TotalApp Docs
Ctrl K
Home
EN | TR
Go to App

Semantic Search

How TotalApp ranks and auto-attaches relevant sources from My Knowledge across every Writer Tool screen.

Overview

Semantic Search is the feature behind TotalApp's AI-powered tools that lets them find and surface relevant material from your My Knowledge library automatically, instead of relying purely on the AI model's general knowledge. Rather than showing your library in a flat, unordered list, it ranks sources by how closely their content relates to what you're currently working on — using text embeddings and cosine similarity, not keyword matching.

This is not full Retrieval-Augmented Generation (RAG) — it does not chunk or index documents into a vector database. It's a lightweight ranking layer: your whole library is embedded, your current input is embedded, and the two are compared directly. For most personal knowledge libraries (tens to low hundreds of sources), this is fast enough to run live as you type.

Today, every Writer Tool screen (Blog Writer, Essay Writer, Article Writer, Storyteller, Anime Writer, Scenario Writer, Smart Rewriter) uses Semantic Search via its Attach Knowledge button — this is simply the first place it's wired up. As more AI tools are added across TotalApp, Semantic Search is designed to power source discovery in those places too.

What Semantic Search Does

Relevance Ranking

As you type, your knowledge sources are re-ranked in real time so the most relevant ones surface at the top of the picker.

Relevance Score Badge

Each ranked source shows a percentage badge in the picker — either raw cosine similarity or a normalized 0-100% spread, depending on your display preference.

Auto-Attach (Automatic Mode)

Optionally, the top-ranked sources can be attached automatically as you type, with a "N sources auto-attached" indicator — no need to open the picker at all.

Graceful Degradation

If no embedding model is configured, every screen keeps working exactly as before — sources just appear in their original, unranked order.

Semantic Search in Writer Tools

As an example of Semantic Search in practice: every Writer Tool screen (Blog Writer, Essay Writer, Article Writer, Storyteller, Anime Writer, Scenario Writer, Smart Rewriter) has an Attach Knowledge button above its main text field. As you type your topic, description, or source text, your knowledge library is ranked by relevance and the best matches surface first.

The Three Building Blocks

  1. Embedding Model — a local Ollama model that converts text into a numeric vector. See Embedding Model for setup.
  2. Search Mode — Assisted (you confirm attachments manually) or Automatic (top matches auto-attach). Covered in Using Semantic Search.
  3. Score Display — how relevance percentages are presented in the Attach Knowledge picker: raw cosine similarity or normalized. Also covered in Using Semantic Search.

Where It's Configured

All Semantic Search settings live in Settings → Agentic, in the Embedding Engine card — completely independent of which Writer Engine (API, Local CLI, Local LLM, Ollama) you've selected for text generation itself.

FAQ

Does Semantic Search send my documents anywhere?
No. Embeddings are computed by a local Ollama instance running on your own machine — your knowledge library never leaves your computer for the purpose of ranking.
Is Semantic Search required to use My Knowledge?
No. You can add, preview, and manually attach knowledge sources with zero embedding configuration. Semantic Search only adds ranking and auto-attach on top of that.
Which parts of TotalApp support Semantic Search today?
Every Writer Tool screen with an Attach Knowledge button: Blog Writer, Essay Writer, Article Writer, Storyteller, Anime Writer, Scenario Writer, and Smart Rewriter. This is expected to expand to other AI-powered tools over time.