langchain-hippodid 0.1.0

LangChain integration for HippoDid — persistent character memory for AI agents

langchain-hippodid

Persistent character memory for LangChain agents — powered by HippoDid.

Add cloud-persistent, character-scoped memory to any LangChain chain in a few lines.

Install

pip install langchain-hippodid

Quick Start

from langchain_anthropic import ChatAnthropic
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.runnables.history import RunnableWithMessageHistory
from langchain_hippodid import HippoDidMemory

memory = HippoDidMemory(api_key="hd_...", character_id="your-character-uuid")

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    MessagesPlaceholder(variable_name="history"),
    ("human", "{input}"),
])

chain_with_memory = RunnableWithMessageHistory(
    prompt | ChatAnthropic(model="claude-sonnet-4-20250514"),
    memory.get_history,
    input_messages_key="input",
    history_messages_key="history",
)

chain_with_memory.invoke(
    {"input": "My name is Yang, I'm building SameThoughts"},
    config={"configurable": {"session_id": "default"}},
)
# Memory persists to HippoDid cloud — survives restarts, redeploys, everything.

Why HippoDid?

Feature	Default LangChain	HippoDidMemory
Persistence	In-process only	Cloud-persistent across sessions
Scope	Global	Character-scoped (one per user)
Retrieval	Flat chat log	Semantic search across structured categories
Context assembly	Append all messages	Profile + category summaries + relevant facts
LLM provider	N/A	Works with any provider (BYOK)

External ID Pattern — One Character Per User

Auto-create and resolve characters by your system's user ID:

from langchain_hippodid import HippoDidMemory

def get_memory(user_id: str) -> HippoDidMemory:
    return HippoDidMemory(
        api_key="hd_...",
        external_id=user_id,  # Auto-creates character if first time
    )

memory = get_memory("user_yang_001")

Memory Modes

Mode	Behavior	Best For
VERBATIM	Stores exact message text, zero AI cost	Chat history replay
EXTRACTED	AI extracts structured facts via AUDN pipeline	Long-term user knowledge
HYBRID	Both verbatim archive and extracted facts	Full fidelity + structured recall

memory = HippoDidMemory(
    api_key="hd_...",
    character_id="...",
    memory_mode="EXTRACTED",
)

Advanced: Custom Search & Context Assembly

Use HippoDidChatMessageHistory directly for fine-grained control:

from langchain_hippodid import HippoDidChatMessageHistory

history = HippoDidChatMessageHistory(
    api_key="hd_...",
    character_id="your-character-uuid",
)

# Semantic search across memories
results = history.search("travel preferences", top_k=5)

# Assemble full context (profile + memories) for a system prompt
context = history.get_context("current user preferences")

Links

• HippoDid — Persistent memory for AI agents
• Documentation
• GitHub
• PyPI
• hippodid Python SDK