langfun 0.1.2.dev202411020804__tar.gz → 0.1.2.dev202411060804__tar.gz

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: langfun
-Version: 0.1.2.dev202411020804
+Version: 0.1.2.dev202411060804
 Summary: Langfun: Language as Functions.
 Home-page: https://github.com/google/langfun
 Author: Langfun Authors

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/component.py

@@ -253,7 +253,9 @@ def _get_scoped_value(
   return scoped_values.get(var_name, default)
 
 
-class ContextualAttribute(pg.symbolic.ValueFromParentChain):
+class ContextualAttribute(
+    pg.symbolic.ValueFromParentChain, pg.views.HtmlTreeView.Extension
+):
   """Attributes whose values are inferred from the context of the component.
 
   Please see go/langfun-component#attribute-value-retrieval for details.
@@ -290,8 +292,8 @@ class ContextualAttribute(pg.symbolic.ValueFromParentChain):
       self,
       *,
       view: pg.views.HtmlTreeView,
-      parent: Any,
-      root_path: pg.KeyPath,
+      parent: Any = None,
+      root_path: pg.KeyPath | None = None,
       **kwargs,
   ) -> pg.Html:
     inferred_value = pg.MISSING_VALUE
@@ -301,7 +303,8 @@ class ContextualAttribute(pg.symbolic.ValueFromParentChain):
     if inferred_value is not pg.MISSING_VALUE:
       kwargs.pop('name', None)
       return view.render(
-          inferred_value, parent=self, root_path=root_path + '<inferred>',
+          inferred_value, parent=self,
+          root_path=pg.KeyPath('<inferred>', root_path),
           **view.get_passthrough_kwargs(**kwargs)
       )
     return pg.Html.element(

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/llms/anthropic.py

@@ -141,6 +141,8 @@ class Anthropic(rest.REST):
         'x-api-key': self._api_key,
         'anthropic-version': self.api_version,
         'content-type': 'application/json',
+        # TODO(yifenglu): Remove beta flag once the feature is fully supported.
+        'anthropic-beta': 'pdfs-2024-09-25',
     }
 
   @property
@@ -223,7 +225,6 @@ class Anthropic(rest.REST):
         if isinstance(chunk, str):
           item = dict(type='text', text=chunk)
         elif isinstance(chunk, lf_modalities.Image):
-          # NOTE(daiyip): Anthropic only support image content instead of URL.
           item = dict(
               type='image',
               source=dict(
@@ -232,6 +233,15 @@ class Anthropic(rest.REST):
                   data=base64.b64encode(chunk.to_bytes()).decode(),
               ),
           )
+        elif isinstance(chunk, lf_modalities.PDF):
+          item = dict(
+              type='document',
+              source=dict(
+                  type='base64',
+                  media_type=chunk.mime_type,
+                  data=base64.b64encode(chunk.to_bytes()).decode(),
+              ),
+          )
         else:
           raise ValueError(f'Unsupported modality object: {chunk!r}.')
         content.append(item)

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/llms/anthropic_test.py

@@ -59,18 +59,30 @@ image_content = (
     b'\xdao\xd0|\x00\x00\x00\x00IEND\xaeB`\x82'
 )
 
+pdf_content = (
+    b'%PDF-1.4\n1 0 obj\n<< /Type /Catalog /Pages 2 0 R >>\nendobj\n2 0 obj\n<<'
+    b' /Type /Pages /Count 1 /Kids [3 0 R] >>\nendobj\n3 0 obj\n<< /Type /Page'
+    b' /Parent 2 0 R /MediaBox [0 0 612 792] /Contents 4 0 R >>\nendobj\n4 0'
+    b' obj\n<< /Length 44 >>\nstream\nBT /F1 24 Tf 100 700 Td (Hello, PDF'
+    b' content!) Tj ET\nendstream\nendobj\n5 0 obj\n<< /Type /Font /Subtype'
+    b' /Type1 /BaseFont /Helvetica >>\nendobj\nxref\n0 6\n0000000000 65535 f'
+    b' \n0000000010 00000 n \n0000000079 00000 n \n0000000178 00000 n'
+    b' \n0000000278 00000 n \n0000000407 00000 n \ntrailer\n<< /Size 6 /Root 1'
+    b' 0 R >>\nstartxref\n517\n%%EOF'
+)
+
 
 def mock_mm_requests_post(url: str, json: dict[str, Any], **kwargs):
   del url, kwargs
   v = json['messages'][0]['content'][0]
-  image = lf_modalities.Image.from_bytes(base64.b64decode(v['source']['data']))
+  content = lf_modalities.Mime.from_bytes(base64.b64decode(v['source']['data']))
 
   response = requests.Response()
   response.status_code = 200
   response._content = pg.to_json_str({
       'content': [{
           'type': 'text',
-          'text': f'{v["type"]}: {image.mime_type}',
+          'text': f'{v["type"]}: {content.mime_type}',
       }],
       'usage': {
           'input_tokens': 2,
@@ -146,6 +158,13 @@ class AnthropicTest(unittest.TestCase):
       response = lm(lf_modalities.Image.from_bytes(image_content), lm=lm)
       self.assertEqual(response.text, 'image: image/png')
 
+  def test_pdf_call(self):
+    with mock.patch('requests.Session.post') as mock_mm_request:
+      mock_mm_request.side_effect = mock_mm_requests_post
+      lm = anthropic.Claude3Haiku(api_key='fake_key')
+      response = lm(lf_modalities.PDF.from_bytes(pdf_content), lm=lm)
+      self.assertEqual(response.text, 'document: application/pdf')
+
   def test_call_errors(self):
     for status_code, error_type, error_message in [
         (429, 'rate_limit', 'Rate limit exceeded.'),

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/logging.py

@@ -43,7 +43,7 @@ def get_log_level() -> LogLevel:
   return component.context_value('__event_log_level__', 'info')
 
 
-class LogEntry(pg.Object):
+class LogEntry(pg.Object, pg.views.HtmlTreeView.Extension):
   """Event log entry."""
   time: datetime.datetime
   level: LogLevel
@@ -60,7 +60,7 @@ class LogEntry(pg.Object):
       title: str | pg.Html | None = None,
       max_summary_len_for_str: int = 80,
       **kwargs
-      ) -> str:
+  ) -> pg.Html | None:
     if len(self.message) > max_summary_len_for_str:
       message = self.message[:max_summary_len_for_str] + '...'
     else:
@@ -89,7 +89,7 @@ class LogEntry(pg.Object):
   def _html_tree_view_content(
       self,
       view: pg.views.HtmlTreeView,
-      root_path: pg.KeyPath,
+      root_path: pg.KeyPath | None = None,
       max_summary_len_for_str: int = 80,
       collapse_level: int | None = 1,
       extra_flags: dict[str, Any] | None = None,
@@ -118,7 +118,7 @@ class LogEntry(pg.Object):
               view.render(
                   self.metadata,
                   name='metadata',
-                  root_path=root_path + 'metadata',
+                  root_path=pg.KeyPath('metadata', root_path),
                   parent=self,
                   collapse_level=view.get_collapse_level(
                       (collapse_level, -1), collapse_log_metadata_level,

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/message.py

@@ -23,7 +23,11 @@ from langfun.core import natural_language
 import pyglove as pg
 
 
-class Message(natural_language.NaturalLanguageFormattable, pg.Object):
+class Message(
+    natural_language.NaturalLanguageFormattable,
+    pg.Object,
+    pg.views.HtmlTreeView.Extension
+):
   """Message.
 
   ``Message`` is the protocol for users and the system to interact with
@@ -511,7 +515,7 @@ class Message(natural_language.NaturalLanguageFormattable, pg.Object):
       self,
       *,
       view: pg.views.HtmlTreeView,
-      root_path: pg.KeyPath,
+      root_path: pg.KeyPath | None = None,
       collapse_level: int | None = None,
       extra_flags: dict[str, Any] | None = None,
       **kwargs,
@@ -582,7 +586,7 @@ class Message(natural_language.NaturalLanguageFormattable, pg.Object):
           s.write(s.escape(chunk))
         else:
           assert isinstance(chunk, modality.Modality), chunk
-          child_path = root_path + 'metadata' + chunk.referred_name
+          child_path = pg.KeyPath(['metadata', chunk.referred_name], root_path)
           s.write(
               pg.Html.element(
                   'div',
@@ -610,7 +614,7 @@ class Message(natural_language.NaturalLanguageFormattable, pg.Object):
     def render_result():
       if 'result' not in self.metadata:
         return None
-      child_path = root_path + 'metadata' + 'result'
+      child_path = pg.KeyPath(['metadata', 'result'], root_path)
       return pg.Html.element(
           'div',
           [
@@ -632,7 +636,7 @@ class Message(natural_language.NaturalLanguageFormattable, pg.Object):
     def render_usage():
       if 'usage' not in self.metadata:
         return None
-      child_path = root_path + 'metadata' + 'usage'
+      child_path = pg.KeyPath(['metadata', 'usage'], root_path)
       return pg.Html.element(
           'div',
           [
@@ -661,7 +665,7 @@ class Message(natural_language.NaturalLanguageFormattable, pg.Object):
              and not source.has_tag(source_tag)):
         source = source.source
       if source is not None:
-        child_path = root_path + 'source'
+        child_path = pg.KeyPath('source', root_path)
         child_extra_flags = extra_flags.copy()
         child_extra_flags['collapse_source_message_level'] = (
             view.get_collapse_level(
@@ -684,7 +688,7 @@ class Message(natural_language.NaturalLanguageFormattable, pg.Object):
     def render_metadata():
       if not include_message_metadata:
         return None
-      child_path = root_path + 'metadata'
+      child_path = pg.KeyPath('metadata', root_path)
       return pg.Html.element(
           'div',
           [

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/modality.py

@@ -31,7 +31,7 @@ def format_modality_as_ref(enabled: bool = True) -> ContextManager[None]:
   )
 
 
-class Modality(component.Component):
+class Modality(component.Component, pg.views.HtmlTreeView.Extension):
   """Base class for multimodal object."""
 
   REF_START = '<<[['

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/structured/mapping.py

@@ -59,7 +59,9 @@ class MappingError(Exception):  # pylint: disable=g-bad-exception-name
 
 
 @pg.use_init_args(['input', 'output', 'schema', 'context'])
-class MappingExample(lf.NaturalLanguageFormattable, lf.Component):
+class MappingExample(lf.NaturalLanguageFormattable,
+                     lf.Component,
+                     pg.views.HtmlTreeView.Extension):
   """Mapping example between text, schema and structured value."""
 
   input: pg.typing.Annotated[

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/structured/schema.py

@@ -110,7 +110,11 @@ class SchemaError(Exception):   # pylint: disable=g-bad-exception-name
     return r.getvalue()
 
 
-class Schema(lf.NaturalLanguageFormattable, pg.Object):
+class Schema(
+    lf.NaturalLanguageFormattable,
+    pg.Object,
+    pg.views.HtmlTreeView.Extension
+):
   """Base class for structured data schema."""
 
   spec: pg.typing.Annotated[
@@ -193,7 +197,6 @@ class Schema(lf.NaturalLanguageFormattable, pg.Object):
       self,
       *,
       view: pg.views.HtmlTreeView,
-      root_path: pg.KeyPath,
       **kwargs,
   ):
     return pg.Html.element(

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun/core/template.py

@@ -47,6 +47,7 @@ class Template(
     natural_language.NaturalLanguageFormattable,
     component.Component,
     pg.typing.CustomTyping,
+    pg.views.HtmlTreeView.Extension
 ):
   """Langfun string template.
 
@@ -530,7 +531,7 @@ class Template(
       self,
       *,
       view: pg.views.HtmlTreeView,
-      root_path: pg.KeyPath,
+      root_path: pg.KeyPath | None = None,
       collapse_level: int | None = None,
       extra_flags: dict[str, Any] | None = None,
       debug: bool = False,

{langfun-0.1.2.dev202411020804 → langfun-0.1.2.dev202411060804}/langfun.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: langfun
-Version: 0.1.2.dev202411020804
+Version: 0.1.2.dev202411060804
 Summary: Langfun: Language as Functions.
 Home-page: https://github.com/google/langfun
 Author: Langfun Authors