How it works·PR → Release decision

From pull request to human-reviewed release decision

Dulvarn connects PR changes, CI results, test signals and repository policy into one clear workflow for release risk review.

Your team keeps control. Dulvarn explains what changed, what is risky, what should be tested and whether the release is GO, CONDITIONAL GO or NO-GO.

GitHub-first Human-controlled Auditable
Workflow · PR to audit trail
PR opened
Analyze changes
Evaluate signals
Recommend scope
Generate decision
Human review
Audit trail
CI passingPartial NRTCONDITIONAL GOPending review
Workflow

A release decision workflow your team can inspect

Dulvarn does not replace your QA process. It makes the release decision path explicit, reviewable and traceable.

  1. 01
    Connect repository

    Connect your GitHub repository

    Install or connect Dulvarn to the repository you release from. Dulvarn reads PR metadata, changed files and configured repository policy.

    • repository
    • branch
    • PR metadata
    • release policy
  2. 02
    Configure release policy

    Set release guardrails

    Define thresholds and rules that influence GO, CONDITIONAL GO and NO-GO outcomes.

    • risk threshold
    • test integrity threshold
    • status checks enabled
    • ignored files
    • override policy
  3. 03
    Analyze pull request

    Analyze what changed

    Dulvarn evaluates changed files, affected areas and risky hotspots before the release decision is produced.

    • changed files
    • changed areas
    • touched product flows
    • hotspots
  4. 04
    Read CI and test signals

    Collect CI and test evidence

    CI can be green while product risk remains unclear. Dulvarn combines CI state with test integrity and impact signals.

    • CI status
    • test results
    • missing impacted tests
    • coverage gaps
  5. 05
    Recommend test scope

    Choose the right test scope

    Dulvarn recommends Smoke, Partial NRT or Full NRT based on risk, impact and policy.

    • Smoke
    • Partial NRT
    • Full NRT
    • manual checks
  6. 06
    Produce release decision

    Generate GO, CONDITIONAL GO or NO-GO

    The result is a decision summary with reasons, risk score, test integrity score and recommended next actions.

    • GO
    • CONDITIONAL GO
    • NO-GO
    • rationale
  7. 07
    Human review & audit

    Human decides, Dulvarn records

    QA or release owner reviews the evidence, approves or overrides the decision with a reason. Every decision is stored in the audit trail.

    • reviewer
    • override reason
    • audit event
    • decision history
Example scenario

Example: checkout PR with incomplete impacted tests

A representative example of how Dulvarn would evaluate a real-looking PR.

northwind/checkout-service · PR #1284
CI passing
Refactor coupon stacking + idempotent payment retries
Changed areas
  • checkout
  • cart
  • auth
  • payment cfg
Detected risk
  • payment retry hotspot changed
  • coupon stacking logic changed
  • impacted tests missing for payment retry
  • no coverage on partial-refund regression
Release decision · evidence
Conditional Go
Dulvarn recommendationPartial NRT
DecisionCONDITIONAL GO
Human statusWaiting for QA review
Audit eventlogged
Rationale

CI is passing, but the change touches checkout and payment retry logic while impacted test coverage is incomplete. Dulvarn recommends Partial NRT before human release approval.

Decision logic

What influences GO / NO-GO?

A release decision combines code risk, CI state, test integrity, test impact and repository policy.

Code risk

Changed files, touched flows, hotspots and complexity.

CI state

Passing, failing or unknown CI status.

Test integrity

Whether mapped tests exist and whether coverage looks strong enough.

Test impact

Which suites are relevant to the changed areas.

Repository policy

Thresholds configured by the team.

Human override policy

Whether a reviewer must add a reason before overriding.

The exact policy can be configured per repository. Dulvarn explains the decision so humans can review it.

Test scope

Smoke, Partial NRT or Full NRT — with reasoning

Dulvarn helps the team avoid guessing regression scope.

Smoke

low-risk

Best for low-risk changes where critical flows are not affected and test integrity is healthy.

Partial NRT

recommended

Best for medium-risk changes that affect specific product areas such as checkout, auth or payments.

Full NRT

high-risk

Best for broad, high-risk releases, failing CI, low test integrity or changes across critical flows.

Decision matrix
  • Low risk + healthy testsSmoke
  • Medium risk + affected critical areaPartial NRT
  • High risk or failing CIFull NRT / NO-GO
  • Unknown CI or missing testsConditional review
Human control

AI explains risk. Humans decide.

Dulvarn is intentionally designed as decision support, not autonomous release control. QA Leads, Release Managers and Engineering Leads keep ownership of the final release decision.

  • Every decision has a rationale.
  • Every override requires a reason.
  • Every review is audit logged.
  • Release ownership stays with the team.
Integration

Designed for GitHub-first workflows

Dulvarn fits into the tools teams already use for PR review, CI and release preparation.

GitHub PR analysis

Triggered on pull request events and updates.

Status check concept

Release decision can surface as a PR status.

PR comments

Risk summary and reasoning surfaced on the PR.

CI signal ingestion

Reads CI status and test results as evidence.

Repository policy

Required checks and release rules per repo.

Audit events

Every decision and override retained per repository.

Designed for GitHub-first workflows · integration depth depends on beta setup.

Founder-led beta

See how Dulvarn would review your release workflow

Dulvarn is currently being tested with QA and engineering teams that want clearer release decisions, better test scope recommendations and fewer production surprises.

  • Review one real release workflow
  • Identify release risk blind spots
  • Map test scope decision points
  • Discuss beta fit