Skip to content

Similarity-Based Chat Organization #4

@monapdx

Description

@monapdx

Feature: Similarity-Based Chat Organization

Goal

Introduce similarity-based tooling to help users explore, group, and categorize chats more efficiently.

This will be built in three stages, moving from data foundation → scoring logic → UI features.


Implementation Plan


Build Order

1 → 2 → 3

1. Category Fingerprint Caching (Foundation)

Creates a cached vector representation (“fingerprint”) for each category.

This enables fast comparisons without needing to compare against every individual chat.


2. Similarity Scoring (Core Logic)

Uses category fingerprints to calculate similarity scores for uncategorized chats.

Displays suggestions like:

  • “Coding 82%”
  • “GitHub 74%”

This is the core logic that powers all similarity-based features.


3. Similar Chats Panel (UI Layer)

Builds on the similarity logic to show related chats for a selected conversation.

Enables exploration, grouping, and future batch actions.


Dependencies


Notes

  • All features should be local-first (no external APIs required for MVP)
  • Start with TF-IDF + cosine similarity
  • Prioritize performance and responsiveness
  • Treat similarity as suggestive, not authoritative

Future Extensions (Out of Scope for This Issue)

  • Batch categorize chats based on similarity
  • Auto-suggest category assignment actions
  • Conversation clustering / grouping view
  • Embeddings-based similarity (optional upgrade)
  • Adjustable weighting (title vs prompt vs full text)

Why This Matters

  • Makes large chat histories easier to navigate
  • Reduces manual categorization effort
  • Surfaces hidden patterns across conversations

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions