Models — API Reference
Package: dodil.ignite.v1 · Service: ModelService
OpenAI/Cohere-compatible inference over a global model catalog. The HTTP surface is built to drop in behind an existing OpenAI or Cohere client: the paths are top-level OpenAI-style (/v1/chat/completions, /v1/embeddings, /v1/rerank, /v1/audio/transcriptions, /v1/models) — not under /v1/ignite/ — and the JSON body matches the OpenAI/Cohere request exactly (max_tokens, top_p, top_n, response_format, …). Because the body is OpenAI/Cohere-native snake_case, the same JSON works over both HTTP and gRPC.
gRPC reaches every method at dodil.ignite.v1.ModelService/<Method> on $IGNITE_GRPC; the HTTP gateway mirrors each one. See Conventions → Using gRPC for grpcurl setup. HTTP base URL: https://api.dev.dodil.io. Auth is Authorization: Bearer $DODIL_TOKEN — except ListModels and GetModel, which are public (no auth).
A model’s declared input.format (from GetModel) tells you which RPC to call.
Sections
| Section | RPCs | HTTP | streaming |
|---|---|---|---|
| Chat Completions | ChatCompletion | POST /v1/chat/completions | unary |
StreamChatCompletion | POST /v1/chat/completions/stream | server-stream | |
| Embeddings | Embed | POST /v1/embeddings | unary |
| Rerank | Rerank | POST /v1/rerank | unary |
StreamRerank | POST /v1/rerank/stream | server-stream | |
| Transcription | Transcribe | POST /v1/audio/transcriptions | unary |
| Infer (generic) | Infer | POST /v1/infer | unary |
StreamInfer | POST /v1/infer/stream | server-stream | |
| Models (list & get) | ListModels | GET /v1/models | unary (no auth) |
GetModel | GET /v1/models/{model} | unary (no auth) |
Polymorphic fields
OpenAI-polymorphic request/response fields — content, tool_choice, response_format, embedding input — are carried as google.protobuf.Value, so you can paste the exact OpenAI/Cohere JSON object verbatim. See Conventions for wire and streaming notes.
See also
- Model Catalog — the human-readable model list
- Using OpenAI & Cohere SDKs — point an OpenAI/Cohere client at these endpoints
- Models — overview
- Conventions — transport, auth, wire format, streaming