Glossary

Purpose#

This glossary explains recurring technical terms in practical terms: what they are, and why they matter for Fred.

Terms#

Temporal#

What it is
A durable workflow orchestration engine for long-running and stateful tasks, with retries, recovery, and resumability.

Why it matters for Fred
Temporal is already in production for ingestion pipelines. The roadmap extends this foundation to deep-agent execution and long-running business tasks so workflows stay robust across restarts, transient failures, and human approval pauses.

ClickHouse#

What it is
A high-performance analytical database optimized for large-scale, low-latency queries over time-series and event-like data.

Why it matters for Fred
ClickHouse gives Fred a fast analytics store for benchmark, KPI, and data-oriented use cases. It also provides a roadmap path to replace parts currently backed by OpenSearch for high-growth historical data (for example session history and KPIs) when scale requires it.

OpenSearch#

What it is
A search and analytics engine used in Fred for retrieval and operational storage paths.

Why it matters for Fred
OpenSearch remains important today, especially for search use cases. The roadmap evaluates where ClickHouse is a better fit for heavy analytics and long-retention operational datasets.

KPI#

What it is
Key Performance Indicators: operational metrics such as latency, throughput, error rate, and resource usage.

Why it matters for Fred
KPI visibility is required to detect regressions early, compare builds, and guide platform optimization decisions.

Benchmarking#

What it is
Repeatable performance and quality test scenarios executed across versions or environments.

Why it matters for Fred
Benchmarking turns performance claims into measurable evidence and helps teams prioritize real bottlenecks.

MCP (Model Context Protocol)#

What it is
A protocol used to expose tools and capabilities to agents in a consistent way.

Why it matters for Fred
MCP is the integration backbone for external capabilities. Gateway-based MCP patterns are being prototyped to simplify operations and make integrations easier to scale.

Model Drift#

What it is
A change over time in model behavior (quality, style, reliability, latency, or cost) compared to a known baseline.

Why it matters for Fred
Model drift monitoring helps teams detect silent regressions after model/provider changes.

Embedding Quality#

What it is
How well embedding vectors preserve semantic meaning for retrieval and ranking tasks.

Why it matters for Fred
Embedding quality directly affects relevance of RAG answers, source precision, and user trust.

Multimodal Agent#

What it is
An agent able to handle multiple modalities (text, documents, images, tabular data, and potentially others).

Why it matters for Fred
Multimodal support requires strong configuration contracts, because model/tool compatibility and routing rules become more complex.