Install

Current release: 1.4.2.

Prerequisites

• uv: the only external dependency. memsearch runs through uvx, so there is no Python packaging to manage.
• pi >= 0.84.1. The 0.84.x line is the one the integration suite runs against.
• Node >= 22.19.
• Linux or macOS. Windows needs WSL2, for the reason on the limits page.

Installing

Four forms:

pi install npm:pi-memsearch                               # all projects (~/.pi/settings.json)
pi install npm:pi-memsearch -l                            # this project (.pi/settings.json)
pi install npm:pi-memsearch@1.0.0                         # pinned; `pi update` never advances it
pi install https://github.com/espadat-studio/pi-memsearch # unreleased master

The first run

The first run pauses once, on purpose

uvx resolves memsearch[onnx]>=0.4.17,<0.5, and then the first embedding downloads the onnx model, about 560 MB. That is roughly a 10 s pause on a fast connection, and pi announces it as a notice so it is not mistaken for a hang. It happens once per machine, not once per project.

No API key is involved. When no embedding provider is configured anywhere, pi-memsearch writes embedding.provider = onnx into memsearch’s global config, once. An existing config is never touched: whatever the rest of the mesh already agreed on stays.

After that, leave it alone. Capture runs on every settled exchange, and the store indexes itself.

Confirming it worked

memory_status is the doctor. Ask for it in a session and it reports uv and memsearch presence and version, the resolved scope and collection, the index state and chunk count, and the auto-context state with its counters.

If anything looks wrong, troubleshooting is built around reading that output.

Uninstall

pi remove npm:pi-memsearch

The memory markdown under .memsearch/ survives. It is yours, not the package’s. The collection in ~/.memsearch/milvus.db is derived from that markdown and rebuildable at any time, so nothing irreplaceable ever lived inside pi-memsearch.