dry (no-mistakes) (minimal)

server/reflector/hatchet/workflows/diarization_pipeline.py

@@ -558,8 +558,11 @@ async def detect_topics(input: PipelineInput, ctx: Context) -> TopicsResult:
         TranscriptTopic,
         transcripts_controller,
     )
+    from reflector.utils.transcript_constants import (  # noqa: PLC0415
+        TOPIC_CHUNK_WORD_COUNT,
+    )
 
-    chunk_size = 300
+    chunk_size = TOPIC_CHUNK_WORD_COUNT
     chunks = []
     for i in range(0, len(words), chunk_size):
         chunk_words = words[i : i + chunk_size]
@@ -850,6 +853,7 @@ async def generate_recap(input: PipelineInput, ctx: Context) -> RecapResult:
     from reflector.processors.summary.prompts import (  # noqa: PLC0415
         RECAP_PROMPT,
         build_participant_instructions,
+        build_summary_markdown,
     )
 
     subject_summaries = process_result.subject_summaries
@@ -883,18 +887,7 @@ async def generate_recap(input: PipelineInput, ctx: Context) -> RecapResult:
     )
     short_summary = str(recap_response)
 
-    lines = []
-    lines.append("# Quick recap")
-    lines.append("")
-    lines.append(short_summary)
-    lines.append("")
-    lines.append("# Summary")
-    lines.append("")
-    for s in summaries:
-        lines.append(f"**{s['subject']}**")
-        lines.append(s["summary"])
-        lines.append("")
-    long_summary = "\n".join(lines)
+    long_summary = build_summary_markdown(short_summary, summaries)
 
     async with fresh_db_connection():
         transcript = await transcripts_controller.get_by_id(input.transcript_id)

server/reflector/pipelines/topic_processing.py

@@ -18,6 +18,7 @@ from reflector.processors import (
 )
 from reflector.processors.types import TitleSummary
 from reflector.processors.types import Transcript as TranscriptType
+from reflector.utils.transcript_constants import TOPIC_CHUNK_WORD_COUNT
 
 
 class EmptyPipeline:
@@ -38,7 +39,7 @@ async def detect_topics(
     on_topic_callback: Callable,
     empty_pipeline: EmptyPipeline,
 ) -> list[TitleSummary]:
-    chunk_size = 300
+    chunk_size = TOPIC_CHUNK_WORD_COUNT
     topics: list[TitleSummary] = []
 
     async def on_topic(topic: TitleSummary):

server/reflector/processors/summary/prompts.py

@@ -69,3 +69,23 @@ RECAP_PROMPT = dedent(
     As we already know it is a meeting, do not start with 'During the meeting' or equivalent.
     """
 ).strip()
+
+
+def build_summary_markdown(recap: str, summaries: list[dict[str, str]]) -> str:
+    """Build markdown summary from recap and subject summaries."""
+    lines: list[str] = []
+    if recap:
+        lines.append("# Quick recap")
+        lines.append("")
+        lines.append(recap)
+        lines.append("")
+
+    if summaries:
+        lines.append("# Summary")
+        lines.append("")
+        for summary in summaries:
+            lines.append(f"**{summary['subject']}**")
+            lines.append(summary["summary"])
+            lines.append("")
+
+    return "\n".join(lines)

server/reflector/processors/summary/summary_builder.py

@@ -20,6 +20,7 @@ from reflector.processors.summary.prompts import (
     PARAGRAPH_SUMMARY_PROMPT,
     RECAP_PROMPT,
     build_participant_instructions,
+    build_summary_markdown,
 )
 from reflector.settings import settings
 
@@ -556,22 +557,7 @@ class SummaryBuilder:
     # ----------------------------------------------------------------------------
 
     def as_markdown(self) -> str:
-        lines: list[str] = []
-        if self.recap:
-            lines.append("# Quick recap")
-            lines.append("")
-            lines.append(self.recap)
-            lines.append("")
-
-        if self.summaries:
-            lines.append("# Summary")
-            lines.append("")
-            for summary in self.summaries:
-                lines.append(f"**{summary['subject']}**")
-                lines.append(summary["summary"])
-                lines.append("")
-
-        return "\n".join(lines)
+        return build_summary_markdown(self.recap, self.summaries)
 
     def format_list_md(self, data: list[str]) -> str:
         return "\n".join([f"- {item}" for item in data])

server/reflector/utils/transcript_constants.py

@@ -0,0 +1,8 @@
+"""
+Shared transcript processing constants.
+
+Used by both Hatchet workflows and Celery pipelines for consistent processing.
+"""
+
+# Topic detection: number of words per chunk for topic extraction
+TOPIC_CHUNK_WORD_COUNT = 300