last30days-v3-spec

Define the architectural blueprint for the last30days v3 runtime pipeline.

Updated Apr 25, 2026
One-click install
npx skills add https://github.com/bobvarkey/diabetes-buddy --skill last30days-v3-spec
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: last30days-v3-spec
Source: https://github.com/bobvarkey/diabetes-buddy/tree/main/last30days-skill/skills/last30days
Command: npx skills add https://github.com/bobvarkey/diabetes-buddy --skill last30days-v3-spec

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Defines the architectural blueprint for the last30days v3 runtime pipeline, enabling consistent implementation, onboarding, and cross-team collaboration.

Core Features & Use Cases

  • Pipeline blueprint: plan the query, retrieve per (subquery, source), normalize and dedupe, extract best snippets, fuse with weighted ranking, rerank with a single relevance score, cluster evidence, and render ranked clusters.
  • Deterministic execution: specifies required tools and environment checks to ensure reproducible runs across environments.
  • Extensibility: supports optional directories (scripts/, references/, assets/) that can be integrated as needed by the implementation team.

Quick Start

Review the spec to understand the pipeline stages and prerequisites, then locate the scripts and environment setup in the repository to begin implementing the v3 runtime.

Frequently Asked Questions about last30days-v3-spec

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a reproducible runtime pipeline for query planning and source retrieval?

A reproducible runtime pipeline requires a strict architectural blueprint specifying modular steps, data formats, and environment checks to ensure deterministic behavior across query planning, source retrieval, and ranking. This spec defines that layered approach for consistent implementation.

What is the best way to structure a multi-stage retrieval pipeline for evidence extraction and ranking?

The best way to structure a retrieval pipeline is through sequential stages: plan queries, retrieve per subquery and source, normalize and dedupe, extract best snippets, fuse with weighted ranking, rerank with a relevance score, cluster evidence, and render ranked clusters.

How do I ensure deterministic execution across environments when running a runtime pipeline?

To ensure deterministic execution across environments, the pipeline specification mandates required tools and environment checks that must pass before processing begins, guaranteeing reproducible runs and clear error reporting.

Can I extend the runtime pipeline architecture with custom scripts and reference assets?

Yes, the architecture supports extensibility through optional directories including scripts, references, and assets that implementation teams can integrate as needed without altering the core deterministic pipeline stages.

Why does my query planning and deduplication pipeline produce inconsistent results across runs?

Inconsistent results often stem from missing environment checks and unspecified data formats. The spec addresses this by enforcing strict modular steps, format definitions, and runtime prerequisites to guarantee deterministic behavior.

Do I need any external dependencies to implement the last30days v3 spec architecture?

No external dependencies are required to implement the architecture. The spec is self-contained, defining all pipeline stages, data formats, and environment checks internally to guide implementation teams without third-party constraints.