Add configured model connectivity check

@@ -49,6 +49,7 @@ cp .env.example .env
 
 
 ```bash
 ```bash
 python scripts/00_create_kb.py
 python scripts/00_create_kb.py
+python scripts/00_check_models.py
 python scripts/01_upload_docs.py
 python scripts/01_upload_docs.py
 python scripts/02_wait_ingestion.py
 python scripts/02_wait_ingestion.py
 python scripts/03_export_chunks.py
 python scripts/03_export_chunks.py

+from __future__ import annotations
+
+import sys
+import time
+from typing import Any
+
+import _bootstrap  # noqa: F401
+
+import requests
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
+
+from weknora_eval.config import load_config
+
+
+def main() -> int:
+    config = load_config()
+    ragas = config["ragas"]
+    failures: list[str] = []
+
+    print("Checking configured Ragas model services...\n")
+
+    failures.extend(
+        check_chat_model(
+            title="Generator LLM",
+            base_url=require_value(ragas, "llm_base_url"),
+            api_key=require_value(ragas, "llm_api_key"),
+            model=require_value(ragas, "generator_model"),
+            temperature=float(ragas.get("temperature", 0)),
+            max_tokens=min(int(ragas.get("max_tokens", 1024)), 1024),
+        )
+    )
+    failures.extend(
+        check_chat_model(
+            title="Judge LLM",
+            base_url=require_value(ragas, "llm_base_url"),
+            api_key=require_value(ragas, "llm_api_key"),
+            model=require_value(ragas, "judge_model"),
+            temperature=float(ragas.get("temperature", 0)),
+            max_tokens=min(int(ragas.get("max_tokens", 1024)), 1024),
+        )
+    )
+    failures.extend(
+        check_embedding_model(
+            base_url=require_value(ragas, "embedding_base_url"),
+            api_key=require_value(ragas, "embedding_api_key"),
+            model=require_value(ragas, "embedding_model"),
+        )
+    )
+
+    reranker_base_url = str(ragas.get("reranker_base_url") or "")
+    reranker_model = str(ragas.get("reranker_model") or "")
+    if reranker_base_url and reranker_model:
+        failures.extend(
+            check_reranker_model(
+                base_url=reranker_base_url,
+                api_key=str(ragas.get("reranker_api_key") or ""),
+                model=reranker_model,
+            )
+        )
+    else:
+        print("[SKIP] Reranker: RAGAS_RERANKER_BASE_URL or RAGAS_RERANKER_MODEL is empty\n")
+
+    if failures:
+        print("Model service check failed:")
+        for failure in failures:
+            print(f"- {failure}")
+        return 1
+
+    print("All configured model services are reachable.")
+    return 0
+
+
+def check_chat_model(
+    *,
+    title: str,
+    base_url: str,
+    api_key: str,
+    model: str,
+    temperature: float,
+    max_tokens: int,
+) -> list[str]:
+    print(f"[CHECK] {title}: model={model} base_url={base_url}")
+    started = time.monotonic()
+    try:
+        llm = ChatOpenAI(
+            model=model,
+            api_key=api_key,
+            base_url=base_url,
+            temperature=temperature,
+            max_tokens=max_tokens,
+            timeout=120,
+        )
+        response = llm.invoke("Reply with exactly: OK")
+        content = str(response.content or "").strip()
+        elapsed = time.monotonic() - started
+        if not content:
+            return [f"{title} returned an empty response"]
+        print(f"[OK] {title}: {elapsed:.2f}s response={content[:80]!r}\n")
+        return []
+    except Exception as exc:  # noqa: BLE001
+        return [f"{title} failed: {exc}"]
+
+
+def check_embedding_model(*, base_url: str, api_key: str, model: str) -> list[str]:
+    print(f"[CHECK] Embedding: model={model} base_url={base_url}")
+    started = time.monotonic()
+    try:
+        embeddings = OpenAIEmbeddings(
+            model=model,
+            api_key=api_key,
+            base_url=base_url,
+            tiktoken_enabled=False,
+            check_embedding_ctx_length=False,
+            request_timeout=120,
+        )
+        vector = embeddings.embed_query("hello")
+        elapsed = time.monotonic() - started
+        if not vector:
+            return ["Embedding returned an empty vector"]
+        print(f"[OK] Embedding: {elapsed:.2f}s dimensions={len(vector)} first3={vector[:3]}\n")
+        return []
+    except Exception as exc:  # noqa: BLE001
+        return [f"Embedding failed: {exc}"]
+
+
+def check_reranker_model(*, base_url: str, api_key: str, model: str) -> list[str]:
+    print(f"[CHECK] Reranker: model={model} base_url={base_url}")
+    url = base_url.rstrip("/") + "/rerank"
+    headers = {"Content-Type": "application/json"}
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
+    payload = {
+        "model": model,
+        "query": "付款期限是什么？",
+        "documents": [
+            "买方应在收到合法有效发票后30日内完成付款。",
+            "本合同自双方签字盖章之日起生效。",
+        ],
+    }
+    started = time.monotonic()
+    try:
+        response = requests.post(url, headers=headers, json=payload, timeout=120)
+        elapsed = time.monotonic() - started
+        if response.status_code >= 400:
+            return [f"Reranker failed with HTTP {response.status_code}: {response.text[:500]}"]
+        payload = response.json()
+        if not _has_rerank_results(payload):
+            return [f"Reranker returned no recognizable results: {payload}"]
+        print(f"[OK] Reranker: {elapsed:.2f}s response_keys={list(payload.keys())}\n")
+        return []
+    except Exception as exc:  # noqa: BLE001
+        return [f"Reranker failed: {exc}"]
+
+
+def _has_rerank_results(payload: dict[str, Any]) -> bool:
+    for key in ("results", "data"):
+        if isinstance(payload.get(key), list) and payload[key]:
+            return True
+    return False
+
+
+def require_value(config: dict[str, Any], key: str) -> str:
+    value = config.get(key)
+    if value in {None, ""}:
+        raise ValueError(f"Missing required config value: ragas.{key}")
+    return str(value)
+
+
+if __name__ == "__main__":
+    sys.exit(main())