feat(search): thread lang through NlSearchRequest → controller → NlQueryParserService → NlpClient

- NlSearchRequest gains @NotBlank @Pattern(regexp="de|en|es") lang field
- NlSearchController passes request.lang() to service
- NlQueryParserService.search signature: (String, String, Pageable); renames ollamaClient→nlpClient; removes redundant length guard (Bean Validation is enforcement point)
- application.yaml: replaces app.ollama.* with app.nlp.base-url; application-dev.yaml: points to localhost:8001
- frontend/documents/+page.svelte: sends lang: languageTag() in POST body

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/src/main/java/org/raddatz/familienarchiv/search/NlQueryParserService.java

@@ -34,18 +34,13 @@ public class NlQueryParserService {
     private static final int MIN_TAG_TERM = 3;
     private static final int MAX_RESOLVED_TAGS = 10;
 
-    private final OllamaClient ollamaClient;
+    private final NlpClient nlpClient;
     private final PersonService personService;
     private final DocumentService documentService;
     private final TagService tagService;
 
-    public NlSearchResponse search(String query, Pageable pageable) {
+    public NlSearchResponse search(String query, String lang, Pageable pageable) {
-        if (query == null || query.length() < MIN_QUERY || query.length() > MAX_QUERY) {
+        NlpExtraction ext = nlpClient.parse(query, lang);
-            throw DomainException.badRequest(ErrorCode.VALIDATION_ERROR,
-                    "Query must be between " + MIN_QUERY + " and " + MAX_QUERY + " characters");
-        }
-
-        OllamaExtraction ext = ollamaClient.parse(query);
 
         List<String> personNames = ext.personNames() != null ? ext.personNames() : List.of();
         List<String> keywords = ext.keywords() != null ? ext.keywords() : List.of();

backend/src/main/java/org/raddatz/familienarchiv/search/NlSearchController.java

@@ -23,6 +23,6 @@ public class NlSearchController {
                                    Pageable pageable,
                                    @AuthenticationPrincipal UserDetails principal) {
         rateLimiter.checkAndConsume(principal.getUsername());
-        return nlQueryParserService.search(request.query(), pageable);
+        return nlQueryParserService.search(request.query(), request.lang(), pageable);
     }
 }

backend/src/main/java/org/raddatz/familienarchiv/search/NlSearchRequest.java

@@ -1,11 +1,15 @@
 package org.raddatz.familienarchiv.search;
 
 import jakarta.validation.constraints.NotBlank;
+import jakarta.validation.constraints.Pattern;
 import jakarta.validation.constraints.Size;
 
 public record NlSearchRequest(
         @NotBlank
         @Size(min = 3, max = 500)
-        String query
+        String query,
+        @NotBlank
+        @Pattern(regexp = "de|en|es")
+        String lang
 ) {
 }

backend/src/main/resources/application-dev.yaml

@@ -13,5 +13,5 @@ springdoc:
     path: /swagger-ui.html
 
 app:
-  ollama:
+  nlp:
-    base-url: http://localhost:11434
+    base-url: http://localhost:8001

backend/src/main/resources/application.yaml

@@ -130,13 +130,8 @@ app:
     # The loader maps columns by header name — no positional indices (see ADR-025).
     dir: ${IMPORT_DIR:/import}
 
-  ollama:
+  nlp:
-    base-url: http://ollama:11434
+    base-url: http://nlp-service:8001
-    model: qwen2.5:7b-instruct-q4_K_M
-    # CPU inference: ~18s warm. Higher ceiling absorbs the cold model load on the
-    # first query after an Ollama (re)start before OLLAMA_KEEP_ALIVE pins it.
-    timeout-seconds: 60
-    health-check-timeout-seconds: 2
 
   nl-search:
     rate-limit:

frontend/src/routes/documents/+page.svelte

@@ -17,6 +17,7 @@ import { bulkSelectionStore } from '$lib/document/bulkSelection.svelte';
 import { getErrorMessage, parseBackendError } from '$lib/shared/errors';
 import { csrfFetch } from '$lib/shared/cookies';
 import * as m from '$lib/paraglide/messages.js';
+import { languageTag } from '$lib/paraglide/runtime';
 import type { components } from '$lib/generated/api';
 
 type NlQueryInterpretation = components['schemas']['NlQueryInterpretation'];
@@ -224,7 +225,7 @@ async function runSmartSearch() {
 		const res = await csrfFetch('/api/search/nl', {
 			method: 'POST',
 			headers: { 'Content-Type': 'application/json' },
-			body: JSON.stringify({ query })
+			body: JSON.stringify({ query, lang: languageTag() })
 		});
 		if (!res.ok) {
 			const backend = await parseBackendError(res);