Deploy and security
Most of our customers run Sahasa.AI inside their own cloud. Here is exactly what that involves, what we can see, and what we can't.
An isolated tenant we operate. You connect a Git provider and a requirements source. Your AI keys or ours. Live the same day.
fastest start
A small set of containers: API, worker, PostgreSQL, a vector store, Redis. Runs on a couple of VMs or in your Kubernetes. AWS, Azure, GCP or your own racks.
the enterprise default
Either of the above, plus our engineers inside your organisation building adapters, connectors and quality gates, and running the platform day to day.
when you want outcomes, not another tool
Inside your boundary
Read-only access to your repositories. Outbound calls only to the AI provider you choose. Nothing phones home.
Two or three VMs with Docker, or a namespace in your cluster. Sized for your repo count; a mid-size org runs comfortably on a pair of 8-vCPU machines.
PostgreSQL for results, a vector store for patterns, object storage for screenshots and reports. All yours, encrypted with your keys, retained on your schedule.
A read-only token to your Git provider. Write access only to open pull requests, if you want that. A service account for Jira. SSO through your IdP.
To Anthropic, OpenAI, Google, or an internal gateway in front of Bedrock or Azure OpenAI. Your keys, your data agreements. Prompts and responses are logged in your database, not ours.
Signed container images from our registry. You pull on your schedule. Licence file with an expiry; on expiry the platform goes read-only, your data stays.
Pilot
We don't do slideware pilots. You pick the service. We deploy, connect, and come back with results you can check.
Deploy in your cloud or connect to ours. Read-only access to one repository and its requirements. We agree the baseline: current coverage, current findings, current time to a green build.
Test suites generated into your repo and run in your CI. Bugs, security findings and the threat model on every pull request in that period. Live UI checks on staging. A traceability matrix for the requirements you gave us.
We triage one real failed build together, walk through the evidence states, and put the numbers next to the baseline. You decide. If you go ahead, the pilot fee is credited against the first year.
One engineer for about two hours a week, a repository, a requirements source (Jira is fine), a staging URL, and an AI provider key or permission to use ours.
Next step
A demo runs on your code, not ours. Forty-five minutes, one service, real findings. If it's useful, we go into a four-week pilot inside your CI.