eidolondb-adapters 0.1.0

LangChain and LlamaIndex memory adapters for EidolonDB

eidolondb-adapters

LangChain and LlamaIndex memory adapters for EidolonDB.

Installation

pip install "eidolondb-adapters[langchain]"
# or
pip install "eidolondb-adapters[llamaindex]"
# or
pip install "eidolondb-adapters[all]"

Works with both self-hosted EidolonDB and the EidolonDB Cloud API (eidolondb.com).

LangChain Example

from langchain_core.language_models import FakeListLLM
from langchain.chains import ConversationChain

from eidolondb_adapters import EidolonMemory

memory = EidolonMemory(
    url="http://localhost:3000",
    tenant="my-tenant",
    k=5,
    memory_key="history",
    input_key="input",
    output_key="response",
)

llm = FakeListLLM(responses=["I can help with that."])
chain = ConversationChain(llm=llm, memory=memory)

result = chain.predict(input="We chose PostgreSQL for analytics.")
print(result)

LlamaIndex Example

from llama_index.core.chat_engine import SimpleChatEngine
from llama_index.core.llms import ChatMessage, MessageRole

from eidolondb_adapters import EidolonMemoryBuffer

memory = EidolonMemoryBuffer.from_defaults(
    url="http://localhost:3000",
    tenant="my-tenant",
    k=5,
)

# You can still push messages directly if needed.
memory.put(ChatMessage(role=MessageRole.USER, content="My preferred editor is Neovim."))

chat_engine = SimpleChatEngine.from_defaults(memory=memory)
response = chat_engine.chat("What tools do I prefer?")
print(response)

Configuration

Parameter	Applies To	Default	Description
url	LangChain, LlamaIndex	http://localhost:3000	EidolonDB base URL (local or cloud endpoint).
tenant	LangChain, LlamaIndex	default	Tenant ID used by the EidolonDB SDK.
k	LangChain, LlamaIndex	5	Number of recalled memories to fetch per query.
input_key	LangChain	input	Input key used to read the current user prompt.
output_key	LangChain	output	Output key used to ingest the model response.
memory_key	LangChain	history	Variable name injected into LangChain memory context.
return_messages	LangChain	False	Return HumanMessage objects instead of a joined string.