Overview
Ignite is Dodil’s serverless platform — run code as auto-scaling apps and call managed model inference, with remote image builds and secrets to support them.
Services
| Service | What it does |
|---|---|
| Compute | ComputeService — deploy code as serverless apps; draft → compile → publish → versioned releases; streaming Invoke, executions, logs/events, replicas, MCP tools. |
| Models | ModelService — managed inference: chat, embeddings, rerank, transcription, generic infer. OpenAI/Cohere-compatible. |
| Builds | BuildService — remote container image builds (Kaniko) from uploaded context or git; consumed by image-mode app versions. |
| Secrets | SecretService — Vault-backed git/registry credential slots, referenced by name from builds and deploys. |
How it fits together
An app’s code is either compiled by the managed runtime (Python/Rust/Go/Deno) or runs a container image — and that image is built by Builds, possibly pulling base layers from a private registry via credentials in Secrets. Once published, every app version invokes through the same gateway, auth, and scaling path.
Models is a global inference catalog with its own OpenAI/Cohere-compatible surface — independent of any app you deploy, but powered by the same Ignite infrastructure. Inference is offered serverless today; GPU-backed apps are on the roadmap.
See also
- Conventions — transport, auth, wire format, streaming
- CLI Basics — install + common flags
- Auth and Access