diff --git a/.github/screenshots/google-adk-trace-detail-messages.png b/.github/screenshots/google-adk-trace-detail-messages.png
new file mode 100644
index 00000000000..ebc34fe64a6
Binary files /dev/null and b/.github/screenshots/google-adk-trace-detail-messages.png differ
diff --git a/.github/screenshots/google-adk-trace-full-page.png b/.github/screenshots/google-adk-trace-full-page.png
new file mode 100644
index 00000000000..05299c06484
Binary files /dev/null and b/.github/screenshots/google-adk-trace-full-page.png differ
diff --git a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/detector.ts b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/detector.ts
new file mode 100644
index 00000000000..eb14a5ca153
--- /dev/null
+++ b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/detector.ts
@@ -0,0 +1,86 @@
+import { FormatDetector } from "../../types";
+
+interface GooglePart {
+  text?: string;
+  inline_data?: {
+    data?: string;
+    mime_type?: string;
+  };
+  function_call?: unknown;
+  function_response?: unknown;
+}
+
+interface GoogleContent {
+  role?: string;
+  parts?: GooglePart[];
+}
+
+interface GoogleCandidate {
+  content?: GoogleContent;
+  finish_reason?: string;
+}
+
+/**
+ * Checks if an object looks like a Google GenAI Content item
+ */
+const isGoogleContent = (item: unknown): item is GoogleContent => {
+  if (!item || typeof item !== "object") return false;
+  const c = item as Record<string, unknown>;
+
+  if (!Array.isArray(c.parts)) return false;
+
+  const validRoles = ["user", "model", "function", "system"];
+  if (c.role !== undefined && !validRoles.includes(c.role as string))
+    return false;
+
+  return true;
+};
+
+/**
+ * Detects Google GenAI input format:
+ * { contents: [{ role: "user", parts: [{ text: "..." } | { inline_data: {...} }] }] }
+ */
+const hasGoogleInputFormat = (data: unknown): boolean => {
+  if (!data || typeof data !== "object") return false;
+  const d = data as Record<string, unknown>;
+
+  if (!Array.isArray(d.contents)) return false;
+  if (d.contents.length === 0) return false;
+
+  return d.contents.every(isGoogleContent);
+};
+
+/**
+ * Detects Google GenAI output format:
+ * { candidates: [{ content: { role: "model", parts: [...] } }] }
+ */
+const hasGoogleOutputFormat = (data: unknown): boolean => {
+  if (!data || typeof data !== "object") return false;
+  const d = data as Record<string, unknown>;
+
+  if (!Array.isArray(d.candidates)) return false;
+  if (d.candidates.length === 0) return false;
+
+  return d.candidates.every((candidate: unknown) => {
+    if (!candidate || typeof candidate !== "object") return false;
+    const c = candidate as GoogleCandidate;
+    return c.content === undefined || isGoogleContent(c.content);
+  });
+};
+
+/**
+ * Detects if the provided data is in Google GenAI (ADK) format.
+ */
+export const detectGoogleFormat: FormatDetector = (data, prettifyConfig) => {
+  if (!data) return false;
+
+  const isInput = prettifyConfig?.fieldType === "input";
+  const isOutput = prettifyConfig?.fieldType === "output";
+
+  if (!isInput && !isOutput) return false;
+
+  if (isInput && hasGoogleInputFormat(data)) return true;
+  if (isOutput && hasGoogleOutputFormat(data)) return true;
+
+  return false;
+};
diff --git a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/index.ts b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/index.ts
new file mode 100644
index 00000000000..2fea3eedccc
--- /dev/null
+++ b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/index.ts
@@ -0,0 +1,11 @@
+import { LLMMessageFormatImplementation } from "../../types";
+import { detectGoogleFormat } from "./detector";
+import { mapGoogleMessages } from "./mapper";
+
+export const googleFormat: LLMMessageFormatImplementation = {
+  name: "google",
+  detector: detectGoogleFormat,
+  mapper: mapGoogleMessages,
+};
+
+export { detectGoogleFormat, mapGoogleMessages };
diff --git a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/mapper.ts b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/mapper.ts
new file mode 100644
index 00000000000..aec26b3efb7
--- /dev/null
+++ b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/google/mapper.ts
@@ -0,0 +1,210 @@
+import PrettyLLMMessage from "@/shared/PrettyLLMMessage";
+import {
+  FormatMapper,
+  LLMMessageDescriptor,
+  LLMBlockDescriptor,
+  LLMMapperResult,
+} from "../../types";
+import { MessageRole } from "@/shared/PrettyLLMMessage/types";
+import { isPlaceholder } from "../../utils";
+
+interface GoogleInlineData {
+  data?: string;
+  mime_type?: string;
+}
+
+interface GooglePart {
+  text?: string;
+  inline_data?: GoogleInlineData;
+  function_call?: {
+    name?: string;
+    args?: Record<string, unknown>;
+  };
+  function_response?: {
+    name?: string;
+    response?: unknown;
+  };
+}
+
+interface GoogleContent {
+  role?: string;
+  parts?: GooglePart[];
+}
+
+interface GoogleCandidate {
+  content?: GoogleContent;
+  finish_reason?: string;
+}
+
+interface GoogleInputData {
+  contents: GoogleContent[];
+}
+
+interface GoogleOutputData {
+  candidates: GoogleCandidate[];
+  usage_metadata?: {
+    prompt_token_count?: number;
+    candidates_token_count?: number;
+    total_token_count?: number;
+  };
+}
+
+const generateMessageId = (index: number, prefix: string): string =>
+  `${prefix}-${index}`;
+
+/**
+ * Normalizes Google role names to our internal MessageRole type.
+ */
+const normalizeRole = (role: string | undefined): MessageRole => {
+  if (role === "model") return "assistant";
+  if (role === "function") return "tool";
+  return (role as MessageRole) || "user";
+};
+
+/**
+ * Maps an array of Google parts to block descriptors.
+ */
+const mapParts = (
+  parts: GooglePart[],
+  role: MessageRole,
+): LLMBlockDescriptor[] => {
+  const blocks: LLMBlockDescriptor[] = [];
+  const images: Array<{ url: string; name: string }> = [];
+
+  parts.forEach((part, index) => {
+    if (part.text !== undefined) {
+      // Flush pending images before text
+      if (images.length > 0) {
+        blocks.push({
+          blockType: "image",
+          component: PrettyLLMMessage.ImageBlock,
+          props: { images: [...images] },
+        });
+        images.length = 0;
+      }
+      blocks.push({
+        blockType: "text",
+        component: PrettyLLMMessage.TextBlock,
+        props: {
+          children: part.text,
+          role,
+          showMoreButton: true,
+        },
+      });
+    } else if (part.inline_data) {
+      const data = part.inline_data.data;
+      if (data && data.length > 0) {
+        images.push({
+          url: data,
+          name: isPlaceholder(data) ? data : `Image ${index + 1}`,
+        });
+      }
+    } else if (part.function_call) {
+      const name = part.function_call.name || "function_call";
+      const args = part.function_call.args
+        ? JSON.stringify(part.function_call.args, null, 2)
+        : "";
+      blocks.push({
+        blockType: "code",
+        component: PrettyLLMMessage.CodeBlock,
+        props: { code: args, label: name },
+      });
+    } else if (part.function_response) {
+      const name = part.function_response.name || "function_response";
+      const response = part.function_response.response
+        ? JSON.stringify(part.function_response.response, null, 2)
+        : "";
+      blocks.push({
+        blockType: "code",
+        component: PrettyLLMMessage.CodeBlock,
+        props: { code: response, label: name },
+      });
+    }
+  });
+
+  // Flush any remaining images
+  if (images.length > 0) {
+    blocks.push({
+      blockType: "image",
+      component: PrettyLLMMessage.ImageBlock,
+      props: { images },
+    });
+  }
+
+  return blocks;
+};
+
+/**
+ * Maps a GoogleContent object to an LLMMessageDescriptor.
+ */
+const mapGoogleContent = (
+  content: GoogleContent,
+  index: number,
+  prefix: string,
+): LLMMessageDescriptor => {
+  const role = normalizeRole(content.role);
+  const blocks = content.parts ? mapParts(content.parts, role) : [];
+
+  return {
+    id: generateMessageId(index, prefix),
+    role,
+    blocks,
+  };
+};
+
+/**
+ * Maps Google GenAI input format to LLMMapperResult.
+ */
+const mapGoogleInput = (data: GoogleInputData): LLMMapperResult => {
+  const messages = data.contents.map((content, index) =>
+    mapGoogleContent(content, index, "input"),
+  );
+  return { messages };
+};
+
+/**
+ * Maps Google GenAI output format to LLMMapperResult.
+ */
+const mapGoogleOutput = (data: GoogleOutputData): LLMMapperResult => {
+  const messages: LLMMessageDescriptor[] = [];
+
+  data.candidates.forEach((candidate, index) => {
+    if (!candidate.content) return;
+
+    const message = mapGoogleContent(candidate.content, index, "output");
+    if (candidate.finish_reason) {
+      message.finishReason = candidate.finish_reason;
+    }
+    messages.push(message);
+  });
+
+  const usage = data.usage_metadata
+    ? {
+        prompt_tokens: data.usage_metadata.prompt_token_count,
+        completion_tokens: data.usage_metadata.candidates_token_count,
+        total_tokens: data.usage_metadata.total_token_count,
+      }
+    : undefined;
+
+  return { messages, usage };
+};
+
+/**
+ * Maps Google GenAI format data to normalized LLMMapperResult.
+ */
+export const mapGoogleMessages: FormatMapper = (data, prettifyConfig) => {
+  if (!data) return { messages: [] };
+
+  const isInput = prettifyConfig?.fieldType === "input";
+  const isOutput = prettifyConfig?.fieldType === "output";
+
+  if (isInput && typeof data === "object" && "contents" in data) {
+    return mapGoogleInput(data as GoogleInputData);
+  }
+
+  if (isOutput && typeof data === "object" && "candidates" in data) {
+    return mapGoogleOutput(data as GoogleOutputData);
+  }
+
+  return { messages: [] };
+};
diff --git a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/index.ts b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/index.ts
index f80ceb3e260..9f10f23f003 100644
--- a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/index.ts
+++ b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/index.ts
@@ -1,2 +1,3 @@
 export { openaiFormat, detectOpenAIFormat, mapOpenAIMessages } from "./openai";
+export { googleFormat, detectGoogleFormat, mapGoogleMessages } from "./google";
 export { getFormat, getAllFormats } from "./registry";
diff --git a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/registry.ts b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/registry.ts
index a47c293ff9b..7e75a9edb8e 100644
--- a/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/registry.ts
+++ b/apps/opik-frontend/src/shared/PrettyLLMMessage/llmMessages/providers/registry.ts
@@ -1,6 +1,7 @@
 import { LLMMessageFormat, LLMMessageFormatImplementation } from "../types";
 import { openaiFormat } from "./openai";
 import { langchainFormat } from "./langchain";
+import { googleFormat } from "./google";
 
 const FORMAT_REGISTRY: Record<
   LLMMessageFormat,
@@ -9,7 +10,7 @@ const FORMAT_REGISTRY: Record<
   openai: openaiFormat,
   langchain: langchainFormat,
   anthropic: null,
-  google: null,
+  google: googleFormat,
 };
 
 export const getFormat = (
diff --git a/apps/opik-frontend/src/v2/pages-shared/llm/PromptMessagesReadonly/PromptMessagesReadonly.tsx b/apps/opik-frontend/src/v2/pages-shared/llm/PromptMessagesReadonly/PromptMessagesReadonly.tsx
index 944a56d2532..e1d0aa631b5 100644
--- a/apps/opik-frontend/src/v2/pages-shared/llm/PromptMessagesReadonly/PromptMessagesReadonly.tsx
+++ b/apps/opik-frontend/src/v2/pages-shared/llm/PromptMessagesReadonly/PromptMessagesReadonly.tsx
@@ -25,6 +25,7 @@ type MediaItem = {
   video_url?: { url: string };
   audio_url?: { url: string };
   input_audio?: { data: string };
+  inline_data?: { data: string; mime_type?: string };
 };
 
 const getRoleLabel = (role: string): string => {
@@ -76,6 +77,12 @@ const getTextAndMedia = (
       if (item.type === "audio_url" || item.type === "input_audio") {
         const url = item.audio_url?.url || item.url;
         if (url) audios.push(url);
+        return;
+      }
+
+      // Google GenAI inline_data (images via ADK)
+      if (item.inline_data?.data) {
+        images.push(item.inline_data.data);
       }
     });
 
diff --git a/sdks/python/src/opik/api_objects/attachment/attachments_extractor.py b/sdks/python/src/opik/api_objects/attachment/attachments_extractor.py
index 6e590418574..7f4bac1cbf6 100644
--- a/sdks/python/src/opik/api_objects/attachment/attachments_extractor.py
+++ b/sdks/python/src/opik/api_objects/attachment/attachments_extractor.py
@@ -33,12 +33,13 @@ def __init__(self, min_attachment_size: int):
         self.decoder = decoder_base64.Base64AttachmentDecoder()
 
         # Pattern to match base64 strings (can be embedded in text)
+        # Matches both standard base64 (+/) and URL-safe base64 (-_)
         # Requires at least min_attachment_size characters to reduce false positives
         min_base64_groups = int(min_attachment_size / 4)
         BASE64_PATTERN = (
-            r"(?:[A-Za-z0-9+/]{4}){"
+            r"(?:[A-Za-z0-9+/\-_]{4}){"
             + str(min_base64_groups)
-            + ",}(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?"
+            + r",}(?:[A-Za-z0-9+/\-_]{2}==|[A-Za-z0-9+/\-_]{3}=)?"
         )
         self.pattern = re.compile(BASE64_PATTERN)
 
diff --git a/sdks/python/src/opik/api_objects/attachment/decoder_base64.py b/sdks/python/src/opik/api_objects/attachment/decoder_base64.py
index 21cb39a162c..740fdddca0b 100644
--- a/sdks/python/src/opik/api_objects/attachment/decoder_base64.py
+++ b/sdks/python/src/opik/api_objects/attachment/decoder_base64.py
@@ -35,8 +35,12 @@ def decode(
             return None
 
         try:
-            # Decode base64 string to bytes
-            decoded_bytes = base64.b64decode(raw_data, validate=True)
+            # Decode base64 string to bytes.
+            # Try standard base64 first; fall back to URL-safe base64 (- and _ chars).
+            try:
+                decoded_bytes = base64.b64decode(raw_data, validate=True)
+            except (ValueError, binascii.Error):
+                decoded_bytes = base64.urlsafe_b64decode(raw_data + "==")
 
             # Detect MIME type from content
             mime_type = decoder_helpers.detect_mime_type(decoded_bytes)