patterns: expose composition helpers, add approve/aggregate hooks, ad…

[Koukyosyumei]

…d from-scratch pattern example

The prebuilt patterns stay fork-friendly, but the pieces they are built from are now public API: merge_reviews, review_cycle, verify_and_judge, ask_with_valid_citations, render_evidence, mean_score_verdict and smaller_diff. ensemble(approve=...) swaps the per-review approval convention; judge_panel(aggregate=...) swaps the ballot aggregation, as its docstring already promised. examples/quorum_ensemble.py shows the anatomy of writing a pattern from scratch (majority quorum instead of unanimity — the case the kwargs deliberately don't cover).

[Koukyosyumei]

🪙 100% AI-authored · 10 files · prompt 46/100

1 AI commit

Important

🎯 Goal: Generalize patterns.py: expose private helpers, add approve/aggregate kwargs, add from-scratch pattern example

1 / 1	10	1 : 1.7
AI commits	files touched	READ : EDIT

💬 The ask

"Do you think the current SDK is well generalized framework that allows users to define various strategies? For example, while this library has patterns, i guess it's important to let users define their own pattern easily. Sure, let's do …"

📍 What shipped

• ✓ patterns: expose composition helpers, add approve/aggregate hooks, add from-scratch pattern example

Note

🪴 Prompt maturity: 46/100 · developing · 1 prompt scored (100% of AI commits)
🔧 Recurring weak spots: no acceptance criteria, weak context.
Heuristic signal of prompt craft — not a developer rating.

📊 heuristic breakdown

Dimension	Signal
Objective (core)	███████░░░ 0.69
Grounding (core)	█████░░░░░ 0.51
Direction / acceptance (core)	░░░░░░░░░░ 0.00
Context	███░░░░░░░ 0.25
Examples	███░░░░░░░ 0.30
Structure	██░░░░░░░░ 0.17
Lexical diversity	█████████░ 0.92
Clarity (readability band)	██████████ 1.00
Length adequacy	██████████ 1.00
Evidence attached (bonus)	░░░░░░░░░░ 0.00

Readability (display-only): Flesch 57 · FK grade 9.8 · Gunning Fog 13.2

Offline classical-NLP score — no LLM. Higher = more specific, bounded, and grounded prompting.

Tip

✅ Security scan clean

No credentials leaked in this branch's diff. h5i scanned every added line against the secret rule pack and found nothing to rotate.

Note

Duplicate-code scan clean — no copy-paste blocks introduced.

🧠 Reasoning by file — 235 nodes across 5 lanes, latest 24 only

%%{init: {"flowchart": {"nodeSpacing": 42, "rankSpacing": 48, "diagramPadding": 14}, "themeVariables": {"fontSize": "18px"}} }%%
flowchart TB
  subgraph lane_0["📄 src/h5i/orchestra/patterns.py · 12 ops"]
    direction LR
    n_lane_0_0["EDIT × 12"]:::a
  end
  subgraph lane_1["📄 tests/test_patterns.py · 3 ops"]
    direction LR
    n_lane_1_0["READ"]:::o
    n_lane_1_1["EDIT × 2"]:::a
    n_lane_1_0 --> n_lane_1_1
  end
  subgraph lane_2["📄 src/h5i/orchestra/_conductor.py · 2 ops"]
    direction LR
    n_lane_2_0["READ × 2"]:::o
  end
  subgraph lane_3["📄 src/h5i/orchestra/__init__.py · 1 op"]
    direction LR
    n_lane_3_0["READ"]:::o
  end
  subgraph lane_4["🗂 Other · 6 nodes"]
    direction LR
    n_lane_4_0["READ · read src/h5i/orchestra/policy.py"]:::o
    n_lane_4_1["READ · read examples/custom_control_flow.py"]:::o
    n_lane_4_2["READ · read examples/review_escalation.py"]:::o
    n_lane_4_3["READ · read h5i-python-sdk-layout.md"]:::o
    n_lane_4_4["READ · read tests/test_examples.py"]:::o
    n_lane_4_5["EDIT · wrote examples/quorum_ensemble.py"]:::a
    n_lane_4_0 --> n_lane_4_1
    n_lane_4_1 --> n_lane_4_2
    n_lane_4_2 --> n_lane_4_3
    n_lane_4_3 --> n_lane_4_4
    n_lane_4_4 --> n_lane_4_5
  end
  classDef o fill:#dbeafe,stroke:#1e3a8a,color:#0b1c4a;
  classDef t fill:#fef3c7,stroke:#92400e,color:#3f2d05;
  classDef a fill:#dcfce7,stroke:#166534,color:#0a2e16;
  classDef n fill:#ede9fe,stroke:#5b21b6,color:#221251;
  classDef m fill:#e5e7eb,stroke:#374151,color:#0b0f17;

Loading

📜 Per-commit provenance

7fbf899f patterns: expose composition helpers, add approve/aggregate hooks, add from-scratch pattern example

• prompt — Do you think the current SDK is well generalized framework that allows users to define various strategies? For example, while this library has patterns, i guess it's important to let users define their own pattern easily. Sure, let's do "expose the private helpers, add predica…
  • 🪴 prompt maturity 34/100 developing — no acceptance criteria, weak context
• model claude-fable-5 · agent claude-code

---

_{Generated by h5i · re-run h5i share pr post to refresh.}