ci-static / secret-scan (push) Successful in 21s
ci-static / yamllint (push) Successful in 1m13s
ci-static / ansible-syntax (push) Successful in 29s
ci-static / markdown-links (push) Successful in 20s
ci-static / secret-scan (pull_request) Successful in 21s
ci-static / yamllint (pull_request) Successful in 33s
ci-static / ansible-syntax (pull_request) Successful in 27s
ci-static / markdown-links (pull_request) Successful in 21s
Root causes des bisherigen Haengers: ENABLE_WEB_SEARCH fehlte (Default False in v0.13.3), persistierte config-Tabelle schlug die Env-Werte (web.search.enable=false, engine=""), ENABLE_RAG_WEB_SEARCH ist in v0.13.3 tot, SearXNG lief mit 123 Default-Engines, und meta.defaultFeatureIds des Modells war leer (neue UI-Chats starten ohne Web-Search-Toggle). - web_search-Rolle: settings.yml mit keep_only (43 kuratierte Engines, Namen aus der Default-settings.yml des gepinnten Images 626105eec), alle Engines auf categories:[general] + disabled:false (OpenWebUI sendet keinen categories-Parameter), outgoing.request_timeout 2.5s/max 8s, suspended_times auf Minuten verkuerzt; Probe assertet echte JSON-results. - open_webui-Rolle: ENABLE_WEB_SEARCH=True, RESULT_COUNT=8, WEB_SEARCH_CONCURRENT_REQUESTS=4, WEB_FETCH_MAX_CONTENT_LENGTH=100000, USER_AGENT (Browser-String); ENABLE_RAG_WEB_SEARCH entfernt; Quadlet-Template quotet Environment="KEY=value" (systemd-Split bei Spaces, betraf u.a. OAUTH_SCOPES). - DB-Sync idempotent: config-rows web.search.*/web.fetch.* an Env angleichen (DB gewinnt ueber Env), Model-Meta capabilities.web_search + builtinTools.web_search=true + defaultFeatureIds=["web_search"], Workspace-Tool gnai-web-search (web_search/fetch_url gegen http://searxng:8080) mit public read grant fuer API-Agenten via tool_ids. - Doku GNAI-020 (Umsetzung, Verifikation, UI-/API-Pfade Path A/B, NPM-Timeout-Betreiber-Schritt); Index + Roadmap M6 aktualisiert. Verifiziert 12.11.2026 auf 10.0.0.90: SearXNG 168 Ergebnisse/1,4 s, aus OpenWebUI-Container via Alias HTTP 200/1,9 s; /api/models zeigt defaultFeatureIds=['web_search']; E2E Path A ohne tool_ids (nur session_id + features.web_search): server-seitiger Loop search_web x2 + fetch_url x2, Zitation in sources, korrekte Antwort mit Quellen (v0.13.3, 2026-11-08); Playbook danach changed=0; secret-scan/yamllint/ansible-syntax/links OK. Generated by Mistral Vibe. Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
LLM serving stack
This directory will hold the executable infrastructure for the LLM serving
stack on ghostnet-ki: rootless llama.cpp with ROCm, separated into
Production and Testing environments.
It is created with the internal LLM vertical slice (roadmap milestone 2 in
GNAI-001). The planned
architecture, deployment and serving configuration are documented under
../../docs/platform/llm-server/ai/.