fidelis-memory 0.2.0

Fidelis Memory 0.2.0 — safer local service startup and installation

on Python PyPI

Why this release

Fidelis 0.2.0 makes local service startup safer around existing installs and lets protocol and health inspection succeed before a local Ollama store is available.

Who it is for

Local-first macOS and Linux users of Codex, Claude Code, GitHub Copilot CLI, Gemini CLI, OpenClaw, and local applications. See the user-fit matrix for prerequisites and non-fits.

Install or upgrade

python3 -m pip install --upgrade "fidelis-memory==0.2.0"

Gemini users can install the matching tagged extension:

gemini extensions install https://github.com/hermes-labs-ai/fidelis --ref=v0.2.0

What changed

• fidelis-server serves health checks without reachable Ollama; memory-backed requests remain unavailable until the local store loads.
• fidelis init detects service-label and port collisions, preserves existing environment settings on re-init, and provides explicit coexistence alternatives.
• FIDELIS_PORT is supported by the server and MCP client when supplied to each process.
• The repository includes a portable Claude plugin root pinned to this release.

Evidence and limits

The release PR passed Linux and macOS CI on Python 3.10–3.14, build validation, independent review, and Hermes Gate. Ollama and the local embedding model remain required for retrieval and ingestion. See the release record, known limitations, and outcome-gated roadmap.