From de51cdad4362e9405ea2d1d977c2cc0c616d8eb5 Mon Sep 17 00:00:00 2001 From: Sam Date: Wed, 22 Jul 2026 12:55:22 +0200 Subject: [PATCH] chore(gitops): upgrade models to gemini-3.6-flash --- helm/echo/values-echo-next.yaml | 16 ++++++++-------- helm/echo/values-extended-env.yaml | 12 ++++++------ helm/echo/values.yaml | 2 +- 3 files changed, 15 insertions(+), 15 deletions(-) diff --git a/helm/echo/values-echo-next.yaml b/helm/echo/values-echo-next.yaml index 1e56b6a..ccd6795 100644 --- a/helm/echo/values-echo-next.yaml +++ b/helm/echo/values-echo-next.yaml @@ -4,7 +4,7 @@ # for echo-next without touching prod/testing (which do not reference this file). # # Swaps all three LLM groups (multimodal pro, multimodal fast, text fast) to -# Gemini 3.5 Flash on Vertex, EU-resident. +# Gemini 3.6 Flash on Vertex, EU-resident. # # Gemini 3.x is only served on the GLOBAL Vertex hostname # (aiplatform.googleapis.com), not the regional eu-aiplatform.googleapis.com @@ -15,28 +15,28 @@ # # The api_base hardcodes project + model because litellm appends ":generateContent" # to it verbatim. This file is echo-next-only, so the dembrane-echo project is correct. -# The _2/_3 regional failovers stay disabled (3.5 Flash is global-host only). +# The _2/_3 regional failovers stay disabled (3.6 Flash is global-host only). common: env: - LLM__MULTI_MODAL_PRO__MODEL: "vertex_ai/gemini-3.5-flash" + LLM__MULTI_MODAL_PRO__MODEL: "vertex_ai/gemini-3.6-flash" LLM__MULTI_MODAL_PRO__VERTEX_LOCATION: "eu" - LLM__MULTI_MODAL_PRO__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.5-flash" + LLM__MULTI_MODAL_PRO__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.6-flash" LLM__MULTI_MODAL_PRO_2__MODEL: "" LLM__MULTI_MODAL_PRO_2__VERTEX_LOCATION: "" LLM__MULTI_MODAL_PRO_3__MODEL: "" LLM__MULTI_MODAL_PRO_3__VERTEX_LOCATION: "" - LLM__MULTI_MODAL_FAST__MODEL: "vertex_ai/gemini-3.5-flash" + LLM__MULTI_MODAL_FAST__MODEL: "vertex_ai/gemini-3.6-flash" LLM__MULTI_MODAL_FAST__VERTEX_LOCATION: "eu" - LLM__MULTI_MODAL_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.5-flash" + LLM__MULTI_MODAL_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.6-flash" LLM__MULTI_MODAL_FAST_2__MODEL: "" LLM__MULTI_MODAL_FAST_2__VERTEX_LOCATION: "" LLM__MULTI_MODAL_FAST_3__MODEL: "" LLM__MULTI_MODAL_FAST_3__VERTEX_LOCATION: "" - LLM__TEXT_FAST__MODEL: "vertex_ai/gemini-3.5-flash" + LLM__TEXT_FAST__MODEL: "vertex_ai/gemini-3.6-flash" LLM__TEXT_FAST__VERTEX_LOCATION: "eu" - LLM__TEXT_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.5-flash" + LLM__TEXT_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.6-flash" LLM__TEXT_FAST_2__MODEL: "" LLM__TEXT_FAST_2__VERTEX_LOCATION: "" LLM__TEXT_FAST_3__MODEL: "" diff --git a/helm/echo/values-extended-env.yaml b/helm/echo/values-extended-env.yaml index aa62286..013bd2e 100644 --- a/helm/echo/values-extended-env.yaml +++ b/helm/echo/values-extended-env.yaml @@ -13,17 +13,17 @@ # 2026-06-08 (model swap + AssemblyAI webhook round-trip). common: env: - LLM__MULTI_MODAL_PRO__MODEL: "vertex_ai/gemini-3.5-flash" + LLM__MULTI_MODAL_PRO__MODEL: "vertex_ai/gemini-3.6-flash" LLM__MULTI_MODAL_PRO__VERTEX_LOCATION: "eu" - LLM__MULTI_MODAL_PRO__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.5-flash" + LLM__MULTI_MODAL_PRO__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.6-flash" - LLM__MULTI_MODAL_FAST__MODEL: "vertex_ai/gemini-3.5-flash" + LLM__MULTI_MODAL_FAST__MODEL: "vertex_ai/gemini-3.6-flash" LLM__MULTI_MODAL_FAST__VERTEX_LOCATION: "eu" - LLM__MULTI_MODAL_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.5-flash" + LLM__MULTI_MODAL_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.6-flash" - LLM__TEXT_FAST__MODEL: "vertex_ai/gemini-3.5-flash" + LLM__TEXT_FAST__MODEL: "vertex_ai/gemini-3.6-flash" LLM__TEXT_FAST__VERTEX_LOCATION: "eu" - LLM__TEXT_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.5-flash" + LLM__TEXT_FAST__API_BASE: "https://aiplatform.googleapis.com/v1/projects/dembrane-echo/locations/eu/publishers/google/models/gemini-3.6-flash" EMBEDDING_MODEL: "vertex_ai/text-embedding-004" EMBEDDING_BASE_URL: "https://europe-west4-aiplatform.googleapis.com" diff --git a/helm/echo/values.yaml b/helm/echo/values.yaml index 8b3926f..50b2587 100644 --- a/helm/echo/values.yaml +++ b/helm/echo/values.yaml @@ -91,7 +91,7 @@ agent: memory: "1Gi" env: ECHO_API_URL: "http://echo-api:8000/api" - LLM_MODEL: "gemini-3.5-flash" + LLM_MODEL: "gemini-3.6-flash" VERTEX_LOCATION: "eu" # Gemini 3.x is global-host only; locations/eu stays in the request path. VERTEX_API_ENDPOINT: "aiplatform.googleapis.com"