limbicdb 5.0.0

Production-grade AI memory infrastructure �?Python SDK with full parity to the TypeScript core.

LimbicDB Python SDK

Production-grade AI memory infrastructure for Python. Full feature parity with the TypeScript/Node.js core — same .limbic database files, identical cognitive scoring, native LangChain and LlamaIndex support.

---

Features

Feature	Status
SQLite storage — reads/writes .limbic files (cross-language compatible)	✅
PostgreSQL + pgvector — cloud/enterprise backend	✅
ExplainEngine — identical Ebbinghaus decay + ACT-R scoring to TS	✅
HLC — Hybrid Logical Clock for P2P sync	✅
LangChain — LimbicChatMemory + LimbicVectorStore	✅
LlamaIndex — LimbicReader + LimbicQueryEngine	✅
pytest — 53+ tests, 80% coverage gate	✅
ruff — zero lint errors	✅

---

Installation

# Core (SQLite only — zero extra deps)
pip install limbicdb

# With PostgreSQL + pgvector
pip install "limbicdb[postgres]"

# With LangChain
pip install "limbicdb[langchain]"

# With LlamaIndex
pip install "limbicdb[llamaindex]"

# Full
pip install "limbicdb[postgres,langchain,llamaindex]"

---

Quick Start

from limbicdb import open as ldb_open

db = ldb_open("./agent.limbic")

db.remember("User prefers dark mode")
db.remember("User is building an AI agent in Python", tags=["project", "ai"])

result = db.recall("Python AI preferences")
for memory in result.memories:
    print(memory.content)

# With full cognitive score explanation
result = db.recall("preferences", explain=True)
for memory in result.memories:
    print(f"[score={memory.explanation.score:.3f}] {memory.content}")
    print(f"  decay={memory.explanation.decay:.3f}")
    print(f"  repetition={memory.explanation.repetition:.3f}")

db.close()

Context Manager

with ldb_open("./agent.limbic") as db:
    db.remember("Context manager example")
    result = db.recall("Context manager")
    print(result.memories[0].content)

---

PostgreSQL Backend

from limbicdb import LimbicDB, LimbicDBConfig
from limbicdb.storage.postgres_store import PostgresStore

store = PostgresStore(
    conninfo="postgresql://postgres:password@localhost:5432/mydb",
    schema="agent_memory",
    embedding_dimensions=1536,
)
store.init()

db = LimbicDB(LimbicDBConfig(storage=store))
db.remember("Stored in PostgreSQL!")
db.close()

Requires pip install "limbicdb[postgres]"

---

LangChain Integration

Chat Memory

from limbicdb.integrations.langchain import LimbicChatMemory
from langchain.chains import ConversationChain
from langchain_openai import ChatOpenAI

memory = LimbicChatMemory(
    path="./chat.limbic",
    session_id="user-42",
    max_history=20,
)

chain = ConversationChain(llm=ChatOpenAI(), memory=memory)
response = chain.invoke({"input": "Hello, my name is Alice"})

RAG Vector Store

from limbicdb.integrations.langchain import LimbicVectorStore

store = LimbicVectorStore(path="./knowledge.limbic")
store.add_texts(["LimbicDB uses Ebbinghaus decay", "Python SDK supports PostgreSQL"])

retriever = store.as_retriever(k=5)
# Use retriever in any LangChain RetrievalQA chain

---

LlamaIndex Integration

from limbicdb.integrations.llamaindex import LimbicQueryEngine

engine = LimbicQueryEngine(path="./agent.limbic", top_k=5)
response = engine.query("UI preferences")
print(response.response)
print(response.source_nodes)

# Or as a Reader for VectorStoreIndex
from limbicdb.integrations.llamaindex import LimbicReader
from llama_index.core import VectorStoreIndex

nodes = LimbicReader(path="./agent.limbic").load_data()
index = VectorStoreIndex(nodes)

---

Cross-Language Interoperability

The Python SDK reads and writes the exact same .limbic format as the TypeScript/Node.js core:

# Python writes
with ldb_open("./shared.limbic") as db:
    db.remember("Written by Python!")

# Node.js reads (TypeScript):
# const db = open('./shared.limbic')
# const result = await db.recall('Python')
# → "Written by Python!"

Enables polyglot Agent architectures:

FastAPI (Python) ←── .limbic ──→ Next.js (TypeScript)

---

ExplainEngine — Score Formula

Component	Formula	Weight
Semantic	Cosine similarity	50%
Decay	Ebbinghaus exp(-t/S)	30%
Repetition	Pimsleur sigmoid over access_count	15%
Recency	exp(-elapsed_days/7)	5%

Weights are identical to the TypeScript ExplainEngine — cross-language score parity guaranteed.

---

Running Tests

cd python

uv sync --extra dev

# Unit tests (no PostgreSQL)
uv run pytest tests/test_core.py tests/test_hlc.py tests/test_integrations.py -v

# PostgreSQL integration tests
PG_TEST_URL=postgres://postgres:password@localhost:5432/testdb \
  uv run pytest tests/test_postgres.py -v

# Lint + type check
uv run ruff check src/
uv run mypy src/

---

Project Layout

python/
├── pyproject.toml
├── src/limbicdb/
│   ├── types.py                  # Memory, Session, MemoryKind, …
│   ├── open.py                   # Convenience open() function
│   ├── core/
│   │   ├── limbicdb.py           # LimbicDB (remember/recall/forget)
│   │   └── explain_engine.py     # ExplainEngine (Ebbinghaus + ACT-R)
│   ├── storage/
│   │   ├── interface.py          # IStorage Protocol
│   │   ├── sqlite_adapter.py     # SQLiteStorageAdapter
│   │   ├── memory_adapter.py     # MemoryStorageAdapter (tests)
│   │   └── postgres_store.py     # PostgresStore (psycopg3 + pgvector)
│   ├── sync/
│   │   └── hlc.py                # HybridLogicalClock
│   └── integrations/
│       ├── langchain.py          # LimbicChatMemory + LimbicVectorStore
│       └── llamaindex.py         # LimbicReader + LimbicQueryEngine
└── tests/
    ├── test_core.py              # 20 core + interop tests
    ├── test_hlc.py               # 12 HLC tests
    ├── test_integrations.py      # 21 LangChain + LlamaIndex tests
    └── test_postgres.py          # 8 PostgreSQL integration tests

---

License

MIT © Kousoyu