Index directories and run semantic search over documents with the cementic CLI.
npx skills add https://github.com/mnazaal/dotfiles --skill tool-cementic-mnazaal
Please help me install this Agent Skill. Skill: tool-cementic Source: https://github.com/mnazaal/dotfiles/tree/main/.agents/skills/tool-cementic Command: npx skills add https://github.com/mnazaal/dotfiles --skill tool-cementic-mnazaal
cementic start to extract, chunk, and embed documents into versioned pipeline revisions while old revisions stay searchable. - Semantic search and collection management: Query active revisions with cementic search, and promote, reindex, or remove collections and revisions. - Pipeline debugging and diagnostics: Use cementic doctor for readiness checks and pipe extract | chunk | embed together to inspect the pipeline without a database. - Use Case: Point cementic at a folder of research papers, wait for indexing to finish, then run cementic search "transformer inference" to retrieve semantically relevant passages across the whole library. ## Quick Start Ask the assistant to check cementic readiness with doctor, start Postgres if needed, then index a chosen directory into a named collection and run a semantic search query against it.High-intent search queries and answers about installing and using this skill.
Run `cementic start ~/papers --collection papers` to watch the directory and extract, chunk, and embed its documents into Postgres. Monitor per-file progress with `cementic status -c papers -v` or get machine-readable output via `cementic status --json`.
Run `cementic search "your query"` to search the active revision of the default collection, or add `-n 5 -c name` to limit results and target specific collections. Note that search only queries a collection's active revision, which requires `cementic collection promote` first.
On a fresh Postgres with no schema yet, `status` and `search` fail with `relation "..." does not exist`, which is expected behavior. Tables are only created inside `cementic start`'s worker paths, so bootstrap by starting a watch on a chosen directory and collection.
Most commands require a reachable Postgres; start it via `systemctl --user start cementic-postgres` or `podman start cementic-postgres` if a container exists. No-database commands like `chunk`, `embedding status`, and `config show` still work, and `extract | chunk | embed` pipes debug the pipeline without a database.
Use cementic for full-text semantic retrieval and indexing of document collections. For BibTeX metadata, citekeys, and library maintenance, use a bibliography-focused tool instead; combine both only when a task needs full text plus citation metadata.
yunusgungorProvisions a GBrain database and injects AGENTS.md for initial setup.
Agents4Academia-AISynthesize cited answers from local research papers and notes.
FuZhiyuRetrieve and analyze academic papers from a Zotero library with citations and BibTeX export.
BEKO2210Search knowledge bases semantically with vector embeddings and keyword fallback.