Skip to content

Supported Services

nsportsman edited this page Feb 7, 2026 · 2 revisions

Supported Services (32)

Navigation: Home | Probe YAML Reference | Match Rules | CLI Reference | Architecture


Julius identifies 32 LLM platforms across 5 categories. Each service is detected via HTTP-based fingerprinting using probe YAML files that define request/response signatures.

Self-Hosted LLM Servers (15)

Service Probe Name Port Specificity Models Augustus API Docs
Ollama ollama 11434 100 Yes Yes docs
vLLM vllm 8000 75 Yes Yes docs
LocalAI localai 8080 75 Yes Yes docs
llama.cpp llama-cpp 8080 50 Yes Yes docs
Hugging Face TGI huggingface-tgi 3000 75 Yes Yes docs
LM Studio lm-studio 1234 70 Yes Yes docs
Aphrodite Engine aphrodite-engine 2242 75 Yes Yes docs
FastChat fastchat-controller 21001 75 Yes Yes docs
GPT4All gpt4all 4891 75 Yes Yes docs
Gradio gradio 7860 75 No No docs
Jan jan 1337 75 Yes Yes docs
KoboldCpp koboldcpp 5001 75 Yes Yes docs
NVIDIA NIM nvidia-nim 8000 75 Yes Yes docs
TabbyAPI tabbyapi 5000 75 Yes Yes docs
Text Generation WebUI text-generation-webui 5000 75 Yes Yes docs

Detection Details

Ollama (specificity: 100) -- The highest-confidence detection. Probes /api/tags for Ollama's unique model listing format and / for the "Ollama is running" banner. Uses require: any (first match wins).

vLLM (specificity: 75) -- Uses require: all, meaning both requests must match. Probes /version for version info and /v1/models for the OpenAI-compatible model listing. The combination of both endpoints distinguishes vLLM from other OpenAI-compatible servers.

Gradio (specificity: 75) -- Probes three endpoints to detect Gradio ML demo interfaces. Does not support model extraction since Gradio apps vary widely.

Aphrodite Engine (specificity: 75) -- Large-scale LLM inference engine with OpenAI-compatible API. Detected via distinctive version endpoint responses.

FastChat (specificity: 75) -- Open platform for training, serving, and evaluating LLM chatbots. Detected via controller API endpoints.

KoboldCpp (specificity: 75) -- Easy-to-use AI text-generation software for GGML/GGUF models with KoboldAI-compatible API.

Gateway Services (3)

Service Probe Name Port Specificity Models Augustus API Docs
LiteLLM litellm 4000 85 Yes Yes docs
Kong AI Gateway kong-ai-proxy 8000 80 No Yes docs
Envoy AI Gateway envoy-ai-gateway 80 75 No Yes docs

Detection Details

LiteLLM (specificity: 85) -- High specificity because LiteLLM has distinctive response signatures. Probes three endpoints including its health check and model listing.

Kong AI Gateway (specificity: 80) -- Detects the Kong AI Proxy plugin by probing Kong-specific endpoints.

Envoy AI Gateway (specificity: 75) -- Unified access to generative AI services built on Envoy Gateway. Detected via Envoy-specific API endpoints.

RAG & Orchestration Platforms (12)

Service Probe Name Port Specificity Models Augustus API Docs
Open WebUI openwebui 3000 80 Yes Yes docs
LibreChat librechat 3080 50 No No docs
SillyTavern sillytavern 8000 50 No No docs
BetterChatGPT betterchatgpt 3000 50 No No docs
AnythingLLM anythingllm 3001 50 No No docs
AstrBot astrbot 6185 75 No No docs
Dify dify 80 75 No No docs
Flowise flowise 3000 75 No No docs
HuggingFace Chat UI huggingface-chat-ui 3000 75 No No docs
LobeHub lobehub 3210 75 No No docs
NextChat nextchat 3000 75 No No docs
Onyx onyx 3000 75 No No docs

Detection Details

Open WebUI (specificity: 80) -- Most comprehensive detection in this category with 4 request probes. Checks /api/config for the "Open WebUI" name, feature flags, and /api/version for deployment metadata.

LibreChat, SillyTavern, BetterChatGPT, AnythingLLM (specificity: 50) -- Default specificity. These frontends are detected via their static assets or API endpoints but have less distinctive signatures than dedicated inference servers.

AstrBot (specificity: 75) -- Multi-platform LLM chatbot framework with dashboard and plugin system. Detected via dashboard API endpoints.

Dify (specificity: 75) -- Open-source LLM app development platform with workflow orchestration. Detected via console API endpoints.

Flowise (specificity: 75) -- Low-code platform for building AI agents and LLM workflows. Detected via API node endpoints.

HuggingFace Chat UI (specificity: 75) -- Open source ChatGPT-style interface powering HuggingChat. Detected via settings and model endpoints.

LobeHub (specificity: 75) -- Multi-agent AI collaboration platform with chat UI. Detected via manifest and API endpoints.

NextChat (specificity: 75) -- Self-hosted ChatGPT-style interface supporting multiple AI providers. Detected via config API.

Onyx (specificity: 75) -- Open source AI platform for enterprise search and chat with RAG. Detected via health and settings endpoints.

Cloud-Managed Services (1)

Service Probe Name Port Specificity Models Augustus API Docs
Salesforce Einstein salesforce-einstein 443 75 No No docs

Detection Details

Salesforce Einstein (specificity: 75) -- Detects the Salesforce Einstein Messaging API (MIAW) via its distinctive API endpoints on port 443.

Generic Detection (1)

Service Probe Name Port Specificity Models Augustus API Docs
OpenAI-compatible openai-compatible -- 1 Yes Yes docs

Detection Details

OpenAI-compatible (specificity: 1) -- The lowest-priority fallback probe. Tests 8 different OpenAI API endpoints (/v1/models, /v1/chat/completions, etc.) to detect any server implementing the OpenAI API specification. Because many services implement this API, it has the lowest specificity and only matches when no more specific probe does.

Summary

Category Count Specificity Range
Self-Hosted 15 50-100
Gateway 3 75-85
RAG & Orchestration 12 50-80
Cloud-Managed 1 75
Generic 1 1

Navigation: Home | Probe YAML Reference | Match Rules | CLI Reference | Architecture

Clone this wiki locally