Domain-Specific Embeddings Plugin

Detects the knowledge domain of your memories and recommends optimized embedding models for each domain.

Category

AI Enhancement

Domains

Domain	Icon	Keywords	Example Models
General	Box	—	text-embedding-3-small, nomic-embed-text
Code	Code	function, class, import, API...	nomic-embed-text, text-embedding-3-large
Medical	Heart	patient, diagnosis, treatment...	PubMedBERT, text-embedding-3-large
Legal	Scale	contract, clause, jurisdiction...	nomic-embed-text, text-embedding-3-large
Scientific	FlaskConical	hypothesis, experiment, p-value...	snowflake-arctic-embed, text-embedding-3-large
Financial	TrendingUp	revenue, EBITDA, portfolio...	text-embedding-3-small, nomic-embed-text

API

GET /api/v1/plugins/domain-embeddings

Action	Params	Description
config	—	Domain profiles + provider status
stats	—	Domain distribution of memories
detect	text (required)	Detect domain for given text
models	domain (required)	Available models for a domain

POST /api/v1/plugins/domain-embeddings

Action	Body	Description
save-config	domainModels, autoDetect, etc.	Save domain model configuration
tag-domain	memoryId, domain	Tag a memory with a domain
batch-detect	batchSize (default: 100)	Auto-detect domains for untagged

Detection

• Keyword-based heuristic matching against domain keyword lists
• Score = matched keywords / total domain keywords
• Threshold: 5% match required to assign a domain
• Falls back to "general" for generic content

Dependencies

• Provider availability checked dynamically (OpenAI, Gemini, Ollama)
• No external API calls for detection (pure keyword matching)