wiki_semantic_linklisted

# LLM Wiki — Semantic Linker Skill (wiki_semantic_link) > **CLI (read first):** This skill drives the `magi` CLI (MAGI research workspace tool, assumed installed on PATH). If unsure of your surroundings, run `magi sync` first to locate the workspace. For the full syntax of any command: `magi <command> --help`. This skill scans all markdown files within the `wiki/concepts/` directory, extracts their text, and generates embeddings using a local Ollama model (configured in `config.yaml`, default: `qwen3-embedding:0.6b`). It then calculates pairwise cosine similarity between all concepts and automatically injects bi-directional Obsidian-style links (`[[Concept Name]]`) for pairs that exceed a given similarity threshold. ## Execution This skill is executed via the `magi link` command. **Crucial Parameter Selection**: Depending on the objective, you MUST use the correct parameters: 1. **Normal Link Injection**: If the user just wants to inject semantic links (default behavior): `magi link <TOPIC_DIR>` Model, threshold, and Ollama URL are read from `config.yaml`. Override via CLI flags if needed (e.g. `--threshold 0.80 --model <model>`). 2. **Deduplication Scan (Auto-Merge)**: If the user is running the `wiki_concept_sync` pipeline to merge duplicates: `magi link <TOPIC_DIR> --dedup-only --auto-merge` > **Prerequisites**: Ensure the embedding model is available locally via `ollama pull <model>` (check `config.yaml` for the configured model name). Requires `numpy` and