Removing a few chunk_size/overlap references missed in #1160

src/paperqa/contrib/openreview_paper_helper.py

@@ -89,12 +89,11 @@ async def fetch_relevant_papers(self, question: str) -> dict[str, Any]:
         submissions = self.get_submissions()
         submission_string = self.create_submission_string(submissions)
 
-        if len(submission_string) > self.settings.parsing.chunk_size:
+        chunk_size = self.settings.parsing.reader_config["chunk_chars"]
+        if len(submission_string) > chunk_size:
             chunks = [
-                submission_string[i : i + self.settings.parsing.chunk_size]
-                for i in range(
-                    0, len(submission_string), self.settings.parsing.chunk_size
-                )
+                submission_string[i : i + chunk_size]
+                for i in range(0, len(submission_string), chunk_size)
             ]
         else:
             chunks = [submission_string]

src/paperqa/settings.py

@@ -243,7 +243,7 @@ class ParsingSettings(BaseModel):
         default=250, description="Number of characters to overlap chunks."
     )
     reader_config: dict[str, Any] = Field(
-        default_factory=dict,
+        default_factory=lambda: {"chunk_chars": 5000, "overlap": 250},
         description="Optional keyword arguments for the document reader.",
         examples=[{"dpi": 300}],
     )
@@ -349,6 +349,9 @@ class ParsingSettings(BaseModel):
 
     @model_validator(mode="after")
     def _deprecated_field(self) -> Self:
+        default_reader_config = (
+            type(self).model_fields["reader_config"].default_factory()  # type: ignore[call-arg,misc]
+        )
         if (
             self.pdfs_use_block_parsing
             != type(self).model_fields["pdfs_use_block_parsing"].default
@@ -370,7 +373,9 @@ def _deprecated_field(self) -> Self:
                 category=DeprecationWarning,
                 stacklevel=2,
             )
-            if "chunk_chars" not in self.reader_config:
+            if "chunk_chars" not in self.reader_config or self.reader_config[
+                "chunk_chars"
+            ] == default_reader_config.get("chunk_chars"):
                 self.reader_config["chunk_chars"] = self.chunk_size
         if self.overlap != type(self).model_fields["overlap"].default:
             warnings.warn(
@@ -380,7 +385,9 @@ def _deprecated_field(self) -> Self:
                 category=DeprecationWarning,
                 stacklevel=2,
             )
-            if "overlap" not in self.reader_config:
+            if "overlap" not in self.reader_config or self.reader_config[
+                "overlap"
+            ] == default_reader_config.get("overlap"):
                 self.reader_config["overlap"] = self.overlap
         return self
 
@@ -972,8 +979,8 @@ def get_index_name(self) -> str:
             str(self.agent.index.use_absolute_paper_directory),
             self.embedding,
             str(self.parsing.parse_pdf),  # Don't use __name__ as lambda wouldn't differ
-            str(self.parsing.chunk_size),
-            str(self.parsing.overlap),
+            str(self.parsing.reader_config["chunk_chars"]),
+            str(self.parsing.reader_config["overlap"]),
             self.parsing.chunking_algorithm,
             str(self.parsing.multimodal),
         ]

tests/test_agents.py

@@ -529,7 +529,7 @@ async def test_propagate_options(agent_test_settings: Settings) -> None:
     assert len(result.contexts) >= 2, "Test expects a few contexts"
     # Subtract 2 to allow tolerance for chunks with leading/trailing whitespace
     num_contexts_sufficient_length = sum(
-        len(c.context) >= agent_test_settings.parsing.chunk_size - 2
+        len(c.context) >= agent_test_settings.parsing.reader_config["chunk_chars"] - 2
         for c in result.contexts
     )
     # Check most contexts have the expected length