dsh-semantic-search

CLI usage

sema index                build the full index from the workspace
sema reindex [--full]     incremental refresh (or full rebuild with --full)
sema search     hybrid vector + BM25 search, prints top hits
sema stats [--json]       index health, provider, and sizing numbers

Global options:

--root           workspace root (default: current directory)
--data-dir       index storage directory (default: /.sema)
--provider      embedding provider: lexical | openai (default: lexical)
--dim              embedding dimension (lexical default: 4096; openai 0 = auto)
--base-url       OpenAI-compatible embeddings endpoint base URL
--model         embeddings model name (openai only)
--api-key        API key (openai only; env: SEMA_EMBEDDING_API_KEY)
--top              hits to print for search (default: 20)
--json                machine-readable output where supported
--help                show this help

CLI exit codes

• 0 — success (including a search with zero hits and a --version/--help call).
• 1 — a runtime failure (config error, build/index/search error).
• 2 — a usage error: unknown command, unknown flag, or a missing query.

Configuration

The plugin is configurable through the bundle row's config (see cordis.patch.yml for an example), the CLI flags above, or defaults in code:

Option	Default	Meaning
root	cwd	workspace root to index
dataDir	.sema	index storage directory
provider.kind	lexical	lexical (offline) or openai
provider.dimension	0 (lexical: 4096)	embedding dimension; 0 = auto-infer from the endpoint
provider.baseUrl	https://api.openai.com/v1	OpenAI-compatible endpoint root
provider.model	text-embedding-3-small	embedding model name
provider.apiKeyEnv	SEMA_EMBEDDING_API_KEY	env var holding the API key
allowFallback	true	fall back to the lexical provider when a remote one fails
include / ignore	defaults	glob sets of files to index / skip
maxLinesPerChunk	80	hard chunk size upper bound
nGram	2	CJK n-gram size (1 disables n-gramming)
topK	20	hits returned by default
rrfK	60	RRF fusion constant
vectorK	300	candidates per channel before fusion
autosave	true	persist the index after builds
autoIndex	true	build lazily on first search
watch	true	watch the workspace for changes