skills.rest
junghan0611
junghan0611COMMUNITY
Explore all skills by this author →
Website𝕏@junghanacsGitHub

semantic-memory

Retrieve past sessions and notes via hybrid semantic search across pi, Claude Code, and Denote.

6|Updated Nov 14, 2025
Search Tags:#embeddings#org-mode#retrieval#memory#multilingual#semantic-search
One-click install
npx skills add https://github.com/junghan0611/agent-config --skill semantic-memory-junghan0611
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: semantic-memory
Source: https://github.com/junghan0611/agent-config/tree/main/skills/semantic-memory
Command: npx skills add https://github.com/junghan0611/agent-config --skill semantic-memory-junghan0611

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Semantic memory helps AI agents recall past conversations and knowledge bases, enabling continuity across sessions and projects.

Core Features & Use Cases

  • Hybrid retrieval: vector similarity + full-text search across pi, Claude Code, and Denote org-mode notes.
  • Local embeddings with Ollama/vLLM and LanceDB for fast, private indexing.
  • Korean morphology and cross-lingual expansion via Kiwi and dictcli for Korean↔English queries.

Quick Start

Search past sessions and org-mode notes by meaning to surface relevant context.

Frequently Asked Questions about semantic-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I retrieve past AI conversation sessions by meaning instead of keywords?▼

Semantic memory locates past sessions by employing vector similarity and full-text search across pi, Claude Code conversations, and Denote org-mode notes to surface relevant context.

Can I index org-mode knowledge chunks using local embeddings for private retrieval?▼

Yes, semantic memory uses local embeddings via Ollama or vLLM and stores them in LanceDB to enable fast, private hybrid retrieval of org-mode knowledge chunks.

Does semantic search work for Korean and English cross-lingual queries?▼

Cross-lingual semantic search works by applying Korean morphology via Kiwi and dictcli to expand Korean↔English queries, ensuring relevant notes are located across both languages.

What is the best way to search past Claude Code and pi conversations with vector similarity?▼

The best way to search past conversations is using this skill's LanceDB hybrid retrieval, which combines vector similarity with full-text search across pi and Claude Code session histories.

Do I need Ollama or vLLM to run semantic memory for org-mode notes?▼

Yes, you need a local embedding server like Ollama or vLLM to generate the vector embeddings required by LanceDB for indexing and retrieving your org-mode knowledge base.

Why use LanceDB hybrid retrieval instead of standard vector search for past notes?▼

LanceDB hybrid retrieval is used because it combines vector similarity with full-text search, improving accuracy when locating relevant context across multilingual org-mode notes and past sessions.

Related Skills

Explore More Software Engineering Skills→
junghan0611junghan0611
6

denotecli

Search and retrieve Denote/org-mode notes with JSON output.

Community
Advanced
zjykzkzjykzk

pi-history-ingest

Ingest Pi session history into an Obsidian wiki while redacting sensitive data.

Community
Advanced
RepairYourTechRepairYourTech
6

lancedb

Implement LanceDB vector databases for RAG pipelines and semantic search.

Community
Advanced
an8079an8079
4

vector-search

Implement vector search with FAISS, Milvus, Qdrant, and Weaviate databases.

Community
Advanced