added health endpoints

python/ebook_search/llm_interface.py

@@ -44,6 +44,36 @@ def request_embeddings(texts: Sequence[str], config: EbookSearchConfig) -> list[
         raise RuntimeError(msg) from error
 
 
+def check_embedding_endpoint(config: EbookSearchConfig, *, timeout_seconds: float = 5.0) -> bool:
+    """Return whether the configured embedding endpoint answers a model listing."""
+    try:
+        response = httpx.get(
+            f"{config.embedding_base_url.rstrip('/')}/models",
+            headers=auth_headers(config.embedding_api_key),
+            timeout=timeout_seconds,
+        )
+        response.raise_for_status()
+    except httpx.HTTPError as error:
+        logger.warning("ebook_embedding_endpoint_unreachable base_url=%s error=%s", config.embedding_base_url, error)
+        return False
+    return True
+
+
+def check_chat_endpoint(config: EbookSearchConfig, *, timeout_seconds: float = 5.0) -> bool:
+    """Return whether the configured chat (answering) endpoint answers a model listing."""
+    try:
+        response = httpx.get(
+            f"{config.vllm_base_url.rstrip('/')}/models",
+            headers=auth_headers(config.vllm_api_key),
+            timeout=timeout_seconds,
+        )
+        response.raise_for_status()
+    except httpx.HTTPError as error:
+        logger.warning("ebook_chat_endpoint_unreachable base_url=%s error=%s", config.vllm_base_url, error)
+        return False
+    return True
+
+
 def embedding_vectors_from_response(body: object) -> list[list[float]]:
     """Extract embedding vectors from an OpenAI-compatible embedding response."""
     if not isinstance(body, dict):