npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@orivus-ai/coding-agent

v0.1.1

Published

Orivus Coding Agent — governed change intelligence for Orivus Edge OS.

Downloads

42

Readme

Orivus Coding Agent

Apache-2.0 reference agent module for governed software mutation — technical preview.

Most AI coding tools interpret your intent and mutate your workspace in the same step.
Orivus separates those roles.

| Role | Package | License | |------|---------|---------| | Interpretation — risk, plan, preview, review, handoff preparation | @orivus-ai/coding-agent | Apache-2.0 | | Authority — sessions, attestation, apply, audit, replay | @orivus-ai/edge | BUSL-1.1 (source-available kernel) |

Proposal ≠ authority.
Agents interpret. The kernel authorizes mutation.

This repository is not the kernel. It is the first open reference agent built on Orivus Edge: a local, governance-first runtime where nothing governed touches disk until a human explicitly confirms and the kernel executes apply.

Requires: @orivus-ai/edge ^0.1.0 (peer). Not a standalone coding tool.


Why this exists

Today's AI coding stacks collapse two distinct responsibilities:

| Responsibility | Question it answers | Who should own it | |----------------|---------------------|-------------------| | Interpretation | What change is proposed, and is it reasonable? | Agent module | | Mutation authority | May this workspace change, now, on disk? | Kernel + explicit human confirm |

When both live in the same tool path, you get:

  • patches applied while you are still reading the plan,
  • no durable record of who authorized what,
  • previews that silently go stale,
  • sessions that cannot be replayed or audited,
  • “the model decided” substituted for an authorization gate.

That is not only a model-quality problem. It is a systems architecture problem.

Proposal ≠ authority. An agent that can interpret and write production files without a separate authorization boundary is automation — not governance.

Orivus names the boundary and enforces it: agents prepare governed proposals in session; the kernel alone executes mutation after attestation, drift checks, and --confirm.


Who this is for

| You are… | Why this module matters | |----------|-------------------------| | Senior engineer | You want inspectable proposals, real previews, and explicit confirm before disk writes | | Platform / developer-experience team | You need a reference pattern for agent modules that cannot own final mutation authority | | Local-first / air-gapped environment | You run inference and governance on-machine; mutation stays session-gated until confirm | | AI-native workflow builder | You are composing agents on Edge, not wrapping another copilot SDK | | Team with mutation policy | You require audit trails, attestation, drift detection, and replayable sessions |

Not a fit if you want a chat-first copilot that edits files immediately, fully autonomous production apply, or a tool that works without the Edge kernel peer.


30-second mental model

Agent   → interpretation   (intent → risk → plan → preview → review)
Session → source of truth    (pendingMutation, hashes, stage seals)
Human   → authorization      (apply --confirm)
Kernel  → authority           (attest → drift check → write → audit)
orivus coding change              →  ready_for_confirm (workspace unchanged)
orivus coding apply               →  blocked (confirm required)
orivus coding apply --preview     →  unified diff only (still unchanged)
orivus coding apply --confirm     →  handoff → kernel apply → audit record

One sentence to remember:

Agents interpret. The kernel authorizes mutation.


Governed execution flow

Interpretation and authority are separate code paths. The agent never calls the kernel write API directly.

flowchart LR
  subgraph interpretation["Agent module (@orivus-ai/coding-agent)"]
    A[change pipeline]
  end
  subgraph session["Session (.orivus/session/)"]
    S[current_session.json<br/>pendingMutation · previewHash · seals]
  end
  subgraph authorization["Human gate"]
    H[apply --confirm]
  end
  subgraph authority["Kernel (@orivus-ai/edge)"]
    K[validate session + attestation]
    G[GovernedApplyService]
    W[workspace write]
    U[audit + archive]
  end

  A --> S
  S --> H
  H -->|handoff payload| K
  K --> G
  G --> W
  W --> U

| Step | Component | Disk under governed targets | |------|-----------|----------------------------| | coding change | Agent pipeline | Unchanged — proposal in session only | | coding apply | Kernel gate | Blocked without confirm | | coding apply --preview | Kernel gate | Diff only — still unchanged | | coding apply --confirm | Agent validates → handoff → kernel act | Written + audited |

The agent's apply --confirm does not write your source tree. It validates integrity and returns a handoff. Only the kernel apply path mutates reality.


60-second demo

Reproducible golden path from Orivus Edge (certification/fixtures/v1-golden-path):

# 0. Prerequisites: @orivus-ai/edge installed globally (see Install)
cd certification/fixtures/v1-golden-path

orivus init --yes
orivus index

# 1. Interpretation — full pipeline, session only
orivus coding change --contract contracts/golden-path.orivus.yaml

# 2. Authority check — workspace still clean
git diff    # expect: empty (pending work lives under .orivus/session/)

# 3. Gate — confirm required
orivus coding apply
# expect: blocked, no file mutation

# 4. Inspect — real unified diff, still no write
orivus coding apply --preview
# expect: diff output; git diff still empty

# 5. Authorize — kernel write + audit
orivus coding apply --confirm
# expect: governed target changed; session archived; audit trail updated

What proves governance: after step 1, git diff is empty. After step 5, only then does the governed file change.

Full narrative: Orivus Edge — demo release v0.1.

Contract-driven change (deterministic)

For demos without free-form intent (explicit symbol — no guessed edit locus):

orivus coding change --contract examples/contracts/safe-null-guard.orivus.yaml

See examples/contracts/README.md.


Install

From npm (recommended)

npm install -g @orivus-ai/edge
cd your-project
orivus init    # registers @orivus-ai/coding-agent when configured

You do not need to clone this repository for normal use.
@orivus-ai/[email protected] is on registry.npmjs.org.

Local development (both repos)

git clone https://github.com/orivus/orivus-edge.git
git clone https://github.com/orivus/orivus-coding-agent.git
cd orivus-edge && pnpm install && pnpm run build
cd ../orivus-coding-agent && pnpm install && pnpm run build
orivus init --yes --agent-path /absolute/path/to/orivus-coding-agent

What changes on disk (v0.1)

| Phase | Actor | Governed source files | |-------|-------|------------------------| | Risk → review pipeline | Coding agent | Unchanged | | apply (no flags) | Kernel | Blocked | | apply --preview | Kernel | Preview only | | apply --confirm | Kernel GovernedApplyService | Written + audited |

Guarantees in this preview:

  • Session-only proposals — pending work under .orivus/session/, not your tree
  • Pipeline attestation — incomplete or tampered pipelines cannot apply
  • Preview before review — reviewer sees real proposed content, not summaries
  • Drift detection — stale preview vs on-disk file blocks apply
  • Syntax verification — structural validation before write
  • Kernel-owned audit — apply and block paths recorded

Core law: Agents propose. The kernel governs. The runtime observes.


What this package does

Governed single-file change pipeline (v0.1):

| Stage | Purpose | |-------|---------| | risk | Analyze change risk; abort on CRITICAL | | plan | Build plan from intent + structural context | | preview | Proposed file content in memory | | adversarial_review | Adversarial pass on preview | | review | Reviewer evaluates real preview vs original | | ready_for_confirm | pendingMutation with previewHash + targetFingerprint |

CLI (via orivus coding …):

| Command | Purpose | |---------|---------| | change | Full pipeline + session persistence; normal CLI shows a concise governed mission report (use --verbose for contract envelopes) | | apply --preview | Unified diff; disk unchanged | | apply --confirm | Session validation → handoff to kernel | | risk / plan / review / run | Partial paths — not the certified golden path |


What this package must never do

| Prohibited | Enforcement | |------------|-------------| | Write governed target source files | ApplyCommand returns handoff only (CERT-SOV-001) | | Direct apply via --target / --content | UNSUPPORTED_DIRECT_HANDOFF | | Skip preview-before-review | Fixed stage order in CodingChangePipeline | | Bypass session on governed apply | Requires current_session.json | | Import Edge kernel internals | @orivus-ai/edge/sdk only; pnpm run certify | | Raw filesystem I/O in commands | Workspace reads via granted storage capability |

No final mutation authority — by design.


Integration with Edge

sequenceDiagram
  participant User
  participant EdgeCLI as orivus CLI
  participant Agent as @orivus-ai/coding-agent
  participant Kernel as GovernedApplyService

  User->>EdgeCLI: coding change
  EdgeCLI->>Agent: run pipeline
  Agent->>Agent: session metadata only
  EdgeCLI->>EdgeCLI: SessionPipelineSealer

  User->>EdgeCLI: coding apply --confirm
  EdgeCLI->>Agent: apply
  Agent-->>EdgeCLI: handoff
  EdgeCLI->>Kernel: execute(authorization)
  Kernel->>Kernel: write file + audit
  1. orivus init registers module id: coding.
  2. CapabilityBroker grants declared capabilities at activation.
  3. On confirm, Edge forwards authorization to GovernedApplyService — the only legal workspace write path.

Capabilities required (src/agent.manifest.ts):

INFERENCE_EXECUTE · TOPOLOGY_READ · INDEX_READ · GIT_SIGNALS_READ · SOURCE_PARSE · LOGGER_WRITE · STORAGE_WORKSPACE_READ · STORAGE_SESSION_WRITE

Session files (agent scope):

| File | Writer | Content | |------|--------|---------| | .orivus/session/current_session.json | change | Pipeline state, pendingMutation, stages | | .orivus/session/last_session.json | Edge after apply | Archived session | | .orivus/session/last_plan.json | plan only | Standalone plan artifact |

The agent never writes governed paths under your workspace source tree.

Handoff on apply --confirm:

{
  status: 'success',
  handoff: {
    ready: true,
    payload: {
      type: 'file_mutation',
      file,
      content,
      previewHash,
      // …
    },
    sessionRef: session.sessionId,
  },
}

Kernel semantics: Orivus Edge — governance.


Build another agent

Orivus agents are EdgeAgentModule packages: declare capabilities, run a governed pipeline, persist proposals in session, return handoff on apply — never invoke kernel apply from agent code.


Honest limits (v0.1)

Technical preview — not a finished platform.

| Area | Reality | |------|---------| | Scope | Single-file governed mutation only | | Autonomy | Human confirm required — not certified for unattended production apply | | Sandbox | Trusted binary — no VM isolation for agent code | | Inference | Local-first — first model download can be slow; live LLM behavior varies | | Context | Structural (index, topology) — not a deep semantic call graph | | Commands | change + apply are certified; run and partial commands are secondary |

We block apply rather than guess when attestation, hash, or drift checks fail.

Kernel limits: known limitations v0.1.


What we are not claiming

| Narrative | Status | |-----------|--------| | “Another AI coding assistant” | No — mutation authority is separated and kernel-governed | | “LangChain / copilot wrapper” | No — Edge kernel + agent module architecture | | “Fully autonomous coding” | No — explicit --confirm is mandatory | | “Fully open-source platform” | No — kernel is BUSL-1.1; this agent is Apache-2.0 | | “AI OS” | No — governed local coding runtime preview |


Develop and certify

pnpm install
pnpm run build
pnpm run check      # typecheck + lint + test + build
pnpm run certify    # boundary scan + behavioral certification

| Suite | Proves | |-------|--------| | certification.sovereignty.test.ts | No disk write on agent apply confirm | | ApplyCommand.test.ts | Drift, hash mismatch, blocked direct handoff | | CodingChangePipeline.test.ts | Stage ordering, retries |

docs/certification.md · docs/release-checklist-v0.1.md


Documentation

| Doc | Topic | |-----|-------| | Orivus Edge README | Kernel, CLI, runtime | | Governance | Attestation, drift, audit, replay | | CLI reference | Full commands | | Certification | pnpm run certify | | Contributing | PR rules, governance laws | | Security | Reporting, threat model | | Changelog | Releases |


License

Apache License 2.0LICENSE, NOTICE.

  • Runtime peer: @orivus-ai/edge (BUSL-1.1). Commercial kernel use may require a separate license — Edge licensing.
  • Mutation authority: Edge kernel only. This agent proposes; it does not own the write path.