Codegeist LLM is the workspace for developing a compact language model that can
run on consumer hardware and integrate with codegeist-os.
The project will evaluate an open-source or open-weight base model and develop a reproducible path from an approved upstream source to documented, verifiable artifacts suitable for local inference. The target hardware envelope and model choice remain open decisions that must be supported by measurements.
codegeist-os is the intended consumer. This repository owns model-selection,
provenance, evaluation, and artifact requirements. The operating-system
repository owns installation, runtime isolation, user permissions, and system
integration.
- No base model, architecture, dataset, inference runtime, or quantization format has been selected.
- No model weights, datasets, generated artifacts, or access tokens belong in Git.
- Training, fine-tuning, inference implementation, model packaging, registries, and release automation are deferred.
- Source-code, model-weight, dataset, derivative, and redistribution licenses must be evaluated separately before any artifact is adopted or published.
- This repository is public through its GitHub mirror. Do not commit private planning material or content that cannot be redistributed publicly.
docs/architecture.mddefines the current repository and integration boundaries.docs/model-selection.mddefines the evidence required before choosing a base model.docs/evaluation.mddefines the initial evaluation categories and hardware measurements.docs/security.mddefines provenance, artifact, credential, and supply-chain requirements.docs/memory-bank/chat.mdrecords compact current project state.
.devcontainer/ and .opencode/ are Git submodules that track the release
branches of the shared Codegeist development and agent kits. Initialize them
from this repository with:
git submodule update --init .devcontainer .opencodeGitea at git.codegeist.ai is the primary write target. GitHub at
github.com/codegeist-ai/codegeist-llm is a public push mirror of Git refs.
Issues, pull requests, secrets, permissions, and other platform state are not
automatically synchronized.
No project or model license has been selected. Do not infer permission to use or redistribute future code, weights, datasets, or derivatives until the relevant license decision is recorded.