Relurpify is a fullstack Agent framework
- generic execution Agent library
- LLM oriented memory/context/graph/sandbox management framework
- archaeology memory system
- extensive testsuite
- (Euclo) coding agent
- TUI interfaces
- Default Agent TUI
- Euclo coding agent access
- Go
1.25+ - Docker or another supported container runtime
- gVisor
runsc - Ollama (default inference provider)
Developer documentation lives in the repository's devdocs folder;
developer-documentation.md is the canonical internal reference. Contribution
gates, coverage requirements, and the release process are described in
CONTRIBUTING.md.
The go build examples in this README were verified in this workspace. The
runtime examples are documented invocation shapes, not all re-run here.
Running without Ollama (CI/plumbing only):
go build ./app/relurpish ./relurpish doctor --offline --fixThe built-in
offlinebackend is a deterministic scripted model that exercises the agent plumbing (tool dispatch, streaming, compilation) without requiring Ollama, Docker, or network access. It is intended for testing and CI — not for end-user demo or production use.
In sandboxed environments you may also want repo-local Go caches:
export GOMODCACHE=$PWD/.gomodcache
export GOCACHE=$PWD/.gocachego build ./app/relurpishPre-built artifacts are attached to every
GitHub release: per-platform
archives, checksums.txt, SBOMs (*.sbom.json), .deb/.rpm packages, and
an AUR package. Releases are created as drafts; the maintainer reviews and
publishes them.
relurpish, dev-agent, relurplint, and generate-config are built for
Linux and macOS (amd64 + arm64). The internal tooling binaries archcheck,
domaincheck, and driftcheck additionally build for Windows (zip archives).
Every binary is statically linked (CGO disabled).
wget https://github.com/lexbit/relurpify/releases/download/v0.1.0/relurpify_0.1.0_amd64.deb
sudo dpkg -i relurpify_0.1.0_amd64.debwget https://github.com/lexbit/relurpify/releases/download/v0.1.0/relurpify-0.1.0-1.x86_64.rpm
sudo rpm -i relurpify-0.1.0-1.x86_64.rpmwget https://github.com/lexbit/relurpify/releases/download/v0.1.0/relurpify_0.1.0_linux_amd64.tar.gz
tar -xzf relurpify_0.1.0_linux_amd64.tar.gzDownload checksums.txt from the same release and verify before installing:
sha256sum -c checksums.txtyay -S relurpify-binRelease binaries report their build metadata (--version on relurpish and
dev-agent):
relurpish --version
# relurpish 0.1.0 (commit 594da458..., built 2026-10-08T...)Source builds report dev / none / unknown.
go build ./...Before starting a chat session, ensure Ollama is running and has a model pulled:
# Start the Ollama daemon (in a separate terminal)
ollama serve
# Pull a model that the default catalog references
ollama pull gemma4:e4bThen run doctor to verify the workspace is ready:
go build ./app/relurpish
./relurpish doctorStart the terminal app with:
go run ./app/relurpish chatThis launches relurpish and starts the default Euclo coding workflow in the current workspace.
For a typical first-use flow:
go build ./app/relurpish
go run ./app/relurpish doctor
go run ./app/relurpish chatThe main local checks are:
make test-unit
make test-dev-agent
make test-tape-fidelitySee developer-documentation.md in the repository's devdocs folder for the
full matrix and agent-test workflow.
Relurpify also includes developer tooling for internal workflows and testing:
# List discovered agents
go run ./app/dev-agent-cli agents list
# Run agent tests
go run ./app/dev-agent-cli agenttest run
# Scaffold a skill
go run ./app/dev-agent-cli skill init my-skill --description "My focused workflow" --with-tests
# Validate a skill
go run ./app/dev-agent-cli skill validate my-skill