Niplex Research Brain

approved

by Niplex Foundation

Privacy-first vault connection index with optional Gemini embeddings and a focused research map. - This plugin has not been manually reviewed by Obsidian staff.

49 downloadsUpdated 25d agoMIT

Niplex Research Brain

Niplex Research Brain is an independently distributed, mobile-compatible Niplex Research AI extension for Obsidian. It builds a user-controlled connection index for selected Markdown folders and can contribute a bounded, provenance-labelled research neighborhood to the Niplex agent when the user connects a compatible ecosystem host and grants permission.

The repository, manifest, release history, and Community entry remain separate from Niplex Research AI. Without the compatible host, the Brain remains useful as a local Research Map; it never pretends that a host connection exists.

What it adds to Niplex Research AI

When connected to a compatible host, Research Brain registers the niplex-ecosystem protocol and exposes two optional capabilities: bounded research context and a read-only action for opening a focused map. The host requests context only for a particular agent turn, applies a character and item budget, filters the declared data classes, and shows the contribution as optional extension context. The Brain does not import the core plugin or depend on its private implementation.

The first ecosystem contract permits the Brain to share only the classes the user enables in the core settings. Note metadata may include bounded note titles or paths; map provenance may include relation labels such as explicit links, backlinks, shared tags, shared MOCs, or semantic similarity. The extension never sends whole note bodies through the bridge, never receives provider credentials, and never gains write authority.

Privacy and indexing boundary

This plugin does not train a private neural network. It builds a local index and, only when the user enables it, requests bounded note previews from Gemini to generate semantic vectors. The index is not a whole-vault upload and the plugin does not write notes automatically.

An empty scope indexes nothing. Use * only when you explicitly want the entire vault. .obsidian/ and NIPLEX-OBSIDIAN/ are excluded by default and can be supplemented with additional exclusions.

Features

FeatureBehavior
Local connectionsReads Markdown metadata, headings, tags, links, and bounded previews.
Optional Research AI extensionRegisters with a compatible Niplex ecosystem host only when available; otherwise remains local.
Optional semantic vectorsUses Gemini embeddings only after the user enables the mode and supplies a Gemini key through Obsidian SecretStorage.
Incremental refreshReuses unchanged records by content hash and only re-embeds changed notes.
Refresh scheduleManual, on plugin open, every 3 days, weekly, or monthly while the app is running. Mobile background execution is not assumed.
Research MapShows a focused question neighborhood with provenance-labelled edges.
Delete controlDeletes the visible local index from NIPLEX-OBSIDIAN/Brain/.
Mobile supportUses a responsive list-and-card view rather than requiring a desktop-only canvas.

Connection behavior

Open the Brain view to see whether Niplex Research AI is connected. A connected status does not grant data access by itself. Open the host’s Niplex ecosystem settings panel and enable only the capabilities you want. The default is deny-all. Disabling the Brain also prevents it from contributing context or opening an ecosystem map.

The host uses a versioned in-process API discovered by plugin ID. If the host is missing, disabled, incompatible, or times out, the Brain continues locally and the agent continues without Brain context. No background inference is started because the user is inactive.

Installation and development

Install Niplex Research AI and Niplex Research Brain independently from Obsidian Community when the compatible releases are available. Enable the Brain only if you want its local index and optional host contribution.

npm install
npm test
npm run build
npm run lint

License

MIT.

For plugin developers

Search results and similarity scores are powered by semantic analysis of your plugin's README. If your plugin isn't appearing for searches you'd expect, try updating your README to clearly describe your plugin's purpose, features, and use cases.