agentic-systems-minilm

Small sentence embedding model for semantic search over agentic / RAG / LLMOps project docs.

Comparison note: This is an embeddings model (MiniLM, 384-d), not a generative 7B demo and not a Gradio chat Space. It is meant for retrieval and clustering on CPU.

Model details

Base	sentence-transformers/all-MiniLM-L6-v2
Training	1 epoch CosineSimilarityLoss on CPU; pairs from hharsha/agentic-systems-showcase plus short synthetic Q/A about AgentFleet, RetrievalLab, AgentOps Studio, Vibespace, Agent OS, CareerAgent, Control Tower, agentgrid
Intended use	Semantic search / clustering of short texts about agentic systems, RAG pipelines, MCP tools, and related portfolio docs
Not intended for	Open-domain chat, replacing large embedding models on broad corpora, medical/legal advice
License	Apache-2.0 (base); training data MIT showcase

Light domain adaptation only — weights start from MiniLM-L6-v2.

Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("hharsha/agentic-systems-minilm")
emb = model.encode(["hybrid search RAG with citations", "AgentFleet multi-agent ops"])
print(emb.shape)  # (2, 384)

Related

• Tag generator (text2text): hharsha/agentic-github-tagger
• LoRA adapter (generative tiny): hharsha/agentic-rag-lora
• Dataset: hharsha/agentic-systems-showcase
• Studio: https://agentic-systems-studio.com
• GitHub: https://github.com/hharsha98

How it was trained

Base: sentence-transformers/all-MiniLM-L6-v2
Loss: CosineSimilarityLoss
Epochs: 1 (CPU)
Batch size: 8
Data: name/summary pairs from the showcase dataset + synthetic agentic/RAG pairs