langchain-b12 0.1.0__py3-none-any.whl → 0.1.2__py3-none-any.whl

langchain_b12/genai/genai.py

@@ -145,13 +145,14 @@ class ChatGenAI(BaseChatModel):
         self, messages: list[BaseMessage]
     ) -> tuple[str | None, types.ContentListUnion]:
         contents = convert_messages_to_contents(messages)
-        if isinstance(messages[-1], SystemMessage):
-            system_instruction = messages[-1].content
-            assert isinstance(
-                system_instruction, str
-            ), "System message content must be a string"
-        else:
-            system_instruction = None
+        system_message: SystemMessage | None = next(
+            (message for message in messages if isinstance(message, SystemMessage)),
+            None,
+        )
+        system_instruction = system_message.content if system_message else None
+        assert system_instruction is None or isinstance(
+            system_instruction, str
+        ), "System message content must be a string or None"
         return system_instruction, cast(types.ContentListUnion, contents)
 
     def get_num_tokens(self, text: str) -> int:

langchain_b12/genai/genai_utils.py

@@ -24,22 +24,37 @@ def multi_content_to_part(
     Args:
         contents: A sequence of dictionaries representing content. Examples:
             [
-                {
+                { # Text content
                     "type": "text",
                     "text": "This is a text message"
                 },
-                {
+                { # Image content from base64 encoded string with OpenAI format
                     "type": "image_url",
                     "image_url": {
                         "url": f"data:{mime_type};base64,{encoded_artifact}"
                     },
                 },
-                {
+                { # Image content fro base64 encoded string with LangChain format
+                    "type": "image",
+                    "source_type": "base64",
+                    "data": "<base64 string>",
+                    "mime_type": "image/jpeg",
+                },
+                { # Image content from URL
+                    "type": "image",
+                    "source_type": "url",
+                    "url": "https://...",
+                },
+                { # File content from base64 encoded string
+                    "type": "file",
+                    "source_type": "base64",
+                    "mime_type": "application/pdf",
+                    "data": "<base64 data string>",
+                },
+                { # File content from URL
                     "type": "file",
-                    "file": {
-                        "uri": f"gs://{bucket_name}/{file_name}",
-                        "mime_type": mime_type,
-                    }
+                    "source_type": "url",
+                    "url": "https://...",
                 }
             ]
     """
@@ -60,15 +75,51 @@ def multi_content_to_part(
             mime_type = header.split(":", 1)[1].split(";", 1)[0]
             data = base64.b64decode(encoded_data)
             parts.append(types.Part.from_bytes(data=data, mime_type=mime_type))
+        elif content["type"] == "image":
+            if "data" in content:
+                assert isinstance(content["data"], str), "Expected str data"
+                assert "mime_type" in content, "Expected 'mime_type' in content"
+                assert isinstance(content["mime_type"], str), "Expected str mime_type"
+                data = base64.b64decode(content["data"])
+                parts.append(
+                    types.Part.from_bytes(data=data, mime_type=content["mime_type"])
+                )
+            elif "url" in content:
+                assert isinstance(content["url"], str), "Expected str url"
+                mime_type = content.get("mime_type", None)
+                assert mime_type is None or isinstance(
+                    mime_type, str
+                ), "Expected str mime_type"
+                parts.append(
+                    types.Part.from_uri(file_uri=content["url"], mime_type=mime_type)
+                )
+            else:
+                raise ValueError(
+                    "Expected either 'data' or 'url' in content for image type"
+                )
         elif content["type"] == "file":
-            assert "file" in content, "Expected 'file' in content"
-            file = content["file"]
-            assert isinstance(file, dict), "Expected dict file"
-            assert "uri" in file, "Expected 'uri' in content['file']"
-            assert "mime_type" in file, "Expected 'mime_type' in content['file']"
-            parts.append(
-                types.Part.from_uri(file_uri=file["uri"], mime_type=file["mime_type"])
-            )
+            if "data" in content:
+                assert isinstance(content["data"], str), "Expected str data"
+                assert "mime_type" in content, "Expected 'mime_type' in content"
+                assert isinstance(content["mime_type"], str), "Expected str mime_type"
+                data = base64.b64decode(content["data"])
+                parts.append(
+                    types.Part.from_bytes(data=data, mime_type=content["mime_type"])
+                )
+            elif "url" in content:
+                assert isinstance(content["url"], str), "Expected str url"
+                assert content["url"], "File URI is required"
+                mime_type = content.get("mime_type", None)
+                assert mime_type is None or isinstance(
+                    mime_type, str
+                ), "Expected str mime_type"
+                parts.append(
+                    types.Part.from_uri(file_uri=content["url"], mime_type=mime_type)
+                )
+            else:
+                raise ValueError(
+                    "Expected either 'data' or 'url' in content for file type"
+                )
         else:
             raise ValueError(f"Unknown content type: {content['type']}")
     return parts

{langchain_b12-0.1.0.dist-info → langchain_b12-0.1.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langchain-b12
-Version: 0.1.0
+Version: 0.1.2
 Summary: A reusable collection of tools and implementations for Langchain
 Author-email: Vincent Min <vincent.min@b12-consulting.com>
 Requires-Python: >=3.11
@@ -11,8 +11,6 @@ Description-Content-Type: text/markdown
 
 This repo hosts a collection of custom LangChain components.
 
-- ChatGenAI converts a sequence of LangChain messages to Google GenAI Content objects and vice versa.
-
 ## Installation
 
 To install this package, run

langchain_b12-0.1.2.dist-info/RECORD

@@ -0,0 +1,8 @@
+langchain_b12/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+langchain_b12/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+langchain_b12/genai/embeddings.py,sha256=od2bVIgt7v9aNAHG0PVypVF1H_XgHto2nTd8vwfvyN8,3355
+langchain_b12/genai/genai.py,sha256=gzkgtvs3wNjcslS_KFZYCajUZIsJkVN2Tq2Q1RMIPyc,15910
+langchain_b12/genai/genai_utils.py,sha256=La3FOtRGmzI5IDql3ojaYzz-lJqSgHOn0Wp3KwZ3P4I,10735
+langchain_b12-0.1.2.dist-info/METADATA,sha256=8Uzn2ChQ8scEHLRiMykPTdP58wZEgiAaRXS8f2DjTe4,1204
+langchain_b12-0.1.2.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+langchain_b12-0.1.2.dist-info/RECORD,,

langchain_b12-0.1.0.dist-info/RECORD

@@ -1,8 +0,0 @@
-langchain_b12/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-langchain_b12/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-langchain_b12/genai/embeddings.py,sha256=od2bVIgt7v9aNAHG0PVypVF1H_XgHto2nTd8vwfvyN8,3355
-langchain_b12/genai/genai.py,sha256=o2KLo2QlLDy0hijt6T2HaRmjSO60SV62dR_Cso6Ad-8,15796
-langchain_b12/genai/genai_utils.py,sha256=2ojnSumovXyx3CxM7JwzyOkEdD7mQxfLnk50-usPbw8,8221
-langchain_b12-0.1.0.dist-info/METADATA,sha256=72Ct-UG2KHnPQuyUltOQ_HGpvLbDrXglmPn-eY9blnc,1307
-langchain_b12-0.1.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-langchain_b12-0.1.0.dist-info/RECORD,,