Skip to content

Add Parallelization, Tree of Thoughts, and Agentic RAG (16-18 of 18)#7

Merged
Hefrock merged 2 commits into
mainfrom
claude/intelligent-gates-2s4z93
Jun 20, 2026
Merged

Add Parallelization, Tree of Thoughts, and Agentic RAG (16-18 of 18)#7
Hefrock merged 2 commits into
mainfrom
claude/intelligent-gates-2s4z93

Conversation

@Hefrock

@Hefrock Hefrock commented Jun 20, 2026

Copy link
Copy Markdown
Owner

Summary

  • Adds the final three patterns to bring the reference from 15 to the planned 18: Parallelization (Single-agent — stateless), Tree of Thoughts (Single-agent — agentic), and Agentic RAG (Single-agent — memory), filling the gaps identified against Anthropic's "Building Effective Agents" taxonomy and the broader 2026 agent-pattern canon.
  • Updates CLAUDE.md and README.md pattern counts/lists to match.
  • Tree of Thoughts originally modeled its branching step as a kind:"branch" flow cluster, but combined with loop:true that made the loop-back arc visually collide with the cluster box (the shared renderer's loop code assumes a plain node's height, not a cluster's — no existing pattern combined the two). Reworked to a plain 4-node loop to keep the diagram correct rather than patching the shared renderer for a one-off case.

Test plan

  • npm run validate — schema OK, 18/18 patterns
  • Visually verified all three new cards: correct section placement, pill color/label, and flow diagram (including the parallel-fan-out cluster on Parallelization and the loop-back arcs on Tree of Thoughts / Agentic RAG)
  • Re-ran full export pipeline (export:images, export:pptx, export:pdf, export:svg) — all 18 patterns export cleanly in every format
  • Spot-checked dark mode and the theme-ocean variant on the new cards — no hardcoded colors were introduced, all three reuse existing cat token classes

Generated by Claude Code

claude added 2 commits June 20, 2026 02:44
Sectioning/ensemble-voting workflow per Anthropic's "Building Effective
Agents" taxonomy: fixed subtasks decided in advance, run concurrently by
a single model, then merged or majority-voted — distinct from
Orchestrator-worker's dynamically delegated specialist agents.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LLtTG22WSrRfurHKzaMc8h
Tree of Thoughts (Single-agent — agentic): branch-evaluate-backtrack
search loop, per Yao et al.'s ToT framework. Initially modeled the
branching step as a kind:"branch" cluster, but that combined with
loop:true caused the loop-back arc to visually collide with the cluster
box (the renderer's loop code assumes a plain node's height, not a
cluster's) — reworked to a plain 4-node loop to keep the diagram correct.

Agentic RAG (Single-agent — memory): adds iterative query decomposition
and a sufficiency check around the retrieve step, distinguishing it from
the single-pass RAG pattern already in the deck.

Verified schema validity (18/18) and visually confirmed section
placement, pill styling, and flow diagrams for both, across the live
page and the standalone SVG export.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LLtTG22WSrRfurHKzaMc8h
@Hefrock
Hefrock merged commit 0985bfb into main Jun 20, 2026
1 check passed
@Hefrock
Hefrock deleted the claude/intelligent-gates-2s4z93 branch June 20, 2026 03:37
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants