Refactor get_sources_content to return DataPoints

Author: pamelafox

app/backend/approaches/approach.py

@@ -340,15 +340,17 @@ async def get_sources_content(
         use_semantic_captions: bool,
         download_image_sources: bool,
         user_oid: Optional[str] = None,
-    ) -> tuple[list[str], list[str], list[str]]:
-        """
-        Extracts text and image sources from the search results.
-        If use_semantic_captions is True, it will use the captions from the results.
-        If use_image_sources is True, it will extract image URLs from the results.
+    ) -> DataPoints:
+        """Extract text/image sources & citations from documents.
+
+        Args:
+            results: List of retrieved Document objects.
+            use_semantic_captions: Whether to use semantic captions instead of full content text.
+            download_image_sources: Whether to attempt downloading & base64 encoding referenced images.
+            user_oid: Optional user object id for per-user storage access (ADLS scenarios).
+
         Returns:
-            - A list of text sources (captions or content).
-            - A list of image sources (base64 encoded).
-            - A list of allowed citations for those sources.
+            DataPoints: with text (list[str]), images (list[str - base64 data URI]), citations (list[str]).
         """
 
         def nonewlines(s: str) -> str:
@@ -380,8 +382,9 @@ def nonewlines(s: str) -> str:
                     if url:
                         image_sources.append(url)
                     citations.append(self.get_image_citation(doc.sourcepage or "", img["url"]))
-
-        return text_sources, image_sources, citations
+        if download_image_sources:
+            return DataPoints(text=text_sources, images=image_sources, citations=citations)
+        return DataPoints(text=text_sources, citations=citations)
 
     def get_citation(self, sourcepage: Optional[str]):
         return sourcepage or ""

app/backend/approaches/chatreadretrieveread.py

@@ -339,12 +339,14 @@ async def run_search_approach(
         )
 
         # STEP 3: Generate a contextual and content specific answer using the search results and chat history
-        text_sources, image_sources, citations = await self.get_sources_content(
+        data_points = await self.get_sources_content(
             results, use_semantic_captions, download_image_sources=send_image_sources, user_oid=auth_claims.get("oid")
         )
+        if not send_text_sources:
+            data_points = DataPoints(text=[], images=data_points.images, citations=data_points.citations)
 
         extra_info = ExtraInfo(
-            DataPoints(text=text_sources if send_text_sources else [], images=image_sources, citations=citations),
+            data_points,
             thoughts=[
                 self.format_thought_step_for_chatcompletion(
                     title="Prompt to generate search query",
@@ -405,15 +407,17 @@ async def run_agentic_retrieval_approach(
             results_merge_strategy=results_merge_strategy,
         )
 
-        text_sources, image_sources, citations = await self.get_sources_content(
+        data_points = await self.get_sources_content(
             results,
             use_semantic_captions=False,
             download_image_sources=send_image_sources,
             user_oid=auth_claims.get("oid"),
         )
+        if not send_text_sources:
+            data_points = DataPoints(text=[], images=data_points.images, citations=data_points.citations)
 
         extra_info = ExtraInfo(
-            DataPoints(text=text_sources if send_text_sources else [], images=image_sources, citations=citations),
+            data_points,
             thoughts=[
                 ThoughtStep(
                     "Use agentic retrieval",

app/backend/approaches/retrievethenread.py

@@ -8,7 +8,6 @@
 
 from approaches.approach import (
     Approach,
-    DataPoints,
     ExtraInfo,
     ThoughtStep,
 )
@@ -183,12 +182,12 @@ async def run_search_approach(
             use_query_rewriting,
         )
 
-        text_sources, image_sources, citations = await self.get_sources_content(
+        data_points = await self.get_sources_content(
             results, use_semantic_captions, download_image_sources=send_image_sources, user_oid=auth_claims.get("oid")
         )
 
         return ExtraInfo(
-            DataPoints(text=text_sources, images=image_sources, citations=citations),
+            data_points,
             thoughts=[
                 ThoughtStep(
                     "Search using user query",
@@ -238,15 +237,15 @@ async def run_agentic_retrieval_approach(
             results_merge_strategy=results_merge_strategy,
         )
 
-        text_sources, image_sources, citations = await self.get_sources_content(
+        data_points = await self.get_sources_content(
             results,
             use_semantic_captions=False,
             download_image_sources=send_image_sources,
             user_oid=auth_claims.get("oid"),
         )
 
         extra_info = ExtraInfo(
-            DataPoints(text=text_sources, images=image_sources, citations=citations),
+            data_points,
             thoughts=[
                 ThoughtStep(
                     "Use agentic retrieval",