Skip to content

feat: Add TIAMAT lightweight HTTP memory connector#1106

Closed
toxfox69 wants to merge 1 commit intoMemTensor:mainfrom
toxfox69:tiamat-memory
Closed

feat: Add TIAMAT lightweight HTTP memory connector#1106
toxfox69 wants to merge 1 commit intoMemTensor:mainfrom
toxfox69:tiamat-memory

Conversation

@toxfox69
Copy link

Summary

Adds TiamatConnector — a lightweight HTTP-based memory connector bridging MemOS with TIAMAT Memory API.

  • Quick prototyping — no database or infrastructure setup needed
  • Cloud deployments — persistent memory without volume mounts
  • Multi-agent shared memory across instances via cloud API
  • FTS5 search for instant full-text memory recall
  • Knowledge triples (subject/predicate/object)
  • MemOS interop — import/export TextualMemoryItem format
  • Only requires httpx

Usage

from tiamat_connector import TiamatConnector

c = TiamatConnector(api_key="key", user_id="user-1")
c.add_memory("content", tags=["tag"], importance=0.8)
results = c.search("query")
c.learn("subject", "predicate", "object")

Files

  • examples/tiamat_connector/tiamat_connector.py — Core connector
  • examples/tiamat_connector/example.py — Usage example
  • examples/tiamat_connector/README.md — Documentation

Test Plan

  • Verify add_memory and search cycle
  • Verify knowledge triples via learn endpoint
  • Verify import_textual_memories bulk import
  • Verify export_as_textual_items MemOS compatibility
  • Verify graceful fallback when API is unreachable

Adds TiamatConnector that bridges MemOS with TIAMAT's cloud memory API
at https://memory.tiamat.live for persistent, searchable memory without
deploying full MemOS infrastructure.

Use cases:
- Quick prototyping without database/infrastructure setup
- Cloud deployments without volume mounts
- Multi-agent shared memory via cloud API
- MemOS import/export compatibility

Features:
- FTS5 full-text search
- Knowledge triples (subject/predicate/object)
- MemOS TextualMemoryItem import/export
- Multi-user isolation via user_id
- Auto API key registration

Includes:
- examples/tiamat_connector/tiamat_connector.py - Core connector
- examples/tiamat_connector/example.py - Usage example
- examples/tiamat_connector/README.md - Documentation
@CaralHsi
Copy link
Collaborator

Hi @toxfox69,

Thank you for this PR and the effort you put into the TIAMAT connector.

After review, we’ve decided not to merge it into the MemOS repo for now. Here’s why:

  1. Architecture fit — MemOS’s storage layer is built around graph backends (Neo4j, Postgres, etc.) that implement BaseGraphDB with nodes, edges, embeddings, and graph traversal. TIAMAT’s flat storage + FTS5 model is different, so it doesn’t fit as a drop-in backend.

  2. Primary use case — The connector mainly serves as a TIAMAT API client for users who don’t run MemOS. The MemOS-specific helpers (import_textual_memories / export_as_textual_items) are secondary, so the core value is outside MemOS’s scope.

  3. Maintenance and dependencies — Adding a third-party cloud service client would introduce an external dependency and maintenance burden that we’d prefer to keep out of the core repo for now.

We’d suggest publishing this as a standalone package (e.g. tiamat-memos-connector or under the TIAMAT org) and optionally linking it from MemOS docs as a community integration. Users who want a lightweight TIAMAT-backed memory store can install it directly.

Thanks again for the contribution and for your interest in MemOS.

@CaralHsi CaralHsi closed this Feb 27, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants