Arize Phoenix Prompt Playground: File Attachment Support?

Andrew S. · 2025-08-29T16:08:15.006Z

does the arize phoenix prompt playground support working with file attachments for models like gemini? I don’t see anywhere to add a file attachment like I can do in aistudio

Arize Phoenix Prompt Playground: File Attachment Support? | Arize AI Community

6 comments

· Sorted by Oldest

RunLLM
APP
·
Phoenix's Prompt Playground does not currently support direct file uploads (e.g., images, PDFs) via the UI for prompts, unlike Google AI Studio. However, you can work with multimodal inputs (such as images or audio) by encoding files as base64 and programmatically constructing prompts using the Python SDK or API, especially for Gemini models. This approach is documented for code-based workflows, but not exposed as a UI feature in the playground yet. File upload support in the UI is a requested and planned enhancement, but not available as of now [source 1] [source 2] [source 3].Would you like more detail on how to use file attachments programmatically with Gemini in Phoenix?Sources: - https://github.com/Arize-ai/phoenix/issues/8814 - https://github.com/Arize-ai/phoenix/issues/7528 - https://arize.com/docs/phoenix/integrations/llm-providers/google-gen-ai/google-gen-ai-evals#«R4l38ut1letrnb»
💡 Hint: Mention RunLLM in the thread for followups.
Andrew S.
·
yes, how would I create a dataset with PDFs? is it possible to store the entire PDF as a binary blob in a dataset? RunLLM
RunLLM
APP
·
Phoenix does not currently support storing entire PDFs as binary blobs directly in a dataset. The recommended workaround is to store a reference (such as a URL or file_id) to the PDF in your dataset and handle the file retrieval and processing programmatically during experiment runs. Native blob storage for datasets is a requested feature but not yet available [source 1] [source 2].Would you like more detail on how to implement this workaround or manage PDF references in your datasets?Sources: - https://github.com/Arize-ai/phoenix/issues/7528 - https://github.com/Arize-ai/phoenix/issues/9189
💡 Hint: Mention RunLLM in the thread for followups.
Mikyo
·
Hey Andrew S. - actively scoping this one out - multi-part content is top of mind. Will keep y'all in the loop as we go. Let us know if you have any thoughts or feedback on how you'd like the files stored or referenced.
❤️1
Mikyo
·
Feel free to add to some of the tickets with your thoughts. Would appreciate it https://github.com/Arize-ai/phoenix/issues/9189
Andrew S.
·
👌

RunLLM
APP
·
Phoenix's Prompt Playground does not currently support direct file uploads (e.g., images, PDFs) via the UI for prompts, unlike Google AI Studio. However, you can work with multimodal inputs (such as images or audio) by encoding files as base64 and programmatically constructing prompts using the Python SDK or API, especially for Gemini models. This approach is documented for code-based workflows, but not exposed as a UI feature in the playground yet. File upload support in the UI is a requested and planned enhancement, but not available as of now [source 1] [source 2] [source 3].Would you like more detail on how to use file attachments programmatically with Gemini in Phoenix?Sources: - https://github.com/Arize-ai/phoenix/issues/8814 - https://github.com/Arize-ai/phoenix/issues/7528 - https://arize.com/docs/phoenix/integrations/llm-providers/google-gen-ai/google-gen-ai-evals#«R4l38ut1letrnb»
💡 Hint: Mention RunLLM in the thread for followups.
Andrew S.
·
yes, how would I create a dataset with PDFs? is it possible to store the entire PDF as a binary blob in a dataset? RunLLM
RunLLM
APP
·
Phoenix does not currently support storing entire PDFs as binary blobs directly in a dataset. The recommended workaround is to store a reference (such as a URL or file_id) to the PDF in your dataset and handle the file retrieval and processing programmatically during experiment runs. Native blob storage for datasets is a requested feature but not yet available [source 1] [source 2].Would you like more detail on how to implement this workaround or manage PDF references in your datasets?Sources: - https://github.com/Arize-ai/phoenix/issues/7528 - https://github.com/Arize-ai/phoenix/issues/9189
💡 Hint: Mention RunLLM in the thread for followups.
Mikyo
·
Hey Andrew S. - actively scoping this one out - multi-part content is top of mind. Will keep y'all in the loop as we go. Let us know if you have any thoughts or feedback on how you'd like the files stored or referenced.
❤️1
Mikyo
·
Feel free to add to some of the tickets with your thoughts. Would appreciate it https://github.com/Arize-ai/phoenix/issues/9189
Andrew S.
·
👌