32 lines
572 B
Markdown
32 lines
572 B
Markdown
---
|
|
title: "Semantic Search Research"
|
|
date: "10-11-2025"
|
|
last_modified: "10-11-2025:19:21"
|
|
tags: ["research", "ai", "search"]
|
|
---
|
|
|
|
# Semantic Search
|
|
|
|
## Current Search
|
|
|
|
Keyword-based with scoring.
|
|
|
|
## Semantic Search
|
|
|
|
Use embeddings for similarity:
|
|
- OpenAI embeddings API
|
|
- Local models (sentence-transformers)
|
|
- Vector database (Pinecone, Weaviate)
|
|
|
|
## Implementation
|
|
|
|
1. Generate embeddings for all notes
|
|
2. Store in vector DB
|
|
3. Query with user search
|
|
4. Return top-k similar
|
|
|
|
## Cost Analysis
|
|
|
|
OpenAI: /tmp/generate_notes.sh.0001 per 1K tokens
|
|
Local: Free but slower
|