fix approx tokens call

src/memory/api/MCP/memory.py

@@ -180,6 +180,7 @@ async def observe(
         session_id: UUID to group observations from same conversation
         agent_model: AI model making observations (for quality tracking)
     """
+    logger.info("MCP: Observing")
     tasks = [
         (
             observation,
@@ -237,6 +238,7 @@ async def search_observations(
     Returns: List with content, tags, created_at, metadata
     Results sorted by relevance to your query.
     """
+    logger.info("MCP: Searching observations for %s", query)
     semantic_text = observation.generate_semantic_text(
         subject=subject or "",
         observation_type="".join(observation_types or []),
@@ -297,6 +299,7 @@ async def create_note(
         confidences: Dict of scores (0.0-1.0), e.g. {"observation_accuracy": 0.9}
         tags: Organization tags for filtering and discovery
     """
+    logger.info("MCP: creating note: %s", subject)
     if filename:
         path = pathlib.Path(filename)
         if not path.is_absolute():

src/memory/api/MCP/tools.py

@@ -108,10 +108,3 @@ async def get_authenticated_user() -> dict:
         "client_id": access_token.client_id,
         "user": user_info,
     }
-
-
-@mcp.tool()
-async def send_response(response: str) -> dict:
-    """Send a response to the user."""
-    logger.info(f"Sending response: {response}")
-    return {"response": response}

src/memory/common/db/models/source_item.py

@@ -28,7 +28,7 @@ from sqlalchemy.dialects.postgresql import BYTEA
 from sqlalchemy.orm import Session, relationship
 from sqlalchemy.types import Numeric
 
-from memory.common import settings
+from memory.common import settings, tokens
 import memory.common.extract as extract
 import memory.common.collections as collections
 import memory.common.chunker as chunker
@@ -125,8 +125,7 @@ def chunk_mixed(content: str, image_paths: Sequence[str]) -> list[extract.DataCh
     )
 
     chunks: list[extract.DataChunk] = [full_text]
-    tokens = chunker.approx_token_count(content)
-    if tokens > chunker.DEFAULT_CHUNK_TOKENS * 2:
+    if tokens.approx_token_count(content) > chunker.DEFAULT_CHUNK_TOKENS * 2:
         chunks += [
             extract.DataChunk(data=add_pics(c, images), metadata={"tags": tags})
             for c in chunker.chunk_text(content)