diff --git a/llm-server/llm/ansible/inventories/production/group_vars/llm_servers.yml b/llm-server/llm/ansible/inventories/production/group_vars/llm_servers.yml new file mode 100644 index 0000000..cbe1841 --- /dev/null +++ b/llm-server/llm/ansible/inventories/production/group_vars/llm_servers.yml @@ -0,0 +1,61 @@ +--- +# Environment-specific values for the Production stack (GNAI-008 section +# 7.4, section 7.2 isolation table; GNAI-001 milestone 4). Same role logic +# as Testing with its own workload account, quadlet directory, unit prefix +# and network. Production deploys only the approved reference profile +# (Swift 1.5 Q6_K, operator decision 2026-11-05, GNAI-009 10.4) plus the +# rollback profile (Unsloth UD-Q6_K); profile parameters are kept in sync +# with the role defaults until the promotion chain (GNAI-008 section 12) +# automates the release. +inference_engine_user: gnai-production +inference_engine_uid: 980 +inference_engine_user_home: /var/lib/ghostnet-ai/production +inference_engine_unit_prefix: gnai-production-llama +inference_engine_network_name: gnai-production-llama +inference_engine_network_alias: llama-cpp + +# Approved reference and rollback profile only (see the comment above). +inference_engine_profiles: + swift-1.5-qwen3.8-27b-q6k: + unit_suffix: swift15 + description: Swift 1.5 reasoning-efficient finetune of Qwen3.8-27B (Q6_K) + model_directory: swift-1.5-qwen3.8-27b + model_file: Swift-1.5-Qwen3.8-27B-Q6_K.gguf + alias: swift-1.5-qwen3.8-27b-q6k + ctx_size: 131072 + n_predict: 32768 + parallel: 1 + reasoning_effort: xhigh + temp: 1.0 + top_p: 0.95 + top_k: 20 + min_p: 0.0 + presence_penalty: 0.0 + repeat_penalty: 1.0 + mtp: + enabled: false + spec_draft_n_max: 2 + unsloth-qwen3.8-27b-ud-q6k: + unit_suffix: unsloth + description: Unsloth UD quant of Qwen3.8-27B (Q6_K) + model_directory: unsloth-qwen3.8-27b + model_file: Qwen3.8-27B-UD-Q6_K.gguf + alias: unsloth-qwen3.8-27b-ud-q6k + ctx_size: 131072 + n_predict: 32768 + parallel: 1 + reasoning_effort: xhigh + temp: 1.0 + top_p: 0.95 + top_k: 20 + min_p: 0.0 + presence_penalty: 0.0 + repeat_penalty: 1.0 + mtp: + enabled: false + spec_draft_n_max: 2 + +inference_engine_active_profile: swift-1.5-qwen3.8-27b-q6k + +# 128k target context, same as Testing (GNAI-010 section 10.9). +inference_engine_ctx_size_override: null diff --git a/llm-server/llm/ansible/inventories/production/hosts.yml b/llm-server/llm/ansible/inventories/production/hosts.yml new file mode 100644 index 0000000..8c61b48 --- /dev/null +++ b/llm-server/llm/ansible/inventories/production/hosts.yml @@ -0,0 +1,7 @@ +all: + children: + llm_servers: + hosts: + ghostnet-ki: + ansible_host: 10.0.0.90 + ansible_user: ansible-deploy