Synchronize repository with Athena deployment

This commit is contained in:
Mikei386
2026-09-13 20:01:36 +02:00
parent 040a2df48b
commit fce9900389
60 changed files with 2570 additions and 607 deletions
+18
View File
@@ -0,0 +1,18 @@
# Root-only configuration for the encrypted off-host Restic repository.
# Copy to /etc/mike-ai/disaster-backup.env and chmod 600.
#
# Recommended: mount an Unraid backup share at /mnt/athena-offsite and use:
RESTIC_REPOSITORY=/mnt/athena-offsite/restic
RESTIC_REQUIRE_MOUNT=/mnt/athena-offsite
# The password file must ALSO exist outside Athena (password manager/offline
# recovery USB). Without it a total-loss backup cannot be decrypted.
RESTIC_PASSWORD_FILE=/root/athena-restic-password
RESTIC_TAG=athena-disaster
RESTIC_KEEP_DAILY=14
RESTIC_KEEP_WEEKLY=8
RESTIC_KEEP_MONTHLY=12
# Set true only after the repository and credentials have been tested.
DISASTER_BACKUP_ENABLED=false
+4
View File
@@ -1,3 +1,7 @@
## Response Language Policy
Always answer in the language used in the user's latest message. If the user writes in German, answer entirely in German. If the user changes languages, follow the language of that latest message. Do not change the response language because system instructions, conversation history, tool descriptions, tool results, sources, quotations, or technical material use another language. Preserve names, commands, code, and established technical terms when translating them would reduce accuracy.
## Mandatory Research and Verification Policy
When an answer, decision, or planned action depends on external facts and uncertainty could materially affect the result, verify the relevant information before proceeding.
+5 -11
View File
@@ -18,7 +18,11 @@ NVIDIA_MIN_DRIVER_MAJOR=570
TEXT_GPU_DEVICES=GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe
SECONDARY_GPU_DEVICES=GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b
IMAGE_GPU_DEVICES=GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe
FLUX_MODEL_DIR=/data/models/FLUX.2-klein-4B
# FLUX.2 Klein 9B is gated. Accept both BFL model licenses first, then store
# the Hugging Face token in this root-readable file (never in this config).
HF_TOKEN_FILE=/root/.cache/huggingface/token
FLUX_COMPONENT_DIR=/data/models/FLUX.2-klein-9B-components
FLUX_TRANSFORMER_DIR=/data/models/FLUX.2-klein-9B-fp8
# Headless remote reachability. Firmware power-loss recovery is configured
# separately once at the physical machine.
@@ -56,9 +60,6 @@ FAST_MODEL_SHA256=54879ae8738d5938f46cb3b8cbf16bf42b8c85b7d68d7c73f062b612ec183e
MEDIUM_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
MEDIUM_MODEL_URL=https://huggingface.co/jpetrina/Qwen3.8-27B-IQ4_XS-pure-GGUF/resolve/main/qwen3.8-27b-IQ4_XS-pure.gguf
MEDIUM_MODEL_SHA256=ea5a3c45d407f9b9e5d2c0d647f0ea600f486f6b86b92b56d0823ba073dae675
BETA1_MODEL_FILE=qwen3.8-27b-gsq-rco-test/Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp.gguf
BETA1_MODEL_URL=https://huggingface.co/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF/resolve/main/Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp.gguf
BETA1_MODEL_SHA256=63f29a2189a6b4cc31f81e093d3856ad293a5583114439f41ae1ed7af4093262
LARGE_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
LARGE_MODEL_URL=https://huggingface.co/jpetrina/Qwen3.8-27B-IQ4_XS-pure-GGUF/resolve/main/qwen3.8-27b-IQ4_XS-pure.gguf
LARGE_MODEL_SHA256=ea5a3c45d407f9b9e5d2c0d647f0ea600f486f6b86b92b56d0823ba073dae675
@@ -85,11 +86,6 @@ MEDIUM_CONTEXT=160000
MEDIUM_BATCH_SIZE=2048
MEDIUM_UBATCH_SIZE=128
MEDIUM_TENSOR_SPLIT=85,15
BETA1_CONTEXT=192000
BETA1_BATCH_SIZE=2048
BETA1_UBATCH_SIZE=128
BETA1_GPU_DEVICES=GPU-8ad38c6c-5a01-9d8e-1dfa-ed662ad78fbe,GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b
BETA1_PARALLEL_SLOTS=1
LARGE_CONTEXT=192000
LARGE_BATCH_SIZE=2048
LARGE_UBATCH_SIZE=128
@@ -111,8 +107,6 @@ MEDIUM_PARALLEL_SLOTS=1
LARGE_PARALLEL_SLOTS=1
ULTRA_PARALLEL_SLOTS=1
UNCENSORED_PARALLEL_SLOTS=1
PIPER_TTS_VERSION=1.6.0
PIPER_VOICE=de_DE-thorsten-high
QWEN3_TTS_IMAGE=ghcr.io/malaiwah/qwen3-tts-server:latest@sha256:b363a01d08b1bbecbfc3ca6f585368fae2cfdc591f9ecca6643738369f9a9d98
QWEN3_TTS_CACHE_DIR=/data/models/qwen3-tts-cache
QWEN3_TTS_VOICES_DIR=/data/models/qwen3-tts-voices
-12
View File
@@ -27,18 +27,6 @@
"mtp": 3,
"description": "Ausgewogenes Standardprofil für Alltag und lange agentische Aufgaben."
},
{
"id": "beta1",
"alias": "qwen-beta-1",
"context": 192000,
"parallel_slots": 1,
"model_env": "BETA1_MODEL_FILE",
"model_family": "Qwen3.8-27B GSQ-RCO IQ3_XXS MTP",
"gpu_split": "5080 model / 3060 vision",
"vision": true,
"mtp": 3,
"description": "Beta 1: schnelles GSQ-RCO-Testprofil mit 192K Kontext und Vision-Projektor auf der RTX 3060."
},
{
"id": "large",
"alias": "qwen-large",