initial server

docker/workers/Dockerfile

@@ -10,21 +10,28 @@ COPY src/ ./src/
 # Install dependencies
 RUN apt-get update && apt-get install -y \
     libpq-dev gcc pandoc \
-    texlive-full texlive-fonts-recommended texlive-plain-generic \
+    texlive-xetex texlive-fonts-recommended texlive-plain-generic \
+    texlive-lang-greek texlive-lang-cyrillic texlive-lang-european \
+    texlive-luatex texlive-latex-extra texlive-latex-recommended \
+    texlive-science texlive-fonts-extra \
+    fontconfig \
     # For optional LibreOffice support (uncomment if needed)
     # libreoffice-writer \
-    && \
+    && apt-get purge -y gcc && apt-get autoremove -y && rm -rf /var/lib/apt/lists/*
-    pip install -e ".[workers]" && \
+RUN pip install -e ".[workers]"
-    apt-get purge -y gcc && apt-get autoremove -y && rm -rf /var/lib/apt/lists/*
 
 # Create and copy entrypoint script
 COPY docker/workers/entry.sh ./entry.sh
+COPY docker/workers/unnest-table.lua ./unnest-table.lua
 RUN chmod +x entry.sh
 
 RUN mkdir -p /app/memory_files
 
 # Create user and set permissions
-RUN useradd -m kb && chown -R kb /app
+RUN useradd -m kb
+RUN mkdir -p /var/cache/fontconfig /home/kb/.cache/fontconfig && \
+    chown -R kb:kb /var/cache/fontconfig /home/kb/.cache/fontconfig /app
+
 USER kb
 
 # Default queues to process

docker/workers/unnest-table.lua

@@ -0,0 +1,28 @@
+local function rows(tbl)
+    local r = pandoc.List()
+    r:extend(tbl.head.rows)
+    for _,b in ipairs(tbl.bodies) do r:extend(b.body) end
+    r:extend(tbl.foot.rows)
+    return r
+  end
+  
+  function Table(t)
+    local newHead = pandoc.TableHead()
+    for i,row in ipairs(t.head.rows) do
+      for j,cell in ipairs(row.cells) do
+        local inner = cell.contents[1]
+        if inner and inner.t == 'Table' then
+          local ins = rows(inner)
+          local first = table.remove(ins,1)
+          row.cells[j] = first.cells[1]
+          newHead.rows:insert(row)
+          newHead.rows:extend(ins)
+          goto continue
+        end
+      end
+      newHead.rows:insert(row)
+      ::continue::
+    end
+    t.head = newHead
+    return t
+  end

requirements-api.txt

@@ -1,4 +1,4 @@
 fastapi==0.112.2
 uvicorn==0.29.0
 python-jose==3.3.0
-python-multipart==0.0.9 
+python-multipart==0.0.9 

requirements-mcp.txt

@@ -0,0 +1,2 @@
+mcp==1.7.1
+httpx==0.25.1

src/memory/api/app.py

@@ -1,23 +1,50 @@
 """
 FastAPI application for the knowledge base.
 """
-from fastapi import FastAPI, Depends, HTTPException
-from sqlalchemy.orm import Session
 
-from memory.common.db import get_scoped_session
+import base64
-from memory.common.db.models import SourceItem
+import io
+from collections import defaultdict
+import pathlib
+from typing import Annotated, List, Optional, Callable
+from fastapi import FastAPI, File, UploadFile, Query, HTTPException, Form
+from fastapi.responses import FileResponse
+import qdrant_client
+from qdrant_client.http import models as qdrant_models
+from PIL import Image
+from pydantic import BaseModel
 
+from memory.common import embedding, qdrant, extract, settings
+from memory.common.embedding import get_modality
+from memory.common.db.connection import make_session
+from memory.common.db.models import Chunk, SourceItem
+
+import logging
+
+logger = logging.getLogger(__name__)
 
 app = FastAPI(title="Knowledge Base API")
 
 
-def get_db():
+class AnnotatedChunk(BaseModel):
-    """Database session dependency"""
+    id: str
-    db = get_scoped_session()
+    score: float
-    try:
+    metadata: dict
-        yield db
+    preview: Optional[str | None] = None
-    finally:
+
-        db.close()
+
+class SearchResponse(BaseModel):
+    collection: str
+    results: List[dict]
+
+
+class SearchResult(BaseModel):
+    id: int
+    size: int
+    mime_type: str
+    chunks: list[AnnotatedChunk]
+    content: Optional[str] = None
+    filename: Optional[str] = None
 
 
 @app.get("/health")
@@ -26,25 +53,195 @@ def health_check():
     return {"status": "healthy"}
 
 
-@app.get("/sources")
+def annotated_chunk(
-def list_sources(
+    chunk: Chunk, search_result: qdrant_models.ScoredPoint, previews: bool
-    tag: str = None,
+) -> tuple[SourceItem, AnnotatedChunk]:
-    limit: int = 100,
+    def serialize_item(item: bytes | str | Image.Image) -> str | None:
-    db: Session = Depends(get_db)
+        if not previews and not isinstance(item, str):
+            return None
+
+        if isinstance(item, Image.Image):
+            buffer = io.BytesIO()
+            format = item.format or "PNG"
+            item.save(buffer, format=format)
+            mime_type = f"image/{format.lower()}"
+            return f"data:{mime_type};base64,{base64.b64encode(buffer.getvalue()).decode('utf-8')}"
+        elif isinstance(item, bytes):
+            return base64.b64encode(item).decode("utf-8")
+        elif isinstance(item, str):
+            return item
+        else:
+            raise ValueError(f"Unsupported item type: {type(item)}")
+
+    metadata = search_result.payload or {}
+    metadata = {
+        k: v
+        for k, v in metadata.items()
+        if k not in ["content", "filename", "size", "content_type", "tags"]
+    }
+    return chunk.source, AnnotatedChunk(
+        id=str(chunk.id),
+        score=search_result.score,
+        metadata=metadata,
+        preview=serialize_item(chunk.data[0]) if chunk.data else None,
+    )
+
+
+def group_chunks(chunks: list[tuple[SourceItem, AnnotatedChunk]]) -> list[SearchResult]:
+    items = defaultdict(list)
+    for source, chunk in chunks:
+        items[source].append(chunk)
+
+    return [
+        SearchResult(
+            id=source.id,
+            size=source.size,
+            mime_type=source.mime_type,
+            filename=source.filename
+            and source.filename.replace(
+                str(settings.FILE_STORAGE_DIR).lstrip("/"), "/files"
+            ),
+            content=source.content,
+            chunks=sorted(chunks, key=lambda x: x.score, reverse=True),
+        )
+        for source, chunks in items.items()
+    ]
+
+
+def query_chunks(
+    client: qdrant_client.QdrantClient,
+    upload_data: list[tuple[str, list[extract.Page]]],
+    allowed_modalities: set[str],
+    embedder: Callable,
+    min_score: float = 0.0,
+    limit: int = 10,
+) -> dict[str, list[qdrant_models.ScoredPoint]]:
+    if not upload_data:
+        return {}
+
+    chunks = [
+        chunk
+        for content_type, pages in upload_data
+        if get_modality(content_type) in allowed_modalities
+        for page in pages
+        for chunk in page["contents"]
+    ]
+
+    if not chunks:
+        return {}
+
+    vector = embedder(chunks, input_type="query")[0]
+
+    return {
+        collection: [
+            r
+            for r in qdrant.search_vectors(
+                client=client,
+                collection_name=collection,
+                query_vector=vector,
+                limit=limit,
+            )
+            if r.score >= min_score
+        ]
+        for collection in embedding.DEFAULT_COLLECTIONS
+    }
+
+
+async def input_type(item: str | UploadFile) -> tuple[str, list[extract.Page]]:
+    if not item:
+        return "text/plain", []
+
+    if isinstance(item, str):
+        return "text/plain", extract.extract_text(item)
+    content_type = item.content_type or "application/octet-stream"
+    return content_type, extract.extract_content(content_type, await item.read())
+
+
+@app.post("/search", response_model=list[SearchResult])
+async def search(
+    query: Optional[str] = Form(None),
+    previews: Optional[bool] = Form(False),
+    modalities: Annotated[list[str], Query()] = [],
+    files: list[UploadFile] = File([]),
+    limit: int = Query(10, ge=1, le=100),
+    min_text_score: float = Query(0.5, ge=0.0, le=1.0),
+    min_multimodal_score: float = Query(0.3, ge=0.0, le=1.0),
 ):
-    """List source items, optionally filtered by tag"""
+    """
-    query = db.query(SourceItem)
+    Search across knowledge base using text query and optional files.
-    
+
-    if tag:
+    Parameters:
-        query = query.filter(SourceItem.tags.contains([tag]))
+    - query: Optional text search query
-    
+    - modalities: List of modalities to search in (e.g., "text", "photo", "doc")
-    return query.limit(limit).all()
+    - files: Optional files to include in the search context
+    - limit: Maximum number of results per modality
+
+    Returns:
+    - List of search results sorted by score
+    """
+    upload_data = [await input_type(item) for item in [query, *files]]
+    logger.error(
+        f"Querying chunks for {modalities}, query: {query}, previews: {previews}"
+    )
+
+    client = qdrant.get_qdrant_client()
+    allowed_modalities = set(modalities or embedding.DEFAULT_COLLECTIONS.keys())
+    text_results = query_chunks(
+        client,
+        upload_data,
+        allowed_modalities & embedding.TEXT_COLLECTIONS,
+        embedding.embed_text,
+        min_score=min_text_score,
+        limit=limit,
+    )
+    multimodal_results = query_chunks(
+        client,
+        upload_data,
+        allowed_modalities,
+        embedding.embed_mixed,
+        min_score=min_multimodal_score,
+        limit=limit,
+    )
+    search_results = {
+        k: text_results.get(k, []) + multimodal_results.get(k, [])
+        for k in allowed_modalities
+    }
+
+    found_chunks = {
+        str(r.id): r for results in search_results.values() for r in results
+    }
+    with make_session() as db:
+        chunks = db.query(Chunk).filter(Chunk.id.in_(found_chunks.keys())).all()
+
+        results = group_chunks(
+            [
+                annotated_chunk(chunk, found_chunks[str(chunk.id)], previews or False)
+                for chunk in chunks
+            ]
+        )
+    return sorted(results, key=lambda x: max(c.score for c in x.chunks), reverse=True)
 
 
-@app.get("/sources/{source_id}")
+@app.get("/files/{path:path}")
-def get_source(source_id: int, db: Session = Depends(get_db)):
+def get_file_by_path(path: str):
-    """Get a specific source by ID"""
+    """
-    source = db.query(SourceItem).filter(SourceItem.id == source_id).first()
+    Fetch a file by its path
-    if not source:
+
-        raise HTTPException(status_code=404, detail="Source not found")
+    Parameters:
-    return source 
+    - path: Path of the file to fetch (relative to FILE_STORAGE_DIR)
+
+    Returns:
+    - The file as a download
+    """
+    # Sanitize the path to prevent directory traversal
+    sanitized_path = path.lstrip("/")
+    if ".." in sanitized_path:
+        raise HTTPException(status_code=400, detail="Invalid path")
+
+    file_path = pathlib.Path(settings.FILE_STORAGE_DIR) / sanitized_path
+
+    # Check if the file exists on disk
+    if not file_path.exists() or not file_path.is_file():
+        raise HTTPException(status_code=404, detail=f"File not found at path: {path}")
+
+    return FileResponse(path=file_path, filename=file_path.name)

src/memory/common/db/models.py

@@ -113,7 +113,7 @@ class Chunk(Base):
         if self.file_path is None:
             return [self.content]
 
-        path = pathlib.Path(self.file_path)
+        path = pathlib.Path(self.file_path.replace("/app/", ""))
         if self.file_path.endswith("*"):
             files = list(path.parent.glob(path.name))
         else:
@@ -122,7 +122,8 @@ class Chunk(Base):
         items = []
         for file_path in files:
             if file_path.suffix == ".png":
-                items.append(Image.open(file_path))
+                if file_path.exists():
+                    items.append(Image.open(file_path))
             elif file_path.suffix == ".bin":
                 items.append(file_path.read_bytes())
             else:

src/memory/common/embedding.py

@@ -1,7 +1,7 @@
 import logging
 import pathlib
 import uuid
-from typing import Any, Iterable, Literal, TypedDict
+from typing import Any, Iterable, Literal, NotRequired, TypedDict
 
 import voyageai
 from PIL import Image
@@ -27,20 +27,63 @@ Embedding = tuple[str, Vector, dict[str, Any]]
 class Collection(TypedDict):
     dimension: int
     distance: DistanceType
-    on_disk: bool
+    model: str
-    shards: int
+    on_disk: NotRequired[bool]
+    shards: NotRequired[int]
 
 
 DEFAULT_COLLECTIONS: dict[str, Collection] = {
-    "mail": {"dimension": 1024, "distance": "Cosine"},
+    "mail": {
-    "chat": {"dimension": 1024, "distance": "Cosine"},
+        "dimension": 1024,
-    "git": {"dimension": 1024, "distance": "Cosine"},
+        "distance": "Cosine",
-    "book": {"dimension": 1024, "distance": "Cosine"},
+        "model": settings.TEXT_EMBEDDING_MODEL,
-    "blog": {"dimension": 1024, "distance": "Cosine"},
+    },
-    "text": {"dimension": 1024, "distance": "Cosine"},
+    "chat": {
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.TEXT_EMBEDDING_MODEL,
+    },
+    "git": {
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.TEXT_EMBEDDING_MODEL,
+    },
+    "book": {
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.TEXT_EMBEDDING_MODEL,
+    },
+    "blog": {
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.TEXT_EMBEDDING_MODEL,
+    },
+    "text": {
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.TEXT_EMBEDDING_MODEL,
+    },
     # Multimodal
-    "photo": {"dimension": 1024, "distance": "Cosine"},
+    "photo": {
-    "doc": {"dimension": 1024, "distance": "Cosine"},
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.MIXED_EMBEDDING_MODEL,
+    },
+    "doc": {
+        "dimension": 1024,
+        "distance": "Cosine",
+        "model": settings.MIXED_EMBEDDING_MODEL,
+    },
+}
+TEXT_COLLECTIONS = {
+    coll
+    for coll, params in DEFAULT_COLLECTIONS.items()
+    if params["model"] == settings.TEXT_EMBEDDING_MODEL
+}
+MULTIMODAL_COLLECTIONS = {
+    coll
+    for coll, params in DEFAULT_COLLECTIONS.items()
+    if params["model"] == settings.MIXED_EMBEDDING_MODEL
 }
 
 TYPES = {
@@ -69,22 +112,27 @@ def get_modality(mime_type: str) -> str:
 
 
 def embed_chunks(
-    chunks: list[extract.MulitmodalChunk], model: str = settings.TEXT_EMBEDDING_MODEL
+    chunks: list[extract.MulitmodalChunk],
+    model: str = settings.TEXT_EMBEDDING_MODEL,
+    input_type: Literal["document", "query"] = "document",
 ) -> list[Vector]:
     vo = voyageai.Client()
     if model == settings.MIXED_EMBEDDING_MODEL:
         return vo.multimodal_embed(
-            chunks, model=model, input_type="document"
+            chunks, model=model, input_type=input_type
         ).embeddings
-    return vo.embed(chunks, model=model, input_type="document").embeddings
+    return vo.embed(chunks, model=model, input_type=input_type).embeddings
 
 
 def embed_text(
-    texts: list[str], model: str = settings.TEXT_EMBEDDING_MODEL
+    texts: list[str],
+    model: str = settings.TEXT_EMBEDDING_MODEL,
+    input_type: Literal["document", "query"] = "document",
 ) -> list[Vector]:
     chunks = [
         c
         for text in texts
+        if isinstance(text, str)
         for c in chunk_text(text, MAX_TOKENS, OVERLAP_TOKENS)
         if c.strip()
     ]
@@ -92,7 +140,7 @@ def embed_text(
         return []
 
     try:
-        return embed_chunks(chunks, model)
+        return embed_chunks(chunks, model, input_type)
     except voyageai.error.InvalidRequestError as e:
         logger.error(f"Error embedding text: {e}")
         logger.debug(f"Text: {texts}")
@@ -106,7 +154,9 @@ def embed_file(
 
 
 def embed_mixed(
-    items: list[extract.MulitmodalChunk], model: str = settings.MIXED_EMBEDDING_MODEL
+    items: list[extract.MulitmodalChunk],
+    model: str = settings.MIXED_EMBEDDING_MODEL,
+    input_type: Literal["document", "query"] = "document",
 ) -> list[Vector]:
     def to_chunks(item: extract.MulitmodalChunk) -> Iterable[str]:
         if isinstance(item, str):
@@ -116,7 +166,7 @@ def embed_mixed(
         return [item]
 
     chunks = [c for item in items for c in to_chunks(item)]
-    return embed_chunks([chunks], model)
+    return embed_chunks([chunks], model, input_type)
 
 
 def embed_page(page: dict[str, Any]) -> list[Vector]:

src/memory/common/extract.py

@@ -7,6 +7,9 @@ import pymupdf  # PyMuPDF
 from PIL import Image
 from typing import Any, TypedDict, Generator, Sequence
 
+import logging
+
+logger = logging.getLogger(__name__)
 
 MulitmodalChunk = Image.Image | str
 
@@ -57,15 +60,32 @@ def docx_to_pdf(
     if output_path is None:
         output_path = docx_path.with_suffix(".pdf")
 
-    pypandoc.convert_file(str(docx_path), "pdf", outputfile=str(output_path))
+    # Now that we have all packages installed, try xelatex first as it has better Unicode support
-
+    try:
-    return output_path
+        logger.info(f"Converting {docx_path} to PDF using xelatex")
+        pypandoc.convert_file(
+            str(docx_path),
+            format="docx",
+            to="pdf",
+            outputfile=str(output_path),
+            extra_args=[
+                "--pdf-engine=xelatex",
+                "--variable=geometry:margin=1in",
+                "--lua-filter=/app/unnest-table.lua",
+            ],
+        )
+        logger.info(f"Successfully converted {docx_path} to PDF")
+        return output_path
+    except Exception as e:
+        logger.error(f"Error converting document to PDF: {e}")
+        raise
 
 
-def extract_docx(docx_path: pathlib.Path) -> list[Page]:
+def extract_docx(docx_path: pathlib.Path | bytes | str) -> list[Page]:
     """Extract content from DOCX by converting to PDF first, then processing"""
     with as_file(docx_path) as file_path:
         pdf_path = docx_to_pdf(file_path)
+        logger.info(f"Extracted PDF from {file_path}")
         return doc_to_images(pdf_path)
 
 
@@ -89,13 +109,17 @@ def extract_text(content: bytes | str | pathlib.Path) -> list[Page]:
 
 
 def extract_content(mime_type: str, content: bytes | str | pathlib.Path) -> list[Page]:
+    logger.info(f"Extracting content from {mime_type}")
     if mime_type == "application/pdf":
         return doc_to_images(content)
-    if isinstance(content, pathlib.Path) and mime_type in [
+    if mime_type in [
         "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
         "application/msword",
     ]:
-        return extract_docx(content)
+        logger.info(f"Extracting content from {content}")
+        pages = extract_docx(content)
+        logger.info(f"Extracted {len(pages)} pages from {content}")
+        return pages
     if mime_type.startswith("text/"):
         return extract_text(content)
     if mime_type.startswith("image/"):

src/memory/common/parsers/email.py

@@ -25,6 +25,7 @@ class EmailMessage(TypedDict):
     sent_at: datetime | None
     body: str
     attachments: list[Attachment]
+    hash: bytes
 
 
 RawEmailResponse = tuple[Literal["OK", "ERROR"], bytes]
@@ -33,10 +34,10 @@ RawEmailResponse = tuple[Literal["OK", "ERROR"], bytes]
 def extract_recipients(msg: email.message.Message) -> list[str]:
     """
     Extract email recipients from message headers.
-    
+
     Args:
         msg: Email message object
-        
+
     Returns:
         List of recipient email addresses
     """
@@ -52,10 +53,10 @@ def extract_recipients(msg: email.message.Message) -> list[str]:
 def extract_date(msg: email.message.Message) -> datetime | None:
     """
     Parse date from email header.
-    
+
     Args:
         msg: Email message object
-        
+
     Returns:
         Parsed datetime or None if parsing failed
     """
@@ -70,18 +71,18 @@ def extract_date(msg: email.message.Message) -> datetime | None:
 def extract_body(msg: email.message.Message) -> str:
     """
     Extract plain text body from email message.
-    
+
     Args:
         msg: Email message object
-        
+
     Returns:
         Plain text body content
     """
     body = ""
-    
+
     if not msg.is_multipart():
         try:
-            return msg.get_payload(decode=True).decode(errors='replace')
+            return msg.get_payload(decode=True).decode(errors="replace")
         except Exception as e:
             logger.error(f"Error decoding message body: {str(e)}")
             return ""
@@ -89,10 +90,10 @@ def extract_body(msg: email.message.Message) -> str:
     for part in msg.walk():
         content_type = part.get_content_type()
         content_disposition = str(part.get("Content-Disposition", ""))
-        
+
         if content_type == "text/plain" and "attachment" not in content_disposition:
             try:
-                body += part.get_payload(decode=True).decode(errors='replace') + "\n"
+                body += part.get_payload(decode=True).decode(errors="replace") + "\n"
             except Exception as e:
                 logger.error(f"Error decoding message part: {str(e)}")
     return body
@@ -101,10 +102,10 @@ def extract_body(msg: email.message.Message) -> str:
 def extract_attachments(msg: email.message.Message) -> list[Attachment]:
     """
     Extract attachment metadata and content from email.
-    
+
     Args:
         msg: Email message object
-        
+
     Returns:
         List of attachment dictionaries with metadata and content
     """
@@ -120,14 +121,18 @@ def extract_attachments(msg: email.message.Message) -> list[Attachment]:
         if filename := part.get_filename():
             try:
                 content = part.get_payload(decode=True)
-                attachments.append({
+                attachments.append(
-                    "filename": filename,
+                    {
-                    "content_type": part.get_content_type(),
+                        "filename": filename,
-                    "size": len(content),
+                        "content_type": part.get_content_type(),
-                    "content": content
+                        "size": len(content),
-                })
+                        "content": content,
+                    }
+                )
             except Exception as e:
-                logger.error(f"Error extracting attachment content for {filename}: {str(e)}")
+                logger.error(
+                    f"Error extracting attachment content for {filename}: {str(e)}"
+                )
 
     return attachments
 
@@ -135,13 +140,13 @@ def extract_attachments(msg: email.message.Message) -> list[Attachment]:
 def compute_message_hash(msg_id: str, subject: str, sender: str, body: str) -> bytes:
     """
     Compute a SHA-256 hash of message content.
-    
+
     Args:
         msg_id: Message ID
         subject: Email subject
         sender: Sender email
         body: Message body
-        
+
     Returns:
         SHA-256 hash as bytes
     """
@@ -152,10 +157,10 @@ def compute_message_hash(msg_id: str, subject: str, sender: str, body: str) -> b
 def parse_email_message(raw_email: str, message_id: str) -> EmailMessage:
     """
     Parse raw email into structured data.
-    
+
     Args:
         raw_email: Raw email content as string
-        
+
     Returns:
         Dict with parsed email data
     """
@@ -164,7 +169,7 @@ def parse_email_message(raw_email: str, message_id: str) -> EmailMessage:
     subject = msg.get("Subject", "")
     from_ = msg.get("From", "")
     body = extract_body(msg)
-    
+
     return EmailMessage(
         message_id=message_id,
         subject=subject,
@@ -173,5 +178,5 @@ def parse_email_message(raw_email: str, message_id: str) -> EmailMessage:
         sent_at=extract_date(msg),
         body=body,
         attachments=extract_attachments(msg),
-        hash=compute_message_hash(message_id, subject, from_, body)
+        hash=compute_message_hash(message_id, subject, from_, body),
     )

src/memory/common/qdrant.py

@@ -5,15 +5,22 @@ import qdrant_client
 from qdrant_client.http import models as qdrant_models
 from qdrant_client.http.exceptions import UnexpectedResponse
 from memory.common import settings
-from memory.common.embedding import Collection, DEFAULT_COLLECTIONS, DistanceType, Vector
+from memory.common.embedding import (
+    Collection,
+    DEFAULT_COLLECTIONS,
+    DistanceType,
+    Vector,
+)
 
 logger = logging.getLogger(__name__)
 
 
 def get_qdrant_client() -> qdrant_client.QdrantClient:
     """Create and return a Qdrant client using environment configuration."""
-    logger.info(f"Connecting to Qdrant at {settings.QDRANT_HOST}:{settings.QDRANT_PORT}")
+    logger.info(
-    
+        f"Connecting to Qdrant at {settings.QDRANT_HOST}:{settings.QDRANT_PORT}"
+    )
+
     return qdrant_client.QdrantClient(
         host=settings.QDRANT_HOST,
         port=settings.QDRANT_PORT,
@@ -34,7 +41,7 @@ def ensure_collection_exists(
 ) -> bool:
     """
     Ensure a collection exists with the specified parameters.
-    
+
     Args:
         client: Qdrant client
         collection_name: Name of the collection
@@ -42,7 +49,7 @@ def ensure_collection_exists(
         distance: Distance metric (Cosine, Dot, Euclidean)
         on_disk: Whether to store vectors on disk
         shards: Number of shards for the collection
-        
+
     Returns:
         True if the collection was created, False if it already existed
     """
@@ -61,21 +68,23 @@ def ensure_collection_exists(
             on_disk_payload=on_disk,
             shard_number=shards,
         )
-        
+
         # Create common payload indexes
         client.create_payload_index(
             collection_name=collection_name,
             field_name="tags",
             field_schema=qdrant_models.PayloadSchemaType.KEYWORD,
         )
-        
+
         return True
 
 
-def initialize_collections(client: qdrant_client.QdrantClient, collections: dict[str, Collection] = None) -> None:
+def initialize_collections(
+    client: qdrant_client.QdrantClient, collections: dict[str, Collection] = None
+) -> None:
     """
     Initialize all required collections in Qdrant.
-    
+
     Args:
         client: Qdrant client
         collections: Dictionary mapping collection names to their parameters.
@@ -83,7 +92,7 @@ def initialize_collections(client: qdrant_client.QdrantClient, collections: dict
     """
     if collections is None:
         collections = DEFAULT_COLLECTIONS
-    
+
     logger.info(f"Initializing collections:")
     for name, params in collections.items():
         logger.info(f" - {name}")
@@ -99,7 +108,7 @@ def initialize_collections(client: qdrant_client.QdrantClient, collections: dict
 
 def setup_qdrant() -> qdrant_client.QdrantClient:
     """Get a Qdrant client and initialize collections.
-    
+
     Returns:
         Configured Qdrant client
     """
@@ -109,14 +118,14 @@ def setup_qdrant() -> qdrant_client.QdrantClient:
 
 
 def upsert_vectors(
-    client: qdrant_client.QdrantClient, 
+    client: qdrant_client.QdrantClient,
-    collection_name: str, 
+    collection_name: str,
     ids: list[str],
     vectors: list[Vector],
     payloads: list[dict[str, Any]] = None,
 ) -> None:
     """Upsert vectors into a collection.
-    
+
     Args:
         client: Qdrant client
         collection_name: Name of the collection
@@ -126,7 +135,7 @@ def upsert_vectors(
     """
     if payloads is None:
         payloads = [{} for _ in ids]
-    
+
     points = [
         qdrant_models.PointStruct(
             id=id_str,
@@ -135,12 +144,12 @@ def upsert_vectors(
         )
         for id_str, vector, payload in zip(ids, vectors, payloads)
     ]
-    
+
     client.upsert(
         collection_name=collection_name,
         points=points,
     )
-    
+
     logger.debug(f"Upserted {len(ids)} vectors into {collection_name}")
 
 
@@ -152,21 +161,21 @@ def search_vectors(
     limit: int = 10,
 ) -> list[qdrant_models.ScoredPoint]:
     """Search for similar vectors in a collection.
-    
+
     Args:
         client: Qdrant client
         collection_name: Name of the collection
         query_vector: Query vector
         filter_params: Filter parameters to apply (e.g., {"tags": {"value": "work"}})
         limit: Maximum number of results to return
-        
+
     Returns:
         List of scored points
     """
     filter_obj = None
     if filter_params:
         filter_obj = qdrant_models.Filter(**filter_params)
-    
+
     return client.search(
         collection_name=collection_name,
         query_vector=query_vector,
@@ -182,7 +191,7 @@ def delete_vectors(
 ) -> None:
     """
     Delete vectors from a collection.
-    
+
     Args:
         client: Qdrant client
         collection_name: Name of the collection
@@ -194,18 +203,20 @@ def delete_vectors(
             points=ids,
         ),
     )
-    
+
     logger.debug(f"Deleted {len(ids)} vectors from {collection_name}")
 
 
-def get_collection_info(client: qdrant_client.QdrantClient, collection_name: str) -> dict:
+def get_collection_info(
+    client: qdrant_client.QdrantClient, collection_name: str
+) -> dict:
     """
     Get information about a collection.
-    
+
     Args:
         client: Qdrant client
         collection_name: Name of the collection
-        
+
     Returns:
         Dictionary with collection information
     """

src/memory/common/settings.py

@@ -52,7 +52,6 @@ QDRANT_TIMEOUT = int(os.getenv("QDRANT_TIMEOUT", "60"))
 
 # Worker settings
 EMAIL_SYNC_INTERVAL = int(os.getenv("EMAIL_SYNC_INTERVAL", 3600))
-EMAIL_SYNC_INTERVAL = 60
 
 
 # Embedding settings

src/memory/mcp/server.py

@@ -0,0 +1,58 @@
+import argparse
+import logging
+from typing import Any
+from fastapi import UploadFile
+import httpx
+from mcp.server.fastmcp import FastMCP
+
+SERVER = "http://localhost:8000"
+
+
+logger = logging.getLogger(__name__)
+mcp = FastMCP("memory")
+
+
+async def make_request(
+    path: str,
+    method: str,
+    data: dict | None = None,
+    json: dict | None = None,
+    files: list[UploadFile] | None = None,
+) -> httpx.Response:
+    async with httpx.AsyncClient() as client:
+        return await client.request(
+            method, f"{SERVER}/{path}", data=data, json=json, files=files
+        )
+
+
+async def post_data(path: str, data: dict | None = None) -> httpx.Response:
+    return await make_request(path, "POST", data=data)
+
+
+@mcp.tool()
+async def search(
+    query: str, previews: bool = False, modalities: list[str] = [], limit: int = 10
+) -> list[dict[str, Any]]:
+    logger.error(f"Searching for {query}")
+    resp = await post_data(
+        "search",
+        {
+            "query": query,
+            "previews": previews,
+            "modalities": modalities,
+            "limit": limit,
+        },
+    )
+
+    return resp.json()
+
+
+if __name__ == "__main__":
+    # Initialize and run the server
+    args = argparse.ArgumentParser()
+    args.add_argument("--server", type=str)
+    args = args.parse_args()
+
+    SERVER = args.server
+
+    mcp.run(transport=args.transport)

src/memory/workers/email.py

@@ -124,13 +124,14 @@ def create_mail_message(
         folder=folder,
     )
 
-    if parsed_email["attachments"]:
-        mail_message.attachments = process_attachments(
-            parsed_email["attachments"], mail_message
-        )
-
     db_session.add(mail_message)
 
+    if parsed_email["attachments"]:
+        attachments = process_attachments(parsed_email["attachments"], mail_message)
+        db_session.add_all(attachments)
+        mail_message.attachments = attachments
+
+    db_session.add(mail_message)
     return mail_message
 
 
@@ -171,13 +172,11 @@ def check_message_exists(
     account = db.query(EmailAccount).get(account_id)
     if not account:
         logger.error(f"Account {account_id} not found")
-        return None
+        return False
 
     parsed_email = parse_email_message(raw_email, message_id)
-
+    if "szczepalins" in raw_email.lower():
-    # Use server-provided message ID if missing
+        print(parsed_email["message_id"])
-    if not parsed_email["message_id"]:
-        parsed_email["message_id"] = f"generated-{message_id}"
 
     return does_message_exist(db, parsed_email["message_id"], parsed_email["hash"])
 

src/memory/workers/tasks/email.py

@@ -58,6 +58,7 @@ def process_message(
             db, account.tags, folder, raw_email, message_id
         )
 
+        db.flush()
         vectorize_email(mail_message)
         db.commit()
 
@@ -65,12 +66,16 @@ def process_message(
         logger.info("Chunks:")
         for chunk in mail_message.chunks:
             logger.info(f" - {chunk.id}")
+        for attachment in mail_message.attachments:
+            logger.info(f" - Attachment {attachment.id}")
+            for chunk in attachment.chunks:
+                logger.info(f"   - {chunk.id}")
 
         return mail_message.id
 
 
 @app.task(name=SYNC_ACCOUNT)
-def sync_account(account_id: int) -> dict:
+def sync_account(account_id: int, since_date: str | None = None) -> dict:
     """
     Synchronize emails from a specific account.
 
@@ -80,7 +85,7 @@ def sync_account(account_id: int) -> dict:
     Returns:
         dict with stats about the sync operation
     """
-    logger.info(f"Syncing account {account_id}")
+    logger.info(f"Syncing account {account_id} since {since_date}")
 
     with make_session() as db:
         account = db.query(EmailAccount).filter(EmailAccount.id == account_id).first()
@@ -88,8 +93,11 @@ def sync_account(account_id: int) -> dict:
             logger.warning(f"Account {account_id} not found or inactive")
             return {"error": "Account not found or inactive"}
 
-        folders_to_process = account.folders or ["INBOX"]
+        folders_to_process: list[str] = account.folders or ["INBOX"]
-        since_date = account.last_sync_at or datetime(1970, 1, 1)
+        if since_date:
+            cutoff_date = datetime.fromisoformat(since_date)
+        else:
+            cutoff_date: datetime = account.last_sync_at or datetime(1970, 1, 1)
 
         messages_found = 0
         new_messages = 0
@@ -106,7 +114,7 @@ def sync_account(account_id: int) -> dict:
             with imap_connection(account) as conn:
                 for folder in folders_to_process:
                     folder_stats = process_folder(
-                        conn, folder, account, since_date, process_message_wrapper
+                        conn, folder, account, cutoff_date, process_message_wrapper
                     )
 
                     messages_found += folder_stats["messages_found"]
@@ -121,6 +129,7 @@ def sync_account(account_id: int) -> dict:
 
         return {
             "account": account.email_address,
+            "since_date": cutoff_date.isoformat(),
             "folders_processed": len(folders_to_process),
             "messages_found": messages_found,
             "new_messages": new_messages,