#!/usr/bin/env python3
"""
Check LLM backend status and print formatted output.
Usage: python scripts/check_llm_status.py
"""
from pdf_mcp import llm_setup, pdf_tools
def _status_label(available: bool) -> str:
return "OK" if available else "NO"
def main():
info = pdf_tools.get_llm_backend_info()
current = info["current_backend"] or "none"
print("LLM BACKEND STATUS")
print("-" * 60)
print(f"current_backend: {current}")
print()
print("backends:")
ollama_model = llm_setup.get_ollama_model_name()
ollama_installed = llm_setup.ollama_is_installed()
ollama_models = llm_setup.ollama_list_models() if ollama_installed else set()
# Query local server for model info
local_health = llm_setup.get_local_server_health()
local_models = llm_setup.get_local_server_models()
for name, data in info["backends"].items():
available = data.get("available", False)
cost = data.get("cost", "unknown")
print(f"- {name:8} status={_status_label(available)} cost={cost}")
if name == "local":
print(f" url={data.get('url')} config_model={data.get('model')}")
if available:
if local_models:
models_list = local_models.get("models", local_models.get("available", []))
if models_list:
# Handle both list of strings and list of dicts
if isinstance(models_list, list):
names = [m.get("name", str(m)) if isinstance(m, dict) else str(m) for m in models_list]
print(f" loaded_models: {', '.join(names)}")
else:
print(f" loaded_models: {models_list}")
else:
print(" loaded_models: (endpoint returned no model list)")
else:
print(" loaded_models: (no /models endpoint)")
else:
print(" start: cd ~/agentic-ai-research && source .venv/bin/activate && python -m services.model_server.cli serve --port 8100")
if name == "ollama":
if not ollama_installed:
print(" install: curl -fsSL https://ollama.ai/install.sh | sh")
elif not ollama_models:
print(" note: ollama list failed or no models found (is the service running?)")
elif ollama_model in ollama_models:
print(f" model: {ollama_model} (installed)")
else:
print(f" model: {ollama_model} (missing)")
print(f" install: ollama pull {ollama_model}")
if name == "openai" and not available:
print(" set: export OPENAI_API_KEY=your-key")
print()
print("recommended model: Qwen3-VL-30B-A3B (95.7% DocVQA, MoE architecture)")
print()
print("override backend: export PDF_MCP_LLM_BACKEND=local|ollama|openai")
print(f"override ollama model: export {llm_setup.OLLAMA_MODEL_ENV}=model:tag")
print()
if __name__ == "__main__":
main()
