Deployment Boundaries¶

Deployment for agentic-proteins should respect the package boundary instead of assuming the full repository is always present.

The point of this page is to protect the idea that packages are publishable units. Even inside a monorepo, deployment assumptions should stay narrow enough that the package can still be understood and operated as its own surface.

Treat the operations pages for agentic-proteins as the package's explicit operating memory. They should make common tasks repeatable without relearning the workflow from logs or oral history.

Visual Summary¶

flowchart TB
    page["Deployment Boundaries<br/>clarifies: repeat workflows | find diagnostics | release safely"]
    classDef page fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a,stroke-width:2px;
    classDef positive fill:#dcfce7,stroke:#16a34a,color:#14532d;
    classDef caution fill:#fee2e2,stroke:#dc2626,color:#7f1d1d;
    classDef anchor fill:#ede9fe,stroke:#7c3aed,color:#4c1d95;
    classDef action fill:#fef3c7,stroke:#d97706,color:#7c2d12;
    step1["packages/agentic-proteins/pyproject.toml"]
    step1 --> page
    step2["CLI entrypoint in src/agentic_proteins/interfaces/cli.py"]
    step2 --> page
    step3["HTTP app in src/agentic_proteins/api/v1"]
    step3 --> page
    run1["tests/e2e for governed flow behavior"]
    page --> run1
    run2["tests/regression and tests/smoke for replay and storage protection"]
    page --> run2
    run3["tests/unit for api, contracts, core, interfaces, model, and runtime"]
    page --> run3
    release1["pyproject.toml"]
    run1 --> release1
    release2["README.md"]
    run2 --> release2
    release3["CHANGELOG.md"]
    run3 --> release3
    class page page;
    class step1,step2,step3 positive;
    class run1,run2,run3 anchor;
    class release1,release2,release3 action;

Boundary Facts¶

package root: packages/agentic-proteins
public metadata: packages/agentic-proteins/pyproject.toml
release notes: packages/agentic-proteins/CHANGELOG.md when present

Concrete Anchors¶

packages/agentic-proteins/pyproject.toml for package metadata
packages/agentic-proteins/README.md for local package framing
packages/agentic-proteins/tests for executable operational backstops

Use This Page When¶

you are installing, running, diagnosing, or releasing the package
you need repeatable operational anchors rather than architectural framing
you are responding to package behavior in local work, CI, or incident pressure

Decision Rule¶

Use Deployment Boundaries to decide whether a maintainer can repeat the package workflow from checked-in assets instead of memory. If a step works only because someone already knows the trick, the workflow is not documented clearly enough yet.

What This Page Answers¶

how agentic-proteins is installed, run, diagnosed, and released in practice
which checked-in files and tests anchor the operational story
where a maintainer should look first when the package behaves differently

Reviewer Lens¶

verify that setup, workflow, and release statements still match package metadata and current commands
check that operational guidance still points at real diagnostics and validation paths
confirm that maintainer advice still works under current local and CI expectations

Honesty Boundary¶

This page explains how agentic-proteins is expected to be operated, but it does not replace package metadata, actual runtime behavior, or validation in a real environment. A workflow is only trustworthy if a maintainer can still repeat it from the checked-in assets named here.

Next Checks¶

move to interfaces when the operational path depends on a specific surface contract
move to quality when the question becomes whether the workflow is sufficiently proven
move back to architecture when operational complexity suggests a structural problem

Purpose¶

This page reminds maintainers that packages are publishable units, not just folders in one repo.

Stability¶

Keep it aligned with the package's actual distributable surface.