@RunLLM is it possible to hide AI providers in the Phoenix interface?

@RunLLM is it possible to hide AI providers in the Phoenix interface? | Arize AI Community

5 comments

· Sorted by Oldest

RunLLM
APP
·
Yes, Phoenix can hide AI providers in the interface if their required dependencies are not installed. The list of available providers in the playground is filtered to show only those with installed dependencies, so users only see relevant providers based on their environment setup. This is handled automatically and does not require manual hiding of providers in the UI. If you want to further customize which providers appear, you can control this by installing only the client packages you need (e.g., `pip install 'arize-phoenix[openai]'`), and the UI will reflect only those providers. There is currently no built-in UI toggle to manually hide providers beyond this dependency-based filtering. Would you like more detail on how this filtering works or how to install specific provider dependencies? Sources: - <https://github.com/Arize-ai/phoenix/pull/5251|feat: Only show providers if dependencies are installed> - <https://github.com/Arize-ai/phoenix/issues/5116|filter down the list of providers depending on whether or not client packages are installed>
💡 Hint: Mention RunLLM in the thread for followups.
Brandon B.
·
RunLLM What about setting an environment variable in case I'm using the docker container that has all of them by default?
RunLLM
APP
·
It seems likely that Phoenix does not currently support an environment variable to selectively hide AI providers in the UI when using the all-in-one Docker container. The filtering is based on installed dependencies, so with the full container, all providers are shown by default. There is no documented environment variable for hiding providers at runtime in this scenario. Would you like more detail on possible workarounds or custom container builds? Sources: - <https://github.com/Arize-ai/phoenix/blob/main/docs/prompt-engineering/how-to-prompts/configure-ai-providers.md|https://github.com/Arize-ai/phoenix/blob/main/docs/prompt-engineering/how-to-prompts/configure-ai-providers.md>
💡 Hint: Mention RunLLM in the thread for followups.
Brandon B.
·
RunLLM that's okay. How about changing the default model list? Is that possible?
RunLLM
APP
·
Yes, you can change the default model list in Phoenix. The platform allows you to save and load model configurations per provider, and these defaults are persisted (e.g., in local storage) and used when switching providers in the playground. You can add custom model names, and the UI supports entering arbitrary model names, not just those in the predefined list. Would you like more detail on how to configure or persist these defaults? Sources: - <https://github.com/Arize-ai/phoenix/issues/5069|save off model config by provider and use as default in playground> - <https://github.com/Arize-ai/phoenix/pull/5645|feat(playground): support arbitrary string model names> - <https://github.com/Arize-ai/phoenix/pull/5426|fix(playgroun): use saved model config for default model when switching providers>
💡 Hint: Mention RunLLM in the thread for followups.

RunLLM
APP
·
Yes, Phoenix can hide AI providers in the interface if their required dependencies are not installed. The list of available providers in the playground is filtered to show only those with installed dependencies, so users only see relevant providers based on their environment setup. This is handled automatically and does not require manual hiding of providers in the UI. If you want to further customize which providers appear, you can control this by installing only the client packages you need (e.g., `pip install 'arize-phoenix[openai]'`), and the UI will reflect only those providers. There is currently no built-in UI toggle to manually hide providers beyond this dependency-based filtering. Would you like more detail on how this filtering works or how to install specific provider dependencies? Sources: - <https://github.com/Arize-ai/phoenix/pull/5251|feat: Only show providers if dependencies are installed> - <https://github.com/Arize-ai/phoenix/issues/5116|filter down the list of providers depending on whether or not client packages are installed>
💡 Hint: Mention RunLLM in the thread for followups.
Brandon B.
·
RunLLM What about setting an environment variable in case I'm using the docker container that has all of them by default?
RunLLM
APP
·
It seems likely that Phoenix does not currently support an environment variable to selectively hide AI providers in the UI when using the all-in-one Docker container. The filtering is based on installed dependencies, so with the full container, all providers are shown by default. There is no documented environment variable for hiding providers at runtime in this scenario. Would you like more detail on possible workarounds or custom container builds? Sources: - <https://github.com/Arize-ai/phoenix/blob/main/docs/prompt-engineering/how-to-prompts/configure-ai-providers.md|https://github.com/Arize-ai/phoenix/blob/main/docs/prompt-engineering/how-to-prompts/configure-ai-providers.md>
💡 Hint: Mention RunLLM in the thread for followups.
Brandon B.
·
RunLLM that's okay. How about changing the default model list? Is that possible?
RunLLM
APP
·
Yes, you can change the default model list in Phoenix. The platform allows you to save and load model configurations per provider, and these defaults are persisted (e.g., in local storage) and used when switching providers in the playground. You can add custom model names, and the UI supports entering arbitrary model names, not just those in the predefined list. Would you like more detail on how to configure or persist these defaults? Sources: - <https://github.com/Arize-ai/phoenix/issues/5069|save off model config by provider and use as default in playground> - <https://github.com/Arize-ai/phoenix/pull/5645|feat(playground): support arbitrary string model names> - <https://github.com/Arize-ai/phoenix/pull/5426|fix(playgroun): use saved model config for default model when switching providers>
💡 Hint: Mention RunLLM in the thread for followups.