Source Map

This page maps external references to this book’s chapters. Use it as a reading guide, not as a replacement for the book’s pattern language.

The sources repeat several core ideas: start simple, separate workflows from agents, use tools through typed contracts, add memory deliberately, evaluate behavior, and reserve multi-agent systems for cases where specialization justifies orchestration cost.

How To Use This Map

Use this page when you need to answer one of three questions:

For normal reading, do not start here. Start with How To Read This Book, then return to this map when you want source context or coverage checks.

Read the map as an editorial filter. A source earns space in the book when it changes a reader’s design decision, exposes a missing failure mode, or gives stronger evidence for a chapter claim.

Source-Backed Decision Shortcuts

Use these shortcuts when a source names a pattern but the engineering decision is still unclear.

Evidence Tiers

Not every source has the same editorial weight. Use this order when sources disagree:

This prevents the online book from drifting toward whatever source has the longest pattern list. The book should privilege engineering evidence over novelty.

Primary References

Secondary References

Sources that were not publicly accessible or only served as discovery routes were excluded from the map.

Local Pattern Repository Coverage

A local flattened digest of the awesome-agentic-patterns project was used as a coverage checklist. Its patterns were not copied into this book. They helped identify missing or underdeveloped areas that now have authored treatment here.

The 2026-06-20 intake scan extracted 167 pattern records across orchestration and control, tool use and environment, reliability and evaluation, context and memory, UX and collaboration, security and safety, feedback loops, and learning and adaptation. The online book should use this catalog as an expansion queue for chapter quality and cross-links, not as a replacement taxonomy.

This local source is best treated as a checklist, not as the book’s taxonomy. Some entries are production-proven, some are emerging, and some are research or product-specific. The book should keep absorbing the useful engineering ideas while preserving its own structure: boundaries first, then patterns, then evaluation and operations.

Local PDF References Reviewed

The following PDFs were reviewed or intake-scanned locally as source-of-knowledge inputs. They are not included in this repository, linked from the book, or reproduced. They inform coverage checks, chapter structure, and topic prioritization only.

Pattern Coverage Map

Maintenance Checklist

Update this page when a source changes the book’s structure, not every time a new link appears.

Before adding a source, check:

• Does it introduce a pattern, failure mode, runtime concern, or implementation technique not already covered?
• Is it stronger than an existing source for the same claim?
• Does it map to at least one chapter where a reader can act on the idea?
• Is it public enough for an online GitHub Pages reader to inspect?
• Does it create a concrete editorial task, such as a missing example, checklist, diagram, lab, or cross-link?

If the answer is mostly no, leave the source out. A shorter map with clear editorial purpose is more useful than a large reference pile.

Editorial Rule

External sources are used to validate coverage and expose missing patterns. The book keeps its own taxonomy:

• foundations first;
• control loops second;
• memory and knowledge as a separate concern;
• tools, skills, and protocols as integration boundaries;
• multi-agent systems only when specialization has value;
• production runtime patterns for safety, evaluation, and operations;
• source-informed pattern selection as the entry point for choosing the right design.

This keeps the book useful as a reference instead of turning it into a link dump.