Skip to main content

Wiki-Links

💡TL;DR

Process file (add links) asks your configured LLM to identify concepts and add Obsidian wiki-links. By default it writes a separate _processed.md file and preserves the source. Optional concept files require a selected concept folder. Inspect the output before expanding to a folder.

Overview​

Use this task to propose links from the meaning of a note. Model-selected terms can differ between runs and can name concepts that do not already have files. A resulting wiki-link is a navigation target, not evidence that two concepts are equivalent or that a generated explanation is correct.

How It Works​

Process​

The task chunks the note, sends each chunk to the selected provider/model, combines the responses and writes the processed output. It can create or update concept notes in the configured concept folder. Check the progress report for partial failures and the actual destination path.

Example​

Source:

Machine learning models use neural networks to learn patterns from examples.

A possible processed output:

[[Machine learning]] models use [[neural networks]] to learn patterns from examples.

This is an illustration. The exact linked terms depend on the source, model and prompt.

Usage​

Open a copied Markdown note and run Process file (add links) from the sidebar or command palette. With default settings, inspect its _processed.md sibling. Custom suffixes, processed folders and move-original settings can change the destination behavior.

Batch: Process Multiple Notes​

Select Process folder (add links), confirm the folder and scope, and begin with a small sample. There is no universal overwrite/skip switch shared by all folder tasks. See batch processing before repeating a run.

Selective: Link Specific Text​

Use Create wiki link and generate note from selection when you want to choose a term yourself. It is a separate action, not an instruction for whole-file add-links to process only the selection. Review the created concept note and any generated content.

Choosing A Linking Approach​

Choose model-assisted linking when discovering candidate concepts from note content is useful. Choose deterministic title matching when reproducibility against an existing set of note names is the priority. Evaluate other plugins against their current documentation; this guide does not claim a feature or performance comparison without a dated benchmark.

Features​

Synonym Suppression​

replaceSynonymsDuringConceptExtraction adds an instruction to avoid synonymous or closely overlapping concepts. It is a Boolean setting, disabled by default, rather than a numerical similarity threshold. Review abbreviations and near-synonyms yourself.

Obsidian owns internal link updates on rename, according to its link-update settings and the operation used. Deleting a target can leave an unresolved wiki-link; that is different from an unlinked mention. Check ambiguous basenames when similarly named notes exist in different folders.

Pure Extraction Mode​

Use Extract concepts (current file) when you want concept files without rewriting the source note. Its minimal-template and backlink controls differ from add-links. See concept notes.

Concept Note Generation​

Automatic Creation​

When concept-folder output is enabled and its path is configured, add-links can create notes containing a title and Linked From backlink. It does not automatically produce a researched description, a related-concepts section or a configurable frontmatter template. Run title generation separately if you want fuller content.

Configuration​

Search settings for the concept folder, processed-file folder, add-links suffix and move-original option. The initial concept folder is empty. Keep output paths distinct from your source collection while testing.

Advanced Options​

Context And Prompts​

chunkWordCount controls approximate chunk size. Focused learning and the custom add-links prompt can narrow the desired concepts. There are no documented sentence/paragraph context selectors, minimum-occurrence thresholds or general concept-template variables in this task.

Tips & Best Practices​

  • Compare the processed output with its source, including quotations, formulas and code.
  • Review whether newly linked terms deserve concept notes.
  • Test prompt changes on a stable sample before processing a folder.
  • Keep a backup before enabling source-moving behavior or rerunning into existing destinations.

Performance​

Latency and cost depend on the provider, model, prompt size, output length and retries. This guide makes no fixed seconds-per-note or cost-per-100-notes promise. Measure a representative sample and check the provider's actual usage before increasing scope.

Troubleshooting​

Inspect the request outcome and model response, confirm the intended note was processed, and check the reported output file. There is no guaranteed minimum word count for useful links.

Narrow the prompt or enable focused learning and review the result. Synonym suppression can help guide a model, but does not enforce a vocabulary.

Check Obsidian's link-update setting and whether the link was already ambiguous. Restore or repair the target deliberately; rerunning an LLM task is not a deterministic link-repair operation.

Next Steps​