Feature Request: Local Semantic Search via Vector Index Add-on

[ChrisRudi]

Feature Request: Local Semantic Search

Part-DB already provides excellent structured search capabilities. However, finding parts often requires knowing the exact part number, category, or keyword.

A local semantic search could make large part collections much easier to navigate.

The idea is simple:

• Generate a searchable text from existing fields
• Create an embedding for each part
• Store the embedding directly in the database (e.g. pgvector)
• Extend the search with an optional semantic ranking

Everything can run locally:

• No cloud services
• No external APIs
• No LLM required
• No data leaves the server
• No subscription costs

Benefits:

• Find parts by function instead of exact keywords
• Detect potential duplicates
• Discover similar components
• Search manufacturer descriptions more effectively
• Better experience for large databases

Examples:

• "quad nand gate" → 74HC00
• "audio opamp" → NE5532
• "schmitt trigger inverter" → 74HC14

This would not replace the existing search but complement it. Users who do not want semantic search could simply disable it.

Because embeddings can be generated locally, this feature would fit well with the self-hosted and privacy-friendly philosophy of Part-DB.