langroid 0.52.1__tar.gz → 0.52.3__tar.gz

{langroid-0.52.1 → langroid-0.52.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langroid
-Version: 0.52.1
+Version: 0.52.3
 Summary: Harness LLMs with Multi-Agent Programming
 Author-email: Prasad Chalasani <pchalasani@gmail.com>
 License: MIT

{langroid-0.52.1 → langroid-0.52.3}/langroid/parsing/file_attachment.py

@@ -3,6 +3,7 @@ import mimetypes
 import uuid
 from pathlib import Path
 from typing import Any, BinaryIO, Dict, Optional, Union
+from urllib.parse import urlparse
 
 from langroid.pydantic_v1 import BaseModel
 
@@ -13,6 +14,8 @@ class FileAttachment(BaseModel):
     content: bytes
     filename: Optional[str] = None
     mime_type: str = "application/octet-stream"
+    url: str | None = None
+    detail: str | None = None
 
     def __init__(self, **data: Any) -> None:
         """Initialize with sensible defaults for filename if not provided."""
@@ -23,7 +26,11 @@ class FileAttachment(BaseModel):
         super().__init__(**data)
 
     @classmethod
-    def from_path(cls, file_path: Union[str, Path]) -> "FileAttachment":
+    def _from_path(
+        cls,
+        file_path: Union[str, Path],
+        detail: Optional[str] = None,
+    ) -> "FileAttachment":
         """Create a FileAttachment from a file path.
 
         Args:
@@ -40,7 +47,74 @@ class FileAttachment(BaseModel):
         if mime_type is None:
             mime_type = "application/octet-stream"
 
-        return cls(content=content, filename=path.name, mime_type=mime_type)
+        return cls(
+            content=content,
+            filename=path.name,
+            mime_type=mime_type,
+            detail=detail,
+        )
+
+    @classmethod
+    def _from_url(
+        cls,
+        url: str,
+        content: Optional[bytes] = None,
+        filename: Optional[str] = None,
+        mime_type: Optional[str] = None,
+        detail: Optional[str] = None,
+    ) -> "FileAttachment":
+        """Create a FileAttachment from a URL.
+
+        Args:
+            url: URL to the file
+            content: Optional raw bytes content (if already fetched)
+            filename: Optional name to use for the file
+            mime_type: MIME type of the content, guessed from filename or url
+
+        Returns:
+            FileAttachment instance
+        """
+        if filename is None and url:
+            # Extract filename from URL if possible
+
+            parsed_url = urlparse(url)
+            path = parsed_url.path
+            filename = path.split("/")[-1] if path else None
+
+        if mime_type is None and filename:
+            mime_type, _ = mimetypes.guess_type(filename)
+
+        return cls(
+            content=content or b"",  # Empty bytes if no content provided
+            filename=filename,
+            mime_type=mime_type or "application/octet-stream",
+            url=url,
+            detail=detail,
+        )
+
+    @classmethod
+    def from_path(
+        cls,
+        path: Union[str, Path],
+        detail: str | None = None,
+    ) -> "FileAttachment":
+        """Create a FileAttachment from either a local file path or a URL.
+
+        Args:
+            path_or_url: Path to the file or URL to fetch
+
+        Returns:
+            FileAttachment instance
+        """
+        # Convert to string if Path object
+        path_str = str(path)
+
+        # Check if it's a URL
+        if path_str.startswith(("http://", "https://", "ftp://")):
+            return cls._from_url(url=path_str, detail=detail)
+        else:
+            # Assume it's a local file path
+            return cls._from_path(path_str, detail=detail)
 
     @classmethod
     def from_bytes(
@@ -135,19 +209,34 @@ class FileAttachment(BaseModel):
         Returns:
             Dictionary with file data
         """
-        if "gemini" in model.lower():
-            return dict(type="image_url", image_url=dict(url=self.to_data_uri()))
-        elif "claude" in model.lower():
-            # optimistically try this: some API proxies like litellm
-            # support this, and others may not.
+        if (
+            self.mime_type
+            and self.mime_type.startswith("image/")
+            or "gemini" in model.lower()
+        ):
+            # for gemini models, we use `image_url` for both pdf-files and images
+
+            image_url_dict = {}
+
+            # If we have a URL and it's a full http/https URL, use it directly
+            if self.url and (
+                self.url.startswith("http://") or self.url.startswith("https://")
+            ):
+                image_url_dict["url"] = self.url
+            # Otherwise use base64 data URI
+            else:
+                image_url_dict["url"] = self.to_data_uri()
+
+            # Add detail parameter if specified
+            if self.detail:
+                image_url_dict["detail"] = self.detail
+
             return dict(
-                type="file",
-                file=dict(
-                    file_data=self.to_data_uri(),
-                ),
+                type="image_url",
+                image_url=image_url_dict,
             )
         else:
-            # fallback: assume file upload is similar to OpenAI API
+            # For non-image files
             return dict(
                 type="file",
                 file=dict(

{langroid-0.52.1 → langroid-0.52.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "langroid"
-version = "0.52.1"
+version = "0.52.3"
 authors = [
     {name = "Prasad Chalasani", email = "pchalasani@gmail.com"},
 ]