Improve Obsidian RAG retrieval for large vaults

- Markdown-aware chunking (split on headers before size-based splitting)
- Prepend note filename to each chunk for self-contained context
- Source-filtered retrieval (obsidian/paperless queries stay isolated)
- MMR search with k=8, fetch_k=24 for better recall and diversity
- Add source metadata to Paperless docs and folder metadata to Obsidian docs

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

blueprints/conversation/agents.py

@@ -121,7 +121,7 @@ async def simba_search(query: str):
         Relevant information from Simba's documents
     """
     print(f"[SIMBA SEARCH] Tool called with query: {query}")
-    serialized, docs = await query_vector_store(query=query)
+    serialized, docs = await query_vector_store(query=query, source="paperless")
     print(f"[SIMBA SEARCH] Found {len(docs)} documents")
     print(f"[SIMBA SEARCH] Serialized result length: {len(serialized)}")
     print(f"[SIMBA SEARCH] First 200 chars: {serialized[:200]}")
@@ -329,8 +329,8 @@ async def obsidian_search_notes(query: str) -> str:
         return "Obsidian integration is not configured. Please set OBSIDIAN_VAULT_PATH environment variable."
 
     try:
-        # Query vector store for obsidian documents
+        # Query vector store filtered to obsidian source only
-        serialized, docs = await query_vector_store(query=query)
+        serialized, docs = await query_vector_store(query=query, source="obsidian")
         return serialized
 
     except Exception as e:

blueprints/rag/logic.py

@@ -3,12 +3,16 @@ import logging
 import os
 import re
 import time
+from pathlib import Path
 
 from dotenv import load_dotenv
 from langchain_core.documents import Document
 from langchain_openai import OpenAIEmbeddings
 from langchain_postgres import PGVector
-from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_text_splitters import (
+    MarkdownHeaderTextSplitter,
+    RecursiveCharacterTextSplitter,
+)
 from sqlalchemy import create_engine, text
 
 from .fetchers import PaperlessNGXService
@@ -69,6 +73,46 @@ text_splitter = RecursiveCharacterTextSplitter(
     add_start_index=True,  # track index in original document
 )
 
+md_header_splitter = MarkdownHeaderTextSplitter(
+    headers_to_split_on=[("#", "h1"), ("##", "h2"), ("###", "h3")],
+    strip_headers=False,
+)
+
+md_chunk_splitter = RecursiveCharacterTextSplitter(
+    chunk_size=1000,
+    chunk_overlap=200,
+    add_start_index=True,
+)
+
+
+def _split_markdown_document(doc: Document) -> list[Document]:
+    """Split a markdown document by headers first, then by size.
+
+    Prepends the note filename to each chunk so chunks are self-contained.
+    """
+    note_name = (
+        Path(doc.metadata.get("filepath", "")).stem
+        if doc.metadata.get("filepath")
+        else ""
+    )
+
+    # Split by markdown headers
+    header_splits = md_header_splitter.split_text(doc.page_content)
+
+    # Carry over original document metadata to each header split
+    for split in header_splits:
+        split.metadata.update(doc.metadata)
+
+    # Then apply size-based splitting on large sections
+    sized_splits = md_chunk_splitter.split_documents(header_splits)
+
+    # Prepend note name for self-contained context
+    if note_name:
+        for split in sized_splits:
+            split.page_content = f"[Note: {note_name}]\n{split.page_content}"
+
+    return sized_splits
+
 
 def _get_collection_id():
     """Get the UUID of our collection from the langchain_pg_collection table."""
@@ -107,6 +151,7 @@ async def fetch_documents_from_paperless_ngx() -> list[Document]:
     documents = []
     for doc in data:
         metadata = {
+            "source": "paperless",
             "created_date": date_to_epoch(doc["created_date"]),
             "filename": doc["original_file_name"],
             "document_type": doctypes.get(doc["document_type"], ""),
@@ -188,6 +233,9 @@ async def fetch_obsidian_documents() -> list[Document]:
             metadata = {
                 "source": "obsidian",
                 "filepath": parsed["filepath"],
+                "folder": str(Path(parsed["filepath"]).parent)
+                if parsed["filepath"]
+                else "",
                 "tags": parsed["tags"],
                 "created_at": parsed["metadata"].get("created_at"),
                 "indexed_at": time.time(),
@@ -224,8 +272,10 @@ async def index_obsidian_documents():
     # Delete existing obsidian chunks
     delete_documents_by_metadata("source", "obsidian")
 
-    # Split, sanitize, and index documents
+    # Split using markdown-aware chunking, sanitize, and index
-    splits = text_splitter.split_documents(documents)
+    splits = []
+    for doc in documents:
+        splits.extend(_split_markdown_document(doc))
     splits = _sanitize_documents(splits)
     vector_store = _get_vector_store()
     await vector_store.aadd_documents(documents=splits)
@@ -315,6 +365,9 @@ async def sync_obsidian_documents() -> dict[str, int]:
                 metadata = {
                     "source": "obsidian",
                     "filepath": parsed["filepath"],
+                    "folder": str(Path(parsed["filepath"]).parent)
+                    if parsed["filepath"]
+                    else "",
                     "tags": parsed["tags"],
                     "created_at": parsed["metadata"].get("created_at"),
                     "indexed_at": now,
@@ -334,7 +387,9 @@ async def sync_obsidian_documents() -> dict[str, int]:
                 continue
 
         if documents:
-            splits = text_splitter.split_documents(documents)
+            splits = []
+            for doc in documents:
+                splits.extend(_split_markdown_document(doc))
             splits = _sanitize_documents(splits)
             if splits:
                 vector_store = _get_vector_store()
@@ -350,9 +405,26 @@ async def sync_obsidian_documents() -> dict[str, int]:
     return {"added": added, "updated": updated, "deleted": deleted}
 
 
-async def query_vector_store(query: str):
+async def query_vector_store(
+    query: str,
+    source: str | None = None,
+    k: int = 8,
+):
+    """Query the vector store with optional source filtering and MMR.
+
+    Args:
+        query: Search query text
+        source: Filter by source metadata (e.g., "obsidian", "paperless")
+        k: Number of results to return
+    """
     vector_store = _get_vector_store()
-    retrieved_docs = await vector_store.asimilarity_search(query, k=6)
+    filter_dict = {"source": source} if source else None
+    retrieved_docs = await vector_store.amax_marginal_relevance_search(
+        query,
+        k=k,
+        fetch_k=k * 3,
+        filter=filter_dict,
+    )
     serialized = "\n\n".join(
         (f"Source: {doc.metadata}\nContent: {doc.page_content}")
         for doc in retrieved_docs