setup

setup command

Prefer pali init for installed binaries. go run ./cmd/setup remains the source-checkout equivalent and prepares local project bootstrap state:

• creates required scaffold directories
• creates the target config file from pali.yaml.example when missing
• defaults to a zero-dependency lexical config so first boot does not require Ollama, ONNX, Qdrant, or Neo4j
• downloads missing ONNX assets when embedding.provider=onnx (or when forced with -download-model):
  • models/all-MiniLM-L6-v2/model.onnx
  • models/all-MiniLM-L6-v2/tokenizer.json
• checks ONNX Runtime shared library availability and prints OS-specific install hints when missing
• checks Ollama server/model readiness when the selected config enables an Ollama-backed embedder, parser, or scorer (/api/version, /api/tags)
  • uses embedding.ollama_base_url and embedding.ollama_model from the selected config file by default

It is safe to run multiple times.

Canonical config reference: docs/configuration.md.

Flags:

• -config: config file path to create/read during setup (default pali.yaml)
• -skip-model-download: do not download model files
• -download-model: force ONNX model download even if provider is not onnx
• -skip-runtime-check: skip ONNX Runtime shared library check
• -skip-ollama-check: skip Ollama readiness checks
• -ollama-base-url: override Ollama URL for checks (default from config)
• -ollama-model: override Ollama model for checks (default from config)
• -model-id: override Hugging Face model id (default sentence-transformers/all-MiniLM-L6-v2)