devmem

Standalone developer memory gateway for cross-agent code assistants (Codex, Claude, Gemini, Grok).

devmem is intentionally separate from the product codebases it serves. It uses isolated namespaces so development-memory data cannot interfere with existing application data.

Goals

• Provide one shared API contract for all code assistants.
• Store semantic context in Milvus and relationship/provenance context in Neo4j.
• Optionally use Aurora Postgres as control-plane storage in a dedicated schema.
• Enforce namespace isolation (devlib_v1 by default) on every operation.

Repo Layout

• devmem/: API/service code.
• docs/: architecture, plan, deployment, and usage docs.
• scripts/: helper CLIs for agents and developers.
• tests/: unit tests.

Quick Start

git clone <your-fork-or-origin> devmem
cd devmem
python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]
uvicorn devmem.main:app --reload --port 8011

Or install the published package (the distribution is named devmem-agents; devmem on PyPI is an unrelated third-party project):

pip install devmem-agents
uvicorn devmem.main:app --port 8011

Local Configuration

Environment-specific values (live backend endpoints, DSNs, SSH tunnel paths) live in devmem.toml (gitignored). Copy the committed template to get started:

cp devmem.toml.example devmem.toml
export DEVMEM_LIVE_TOML_PATH="$(pwd)/devmem.toml"

Any DEVMEM_* environment variable still overrides the TOML.

Documentation

• User Guide — install, configure, run, and troubleshoot.
• Developer Guide — contribute, test, publish to PyPI, autostart with systemd.
• Full docs index
• AGENTS.md — authoritative coding standards for AI coding assistants working on this repo.