to the embedding provider configuration, not the Qdrant vector store.

Important distinction:

• Qdrant is the vector database that stores the embeddings (you mentioned this is working at http://localhost:6333)
• Embedder Provider is the embedding service that generates the vector embeddings (OpenAI, Ollama, OpenAI-Compatible, Gemini, etc.)

Could you please clarify the following:

1. Which embedder provider are you selecting in the settings? (OpenAI, Ollama, OpenAI-Compatible, Gemini, Mistral, Bedrock, or OpenRouter?)

2. What specific settings are you entering for that provider? For example:

   • If using OpenAI: What API key format are you using?
   • If using OpenAI-Compatible: What base URL and API key are you using for Deepseek/Qwen?
   • If using Ollama: What base URL and model ID are you specifying?

3. Are Deepseek and Qwen running locally (e.g., through Ollama or a local server) or are you connecting to their cloud APIs? If cloud, which specific endpoint URL are you using?

4. Could you share the exact configuration you are entering (with API keys redacted) for one of the providers that is failing?

This information will help me determine whether this is a configuration issue or an actual bug in Roo Code.

Please tag @roomote in your response with the answers, and I will be happy to investigate further or propose a fix.

[wmfeng007]

I am using Openai Compatible for Deepseek and Qwen, using the cloud api.
[For Deepseek it is ]
(https://api.deepseek.com/v1)
deepseek-embedding-v2/deepseek-embedding

[Qwen it is ]
(https://dashscope.aliyuncs.com/api/v1)
(https://dashscope.aliyuncs.com/compatible-mode/v1)
/qwen2.5-vl-embedding/text-embedding-v4

all failing.

[EricAtORS]

I have the same issue with using openrouter and using the local qdrant.

When I open the roo code output, I see this:

[CodeIndexManager] Error during background CodeIndexManager configuration/indexing for /var/home/eric/git/Project/src/frontend: Invalid API endpoint. Please check your URL configuration.
Error starting indexing: CodeIndexManager not initialized. Call initialize() first.

[superturbochad]

I have the same error with openrouter and any model I select using local qdrant