MemPalace

3.3.4 verified Sat May 09 auth: no python

MemPalace (v3.3.4) gives AI applications a memory by mining projects and conversations into a searchable, structured knowledge base. It uses a local language model (Ollama/LM Studio) for free setup, no API key required. Release cadence is active with multiple minor releases per month.

pip install mempalace

Common errors

error ModuleNotFoundError: No module named 'mempalace' ↓

cause Package not installed or installed in wrong environment.

fix

pip install mempalace

error AttributeError: module 'mempalace' has no attribute 'MemPalace' ↓

cause Importing the module directly instead of the class.

fix

from mempalace import MemPalace

error mempalace-mcp: command not found ↓

cause Installation of v3.3.2 missing the console script entry point.

fix

pip install mempalace>=3.3.3

error Ollama connection refused ↓

cause Local LLM server (Ollama) is not running or not reachable on default port.

fix

Start Ollama with ollama serve or configure a different host/port via environment variable.

Warnings

breaking In v3.3.2, the `mempalace-mcp` console script was missing from pyproject.toml, causing plugin installation failures. Upgrade to v3.3.4 to ensure the binary is present. ↓

fix pip install mempalace>=3.3.3

gotcha MemPalace by default uses a local LLM (Ollama/LM Studio/llama.cpp) for zero-cost setup. If no local LLM is running, it falls back gracefully but may not mine effectively. ↓

fix Ensure a local LLM server is running, or set environment variables to configure an external API endpoint.

deprecated The old API `import mempalace` without explicit class import is deprecated and may be removed in a future version. ↓

fix Use `from mempalace import MemPalace` instead.

Imports

MemPalace
wrong
```
import mempalace
```
correct
```
from mempalace import MemPalace
```
Direct import of the class is required; bare module import does not expose the API.

Quickstart

Initialize MemPalace, mine a directory, and search for a concept.

from mempalace import MemPalace
mp = MemPalace()
mp.mine(directory='./project')
results = mp.search('key concept')
print(results)