doobidoo/mcp-memory-service v10.14.0

v10.14.0 - conversation_id support for incremental conversation saves

on GitHub

What's New

This release adds a new optional conversation_id parameter to memory_store (MCP tool) and POST /api/memories (REST endpoint), enabling incremental saves from the same conversation without being blocked by semantic deduplication.

Added

• conversation_id parameter for memory_store and REST API (#463): Pass conversation_id when storing memories from the same conversation to allow incremental saves without being blocked by semantic deduplication. Exact hash deduplication is always enforced. The conversation_id is stored in memory metadata for future retrieval/grouping. Applies to both the MCP memory_store tool and the POST /api/memories REST endpoint.

Fixed

• CI: update hash assertion in integration test: test_store_memory_success was asserting "..." in text (truncated hash), but commit 978af00 intentionally changed responses to show the full 64-character content hash. Updated assertion to match the full 64-character content hash pattern to reflect current behaviour.

Files Changed

• src/mcp_memory_service/storage/base.py - skip_semantic_dedup: bool = False on abstract store()
• src/mcp_memory_service/storage/sqlite_vec.py - honours the flag
• src/mcp_memory_service/storage/cloudflare.py - accepts flag (interface compatibility)
• src/mcp_memory_service/storage/hybrid.py - forwards flag to primary backend
• src/mcp_memory_service/services/memory_service.py - conversation_id param + metadata storage
• src/mcp_memory_service/server/handlers/memory.py - extracts conversation_id from MCP arguments
• src/mcp_memory_service/server_impl.py - adds conversation_id to tool inputSchema
• src/mcp_memory_service/web/api/memories.py - adds conversation_id to REST request model
• 3 new tests

Upgrade

pip install --upgrade mcp-memory-service
# or
uv add mcp-memory-service@10.14.0

Full Changelog: https://github.com/doobidoo/mcp-memory-service/blob/main/CHANGELOG.md