feat: ✨ Working again with proper use of enriched content

src/experts/ingestion_agent.py

@@ -1,33 +1,21 @@
 import dspy
-from pydantic import BaseModel, Field
-
-
-# 1. Define the structure of your metadata
-class DocMetadata(BaseModel):
-    synopsis: str = Field(description="A one-sentence summary of the document.")
-    tags: list[str] = Field(description="Relevant tags (NPCs, Locations, Items, Plot Points).")
-    entities: list[str] = Field(description="Key names of people, places, or factions.")
-
+from typing import List
 
 class IngestionSignature(dspy.Signature):
     """You are an expert Dungeon Master's assistant.
     Analyze the provided notes and extract a concise synopsis and relevant metadata.
+    synopsis = A one-sentence summary of the document.
+    tags = Relevant tags (NPCs, Locations, Items, Plot Points).
+    entities = Key names of people, places, or factions.
+    "note -> synopsis:str, tags: list[str], entities: list[str]"
+    /no_think
     """
 
     note: str = dspy.InputField(desc="The DM notes or session recap content.")
-    # By using the Pydantic model as the type, DSPy handles the JSON formatting for you
-    answer: DocMetadata = dspy.OutputField()
+    answer: dict[str,str|List] = dspy.OutputField(desc="the metadata dictionary with the keys; synopsis, tags, entities")
+
 
 
 class IngestionAgent(dspy.Module):
     def __init__(self):
-        super().__init__()
-        # We use TypedPredictor to enforce the Pydantic schema
-        # We use ChainOfThought because it helps 8B models "reason" through the tags
-        # before committing to the final JSON structure.
-        self.process = dspy.TypedPredictor(IngestionSignature)
-
-    def forward(self, note: str):
-        # The .answer will now be a DocMetadata object, not a string!
-        prediction = self.process(note=note)
-        return prediction
+        self.ingest = dspy.Predict(IngestionSignature)