A normal reviewed change
- soltura.yaml
- closed generated clients
- instrumentation + focused tests
- local review package
Business observability, built from your code
Start with the business promise, not a tracking plan. The Soltura skill finds the minimum evidence your code must emit, prepares it for review, and keeps the approved answer accountable in Cloud.
Can customers complete purchases reliably?business commitmentLocal discovery → reviewed pull request → approved aggregates → Cloud

The missing agreement
Event names are not business definitions. “Checkout completed” can mean a button click, a payment attempt, or an authoritative order. Each dashboard can be technically correct and still disagree.
Soltura turns the definition your team would act on into a reviewed contract: the outcome, its boundary, who owns it, and the evidence required to defend the answer.
The first product surface is local
The skill works backwards from the decision, finds the narrowest evidence your code can prove, and asks for approval before editing. The live answer begins only after the reviewed change is deployed and approved signals arrive.
Use the Soltura skill. Can customers complete purchases reliably?
I found a durable order_confirmed boundary and a correlated browser-to-server checkout path. Here is the smallest accountable commitment I can defend.
Browser milestones remain observational. Order confirmation is authoritative. The Cloud receives neither this conversation nor your repository.
checkout_startedcheckout_payment_submittedpayment_resolvedorder_confirmedApprove this evidence, privacy boundary, file scope, and test plan before I edit?
Needs attention
33.8% observed · measurement limitedCommitments, not dashboard tiles
Every commitment keeps the question, reviewed expectation, accountable team, and next action together. Guardrails can qualify the verdict, but they never get averaged into a reassuring score.

One question. One honest visual.
Cloud chooses a bounded visual for the evidence the contract actually supports: change over time, where progress stops, group comparison, or composition. Missing and degraded measurement stay visible.
Open the incident; do not wait for the revenue report.

Start the investigation at that transition—not at a random page.

Localize the investigation without pretending the segment is a cause.

Restore measurement confidence before calling traffic healthy.

Product screens shown with Soltura’s explicitly synthetic demo data.
From departure to investigation
When a commitment departs from its reviewed expectation, Soltura ranks only what the approved aggregates support: facts, concentrations, and associations. It never upgrades correlation into causality.
If the evidence cannot support a firm explanation, the conclusion remains unknown and the investigation continues in the skill.

Deliberately bounded today
The skill can begin from the business question. Today, the approved protocol can make ordered funnels and traffic quality live in Cloud. If the answer needs a new semantic shape, Soltura exposes that gap before instrumentation ships.
No plausible-looking fallback: insufficient evidence stays unknown, not healthy, zero, or guessed.
Designed for review, not blind access
The skill works where your code already lives. Nothing reaches Soltura until your team has reviewed the instrumentation and the exact signal contract.
Stays local
Source code, agent conversations, diffs, test output, logs, datasets, raw payloads, and secrets.
Reaches Soltura
The reviewed signal contract and contract-valid aggregate events needed to keep the commitment answered.
Step one happens locally
Install the skill, review the contract it prepares, and let Cloud keep the answer accountable.
Install the Soltura skill