Under the Hood
The Technology Stack
How Spinacle builds production AI — the tools, protocols, and architecture behind our agentic systems and LLM integrations.
Stack
What We Build With
LLM Layer
GPT-4o, Claude Sonnet, Gemini 1.5 Pro — we work across all major frontier models and help clients pick the right one for their use case.
Orchestration
LangChain, LlamaIndex, custom agentic frameworks. We build deterministic pipelines, not prompt-and-pray systems.
RAG & Vector
Pinecone, Weaviate, pgvector. Retrieval architectures tuned for accuracy, latency, and cost at production scale.
Data & Pipelines
dbt, Airflow, custom ETL. We build the data layer that most AI projects skip — and then wonder why the model fails.
Deployment
AWS, GCP, Azure. Containerised, monitored, observable. AI that stays reliable after day one.
Myve Protocols
SUE & PAO — Spinacle's proprietary intent-capture and agent-routing protocols powering the myve.ai network.
Proprietary Protocols
The protocols behind myve.ai
SUE
Semantic User Expression
A proprietary intent-capture protocol that maps natural-language user requests to structured service graph queries, enabling precision routing in agentic commerce networks.
PAO
Predictive Agent Orchestration
An agent-routing layer that dynamically assigns tasks to the most cost-effective and capable sub-agent, balancing latency, accuracy, and per-token cost in real time.