semantica

Semantica

Instructions

Select the task layer first: package API, semantica CLI, standalone server/worker/explorer, MCP, or a specific optional connector. Confirm the exact extra and backend from the pinned pyproject.toml before proposing installs.
Treat commands and data flow as scoped examples; inspect installed CLI help/source for exact syntax. Build a minimal environment and add only the extra required. Avoid [all] unless user explicitly requests the broad dependency set and accepts its footprint.
Before configuring a model/provider, database, cloud service, or MCP client, obtain explicit approval for credential use, external calls, uploads, cost, data access, and configuration changes. Never ask for or print secret values; prefer documented environment/config references and local endpoints where appropriate.
Distinguish package graph-store backends (e.g. Neo4j/FalkorDB/Neptune/AGE) from the overall toolkit. Explain connector prerequisites and distinguish offline code inspection from runtime, service, or provider verification.

“Which Semantica extras do I need to ingest documents and store triples in Oxigraph?”
“Trace this Semantica CLI/MCP startup failure without using credentials.”
“Plan a minimal local pipeline for parsing and entity extraction.”

The pinned CLI is a Click-based command surface; use semantica --help or its own subcommand help rather than guessing flags.
python -m semantica.mcp_server delegates to the MCP server entrypoint, but that alone does not establish a client configuration or active server.
Read references/upstream.md before backend/provider advice; source metadata does not prove a dependency is installed or service is healthy.

references/upstream.md - supported setup/runtime entrypoints, extras, limits, licensing and safety.
references/source.json - selected pinned file SHA256s; not a whole-revision attestation.
evals/evals.json - authored routing and boundary scenarios, not executed model-quality tests.