Route every report to the right repository

How Novix gets involved

Any report, routed on the way in.

Nothing to route by hand. The second stage of every run resolves the repository and stamps it, before the diagnosis reads a single file.

How the agent handles it

Resolved once, and stamped
ContextPulls live evidence from your connected code and monitoring tools.
The routing happens at the second stage and is written onto the task. Re-deriving it later is the bug this prevents: by the time somebody approves, the text has grown a diagnosis and a fresh match lands somewhere else.
Strongest signal first
Already pinned, then the only repo connected, then the repo the task arrived from, then the repo its text names on a word boundary, then your default. Every step either answers or declines.
Weak words are excluded from the match
Names like api, app, web, service, backend and frontend appear in so many repositories that matching on them says nothing, so they never route a task on their own.
Naming two is a decline, not a coin flip
A report that matches two watched repositories takes the default. Picking the first would be a guess dressed as a decision, and everything downstream inherits it.
One answer, read by everybody after it
The file tree, the excerpts, the drafter, the sandbox clone and the pull request all read the stamp. Anything acting on a pull request that already exists reads the repository off the pull request’s own url instead.

What your team gets

The repository each task was routed to, and the reason, in plain words on the task
One pull request, in the repository the diagnosis was actually written against
A decline you can see, rather than a silent guess

It declines rather than guesses, every time

When the signals do not name a repository, Novix says so and takes your default. A wrong repository is worse than an unrouted one, because the file tree, the excerpts, the patch and the pull request all inherit it.

Works with

GitHub
GitLab
Bitbucket

Try it on a real customer problem.

Connect the place reports or product signals already arrive. The same agent takes it from evidence to a tested pull request.

Start free Chat with the founder

How the agent handles it

Resolved once, and stamped

The routing happens at the second stage and is written onto the task. Re-deriving it later is the bug this prevents: by the time somebody approves, the text has grown a diagnosis and a fresh match lands somewhere else.

Strongest signal first

Already pinned, then the only repo connected, then the repo the task arrived from, then the repo its text names on a word boundary, then your default. Every step either answers or declines.

Weak words are excluded from the match

Names like api, app, web, service, backend and frontend appear in so many repositories that matching on them says nothing, so they never route a task on their own.

Naming two is a decline, not a coin flip

A report that matches two watched repositories takes the default. Picking the first would be a guess dressed as a decision, and everything downstream inherits it.

One answer, read by everybody after it

The file tree, the excerpts, the drafter, the sandbox clone and the pull request all read the stamp. Anything acting on a pull request that already exists reads the repository off the pull request’s own url instead.