Support inline image rendering for LangChain tool results

[Teja-web]

Problem

Tool results can contain multimodal content blocks, including base64-encoded images. The current UI renders these results as raw JSON/text, exposing the base64 string instead of displaying the image.

LangChain’s Python and JavaScript implementations also use different image field names:

• Python: base64, mime_type
• JavaScript: data, mimeType

Expected behavior

When a tool result’s content is an array containing a block with type: "image":

• Render the image inline using a data:<mime-type>;base64,<data> URL.
• Support both Python and JavaScript field conventions.
• Render accompanying type: "text" blocks above the image.
• Do not display the raw base64 data.
• Make the image responsive, width-constrained, rounded, and accessible with descriptive alt text.
• Preserve the existing renderer for JSON and ordinary text results.

Example payloads

Python LangChain

[
  {
    "type": "text",
    "text": "Generated chart"
  },
  {
    "type": "image",
    "base64": "<base64-data>",
    "mime_type": "image/png"
  }
]