mirror of https://github.com/whit3rabbit/anyllm-proxy.git synced 2026-10-06 00:00:12 +00:00

Files

T

whit3rabbitandClaude Sonnet 4.6 545bdbbbd3 feat: add bedrock native passthrough, generic passthrough, provider docs, and managed backend fixes

Adds bedrock_native.rs (Converse/InvokeModel with SigV4) and
generic_passthrough.rs catch-all for Translate mode. Adds comprehensive
provider reference docs (docs/providers/, docs/ENDPOINTS.md). Fixes
managed backend admin UI (BackendForm, ManagedBackendsSection) and
admin route/model handler issues. Adds automated model pricing update
workflow (scripts/update_pricing.py, .github/workflows/update-pricing.yml).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

2026-04-08 16:36:16 -05:00

2.3 KiB

Raw Permalink Blame History

NVIDIA Triton

NVIDIA Triton Inference Server with an OpenAI-compatible frontend via the TensorRT-LLM backend.

LiteLLM prefix: triton/ Status: Stub — routes through OpenAI-compatible client Docs: https://github.com/triton-inference-server/tensorrtllm_backend

Authentication

Variable	Required	Description
`OPENAI_BASE_URL`	Yes	URL of the Triton OpenAI-compatible endpoint (e.g. `http://my-host:8000/v1`)

Quick Start

Single-Backend (env vars)

BACKEND=triton OPENAI_BASE_URL=http://my-host:8000/v1 PROXY_OPEN_RELAY=true cargo run -p anyllm_proxy
# or Docker:
docker run \
  -e BACKEND=triton \
  -e OPENAI_BASE_URL=http://my-host:8000/v1 \
  -e PROXY_OPEN_RELAY=true \
  -p 3000:3000 followthewhit3rabbit/anyllm-proxy

LiteLLM YAML Config

model_list:
  - model_name: local-model
    litellm_params:
      model: triton/ensemble
      api_base: "http://my-host:8000/v1"

Usage Examples

Anthropic Messages API

curl http://localhost:3000/v1/messages \
  -H "x-api-key: $PROXY_API_KEYS" \
  -H "Content-Type: application/json" \
  -d '{"model": "ensemble", "max_tokens": 1024, "messages": [{"role": "user", "content": "Hello"}]}'

OpenAI Chat Completions API

curl http://localhost:3000/v1/chat/completions \
  -H "Authorization: Bearer $PROXY_API_KEYS" \
  -H "Content-Type: application/json" \
  -d '{"model": "ensemble", "messages": [{"role": "user", "content": "Hello"}]}'

Capabilities

Feature	Supported
Chat Completions	✓
Streaming	✓
Tool Use	—
Embeddings	—
Vision	—
Batch	—

Notes

Triton's native protocol is gRPC/HTTP but does not expose an OpenAI-compatible API by default. The OpenAI-compatible frontend requires the TensorRT-LLM backend (tensorrtllm_backend) and its bundled API server.

Triton has no fixed default URL. Set OPENAI_BASE_URL to the address of your deployment.

The model name in requests corresponds to the Triton model repository name (commonly ensemble in TRT-LLM deployments). Check your model repository for the correct name.

Triton is production-grade but requires significant setup: GPU drivers, TensorRT-LLM engine compilation, and a configured model repository. Not suitable for quick local experimentation; use Ollama or LM Studio for that.

2.3 KiB Raw Permalink Blame History