Create a project

AI harness on Temporal. Bundles workflow orchestration, vLLM inference, a model and custom worker images.

Identity and access management, backed by a linked PostgreSQL database. Each instance needs its own database.

High-availability PostgreSQL with automated backups, PITR and read replicas.

CloudNativePG clusters on Kubernetes with PITR, multi-database provisioning and rolling upgrades.

In-memory cache and session store with optional persistence.

S3-compatible object storage tenant with versioning and lifecycle policies.

Temporal workflow server. Requires a postgres connection exposing at least two databases.

OpenAI-compatible inference server on H100 nodepools, with tensor parallelism and paged KV cache.

Open-weight Mistral models cached on node-local NVMe and served through vLLM.

Signed worker containers from the tenant registry, polling a Temporal task queue with autoscaling replicas.

Replicates buckets and database dumps to OVHcloud object storage, outside the primary cloud.