feat(llm): implement llm memory service

README.md

@@ -2,6 +2,8 @@
 
 **AI Notes API** is a production-oriented FastAPI backend for managing AI-related notes, prompts, conversations, and LLM workflows. The project demonstrates clean architecture, async development, PostgreSQL integration, and practical backend patterns for AI engineering.
 
+The assistant is agentic: during chat completions it can call a built-in note toolkit (search, create, read, update, delete notes), and every chat session keeps a long-term memory built from extracted facts and rolling conversation summaries.
+
 ## 📦 Dependencies
 
 * [Python 3.13+](https://www.python.org/downloads/)
@@ -13,7 +15,7 @@ Runtime services:
 
 * [PostgreSQL](https://www.postgresql.org/) - primary data store
 * [Redis](https://redis.io/) - Celery broker and result backend
-* [Celery](https://docs.celeryq.dev/) - background worker for async LLM generation jobs
+* [Celery](https://docs.celeryq.dev/) - background worker for async LLM generation jobs and chat memory updates
 
 ## 📌 API endpoints
 
@@ -41,6 +43,7 @@ Chat session endpoints (authenticated):
 * `PATCH /api/v1/chat/sessions/{session_id}` - update a chat session by ID
 * `DELETE /api/v1/chat/sessions/{session_id}` - delete a chat session by ID
 * `GET /api/v1/chat/sessions/{session_id}/messages` - list messages in a session
+* `GET /api/v1/chat/sessions/{session_id}/memory` - get the long-term memory for a chat session
 
 Chat message endpoints (authenticated):
 

alembic/versions/3e83b1260f95_fix_tables_relations.py

@@ -0,0 +1,34 @@
+"""'Fix tables relations'
+
+Revision ID: 3e83b1260f95
+Revises: 45387e4ce52c
+Create Date: 2026-06-17 01:25:22.088598
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '3e83b1260f95'
+down_revision: Union[str, Sequence[str], None] = '45387e4ce52c'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index(op.f('ix_chat_memories_session_id'), table_name='chat_memories')
+    op.create_index(op.f('ix_chat_memories_session_id'), 'chat_memories', ['session_id'], unique=True)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index(op.f('ix_chat_memories_session_id'), table_name='chat_memories')
+    op.create_index(op.f('ix_chat_memories_session_id'), 'chat_memories', ['session_id'], unique=False)
+    # ### end Alembic commands ###

alembic/versions/45387e4ce52c_create_chat_memories_table.py

@@ -0,0 +1,43 @@
+"""'Create chat_memories table'
+
+Revision ID: 45387e4ce52c
+Revises: a6ca0638c302
+Create Date: 2026-06-17 01:22:47.022259
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision: str = '45387e4ce52c'
+down_revision: Union[str, Sequence[str], None] = 'a6ca0638c302'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('chat_memories',
+    sa.Column('id', sa.Uuid(), nullable=False),
+    sa.Column('session_id', sa.Uuid(), nullable=False),
+    sa.Column('summary', sa.Text(), nullable=False),
+    sa.Column('facts', postgresql.JSONB(astext_type=sa.Text()), nullable=False),
+    sa.Column('created_at', sa.DateTime(timezone=True), nullable=False),
+    sa.Column('updated_at', sa.DateTime(timezone=True), nullable=False),
+    sa.ForeignKeyConstraint(['session_id'], ['chat_sessions.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_chat_memories_session_id'), 'chat_memories', ['session_id'], unique=False)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index(op.f('ix_chat_memories_session_id'), table_name='chat_memories')
+    op.drop_table('chat_memories')
+    # ### end Alembic commands ###

alembic/versions/791080acc188_add_is_summarizing_column_in_chat_.py

@@ -0,0 +1,32 @@
+"""'Add is_summarizing column in chat_memories table'
+
+Revision ID: 791080acc188
+Revises: 3e83b1260f95
+Create Date: 2026-06-18 01:02:05.405819
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '791080acc188'
+down_revision: Union[str, Sequence[str], None] = '3e83b1260f95'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    pass
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    pass
+    # ### end Alembic commands ###

alembic/versions/93228745b965_add_last_summarized_message_id_column_.py

@@ -0,0 +1,36 @@
+"""'Add last_summarized_message_id column in chat_memories table'
+
+Revision ID: 93228745b965
+Revises: 791080acc188
+Create Date: 2026-06-18 01:52:11.606934
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '93228745b965'
+down_revision: Union[str, Sequence[str], None] = '791080acc188'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column('chat_memories', sa.Column('last_summarized_message_id', sa.Uuid(), nullable=True))
+    op.create_index(op.f('ix_chat_memories_last_summarized_message_id'), 'chat_memories', ['last_summarized_message_id'], unique=False)
+    op.create_foreign_key(None, 'chat_memories', 'messages', ['last_summarized_message_id'], ['id'], ondelete='SET NULL')
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_constraint(None, 'chat_memories', type_='foreignkey')
+    op.drop_index(op.f('ix_chat_memories_last_summarized_message_id'), table_name='chat_memories')
+    op.drop_column('chat_memories', 'last_summarized_message_id')
+    # ### end Alembic commands ###

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ai_notes_api"
-version = "0.3.0"
+version = "0.4.0"
 description = "AI Note's API"
 authors = [{ name = "NKTKLN", email = "nktkln@nktkln.com" }]
 readme = "README.md"

src/ai_notes_api/api/v1/chat_sessions.py

@@ -12,10 +12,12 @@
 from ai_notes_api.api.v1.dependencies import (
     get_chat_session_service,
     get_current_user,
+    get_memory_service,
     get_message_service,
 )
 from ai_notes_api.db.models import User
 from ai_notes_api.schemas import (
+    ChatMemoryResponseSchema,
     ChatSessionCreateSchema,
     ChatSessionListQuerySchema,
     ChatSessionListResponseSchema,
@@ -27,7 +29,7 @@
     MessageResponseSchema,
     StatusResponseSchema,
 )
-from ai_notes_api.services import ChatSessionService, MessageService
+from ai_notes_api.services import ChatMemoryService, ChatSessionService, MessageService
 
 router = APIRouter(
     prefix="/chat/sessions",
@@ -276,3 +278,43 @@ async def get_chat_session_messages(
         offset=filters.offset,
         total=len(messages),
     )
+
+
+@router.get(
+    "/{session_id}/memory",
+    summary="Get chat session memory",
+    description="Return memory data for a chat session.",
+    response_model=ChatMemoryResponseSchema,
+    status_code=status.HTTP_200_OK,
+    responses={
+        404: {
+            "model": ErrorResponseSchema,
+            "description": "Chat memory not found",
+        },
+    },
+)
+async def get_chat_session_memory(
+    session_id: UUID,
+    user: Annotated[User, Depends(get_current_user)],
+    service: Annotated[ChatMemoryService, Depends(get_memory_service)],
+) -> ChatMemoryResponseSchema:
+    """Return chat memory for a chat session.
+
+    Args:
+        session_id (UUID): Unique chat session identifier.
+        user (User): Current authenticated user.
+        service (ChatMemoryService): Chat memory service dependency used to
+            retrieve chat memory.
+
+    Returns:
+        ChatMemoryResponseSchema: Chat memory data.
+
+    Raises:
+        ChatMemoryNotFoundError: If no accessible chat memory exists for the
+            given chat session.
+    """
+    logger.info("Chat session memory retrieval requested: session_id={}", session_id)
+
+    memory = await service.get_by_session_id(user.id, session_id)
+
+    return ChatMemoryResponseSchema.model_validate(memory)

src/ai_notes_api/api/v1/dependencies.py

@@ -18,6 +18,7 @@
 from ai_notes_api.integrations import openai_client
 from ai_notes_api.llm import LLMClient
 from ai_notes_api.repositories import (
+    ChatMemoryRepository,
     ChatSessionRepository,
     GenerationJobRepository,
     MessageRepository,
@@ -26,6 +27,7 @@
 )
 from ai_notes_api.services import (
     AuthService,
+    ChatMemoryService,
     ChatSessionService,
     JobService,
     LLMService,
@@ -80,7 +82,7 @@ def get_note_service(
     """
     repository = NoteRepository(session)
 
-    return NoteService(repository=repository)
+    return NoteService(repository)
 
 
 def get_auth_service(
@@ -97,7 +99,7 @@ def get_auth_service(
     """
     repository = UserRepository(session)
 
-    return AuthService(repository=repository)
+    return AuthService(repository)
 
 
 async def get_current_user(
@@ -133,9 +135,13 @@ def get_chat_session_service(
     Returns:
         ChatSessionService: Configured chat session service instance.
     """
-    repository = ChatSessionRepository(session)
+    sessions = ChatSessionRepository(session)
+    memories = ChatMemoryRepository(session)
 
-    return ChatSessionService(repository=repository)
+    return ChatSessionService(
+        session_repository=sessions,
+        memory_repository=memories,
+    )
 
 
 def get_message_service(
@@ -175,9 +181,16 @@ def get_llm_service(
     notes = NoteRepository(session)
     messages = MessageRepository(session)
     sessions = ChatSessionRepository(session)
+    memories = ChatMemoryRepository(session)
     notes_service = NoteService(notes)
-    sessions_service = ChatSessionService(sessions)
-    messages_service = MessageService(messages, sessions)
+    sessions_service = ChatSessionService(
+        session_repository=sessions,
+        memory_repository=memories,
+    )
+    messages_service = MessageService(
+        message_repository=messages,
+        session_repository=sessions,
+    )
 
     return LLMService(
         client=client,
@@ -201,9 +214,30 @@ def get_job_service(
     """
     jobs = GenerationJobRepository(session)
     sessions = ChatSessionRepository(session)
-    sessions_service = ChatSessionService(sessions)
+    memories = ChatMemoryRepository(session)
+    sessions_service = ChatSessionService(
+        session_repository=sessions,
+        memory_repository=memories,
+    )
 
     return JobService(
         job_repository=jobs,
         session_service=sessions_service,
     )
+
+
+def get_memory_service(
+    session: Annotated[AsyncSession, Depends(get_db)],
+) -> ChatMemoryService:
+    """Provide a chat memory service instance.
+
+    Args:
+        session (AsyncSession): Asynchronous database session provided by FastAPI
+            dependency injection.
+
+    Returns:
+        ChatMemoryService: Configured chat memory service instance.
+    """
+    repository = ChatMemoryRepository(session)
+
+    return ChatMemoryService(repository)

src/ai_notes_api/db/models/__init__.py

@@ -4,6 +4,7 @@
 """
 
 from .base import Base
+from .chat_memory import ChatMemory
 from .chat_session import ChatSession, ChatSessionGenerationStatus
 from .datetime import SoftDeleteMixin, TimestampMixin
 from .generation_job import GenerationJob, GenerationJobStatus
@@ -24,4 +25,5 @@
     "GenerationJob",
     "GenerationJobStatus",
     "ChatSessionGenerationStatus",
+    "ChatMemory",
 ]