Skip to main content

MemGPT memory architecture as a Python sidecar for the openclaw-memgpt plugin. Includes Anthropic/Claude compatibility patches against the original MemGPT (Apache 2.0).

Project description

openclaw-memgpt-sidecar

A maintained fork of MemGPT that packages the MemGPT memory architecture as a Python sidecar for the openclaw-memgpt plugin.

It preserves MemGPT's memory design — the core / archival / recall memory tiers, the function-calling syscall vocabulary, and the recursive summariser — while modernising the surrounding plumbing so the architecture runs on the current Python LLM ecosystem and against Anthropic/Claude models through OpenAI-compatible proxies.

Relationship to upstream MemGPT

This project is a fork of MemGPT (originally cpacker/MemGPT, now maintained as letta-ai/letta), described in Packer et al., 2024. Originally authored by Packer et al. (2024); maintained as a fork by Altay Acar. All credit for the memory architecture belongs to the original authors:

  • Charles Packer
  • Vivian Fang
  • Sarah Wooders
  • Shishir Patil
  • Kevin Lin
  • Ion Stoica
  • Joseph E. Gonzalez

The fork exists because the original pymemgpt release was written against the 2023 single-vendor OpenAI ecosystem (the pre-v1 functions API, openai SDK v0.28, llama_index 0.8). Running that architecture unchanged against today's SDKs and against non-OpenAI providers (e.g. Claude via LiteLLM) is not possible without compatibility work. This fork supplies exactly that work and nothing more: the memory architecture is preserved; only the plumbing is modernised.

Scope of modifications

The changes are deliberately narrow and limited to compatibility and dependency plumbing:

  • Wire-format modernisation (Anthropic/Claude compatibility) — translation at the SDK boundary from MemGPT's pre-v1 OpenAI functions vocabulary to the modern OpenAI tools API, so payloads reach Claude (and other providers behind OpenAI-compatible proxies) as valid tool-use exchanges.
  • Dependency modernisationopenai SDK v0.28 → v2.x port, llama_index 0.8 → 0.14 migration, and accompanying dependency-pin loosening.
  • Adapter-layer ports — small, behaviour-preserving ports and reuse-oriented refactors that let the existing code run on the modern dependencies without altering the memory architecture.

A full, file-level account of what changed relative to upstream is in CHANGES.md, per the Apache 2.0 modification-notice requirement.

Use

This package is primarily intended as the sidecar dependency of the openclaw-memgpt plugin. Independent use is possible, but API stability is scoped to the plugin's needs and is not guaranteed for other consumers.

Installation

pip install openclaw-memgpt-sidecar==1.0.0

Python 3.11 is required.

License

Licensed under the Apache License, Version 2.0. See LICENSE for the full text. The license is preserved unchanged from upstream MemGPT.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

openclaw_memgpt_sidecar-1.0.0.tar.gz (101.2 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

openclaw_memgpt_sidecar-1.0.0-py3-none-any.whl (139.3 kB view details)

Uploaded Python 3

File details

Details for the file openclaw_memgpt_sidecar-1.0.0.tar.gz.

File metadata

  • Download URL: openclaw_memgpt_sidecar-1.0.0.tar.gz
  • Upload date:
  • Size: 101.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.11.16 {"installer":{"name":"uv","version":"0.11.16","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for openclaw_memgpt_sidecar-1.0.0.tar.gz
Algorithm Hash digest
SHA256 ffc2d1025b10e75c2d66438b1f0e823ebea43e5261d284ee95fb49a9928a72bb
MD5 118fe28a5552ba81fa0042d65adfb77f
BLAKE2b-256 c610bb5887660272bd9afb65444d24e33758e09660c8809f75f2682c2bd7295c

See more details on using hashes here.

File details

Details for the file openclaw_memgpt_sidecar-1.0.0-py3-none-any.whl.

File metadata

  • Download URL: openclaw_memgpt_sidecar-1.0.0-py3-none-any.whl
  • Upload date:
  • Size: 139.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.11.16 {"installer":{"name":"uv","version":"0.11.16","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for openclaw_memgpt_sidecar-1.0.0-py3-none-any.whl
Algorithm Hash digest
SHA256 cfd0b3b1da21371c934f87b5089fd5011e18be91e6b21b941ef7d4060fdd217d
MD5 4f17f163cbf34b7bd311dbbbc562ea93
BLAKE2b-256 6ebf4e50ac98435d661b42064438f49a578ead7960e031c0949ddb543595eaa2

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page