Feature Table

OpenHosta 3.0 main features

• Lightweight MLLM Middleware (sync and async)
  • user defined headers
  • proxy compatibility (zscaler, litellm)
  • easy to debug
• Meta Prompt (jinja2 + dedent)
• pythonic pattern (from python space to LLM space with emulate)

OpenHosta 4.0 main features

• Lightweight MLLM Middleware (sync and async)
  • user defined headers
  • proxy compatibility (zscaler, litellm)
  • easy to debug
• [NEW] Local inference compatiblity (ollama, vllm, transformers)
• Meta Prompt (jinja2 + dedent)
• [NEW] Guardrails for typed data (data scheam verifications)
• [NEW] Safe and quantified uncertainty(if logprobs availables)
• pythonic pattern (from python space to LLM space with emulate)
• [NEW] Data Generation (if assistant completion available)

OpenHosta 5.0 main features

• Lightweight MLLM Middleware (sync and async)
  • user defined headers
  • proxy compatibility (zscaler, litellm)
  • easy to debug
• Local inference compatiblity (ollama, vllm, transformers)
• [NEW] LLM call audit (python hook at llm level)
• [NEW] Reproducible AI
• [NEW] LLM call caching (with reproducible AI enabled)
• Meta Prompt (jinja2 + dedent)
• Guardrails for typed data (data scheam verifications)
• pythonic pattern (from python space to LLM space with emulate)
• [NEW] Agentic workflows (constrained LLM using python code as a jail)
• [NEW] Assistant with tools (emulate in object plus use())
• Safe and quantified uncertainty(if logprobs availables)
• Data Generation (if assistant completion available)
• [NEW] Computer use (with pyautogui)

Model Table :

Tested version 3.0.4 :

• Azure API: preview 2025-xx-xx
• ollama 0.18
• vllm: ...
• transformers: 5.0
• Huggingface Router API:

Engine	Model	Text-to-text	Image-to-text	Logprobs	Generation Branching	Text-to-image
OpenAI API	GPT-4.1 / GPT-5.x	✅	✅	✅	❌	⬜
OpenAI Azure API	GPT-4.1 / GPT-5.x	✅	✅	❌	❌	⬜
Ollama API	Qwen3.5:27b	✅	✅	✅	❌	⬜
OpenAI Ollama API	Qwen3.5:27b	✅	✅	✅	❌	⬜
vllm API	Qwen3.5:27b	✅	✅	✅	❌	⬜
transformer API	Qwen3.5:4b	✅	✅	✅	❔	⬜
transformer API	flux2-klein:9b	❌	❌	❔	⬜	✅
Huggingface API	flux2-klein:9b	❌	❌	❔	⬜	✅
Huggingface API	Qwen3.5:27b	✅	✅	❔	❔	⬜