feat(presets): ready-made Opus 4.8 API preset (cli-agents openai-server bridge)

Assets/Resources/CoreAIPresets/CoreAISettings_OpusApi.asset: OpenAiHttp backend at
http://localhost:8801/v1 (the agent.sh openai-server default port), model 'opus' —
drop-in config for driving CoreAI with Claude Opus 4.8 through the local CLI bridge
without editing the live settings asset each time. README documents the one-line
bridge launch. (.meta files intentionally left for Unity to generate.)

Author: NeoXider

Assets/Resources/CoreAIPresets/CoreAISettings_OpusApi.asset

@@ -0,0 +1,93 @@
+%YAML 1.1
+%TAG !u! tag:unity3d.com,2011:
+--- !u!114 &11400000
+MonoBehaviour:
+  m_ObjectHideFlags: 0
+  m_CorrespondingSourceObject: {fileID: 0}
+  m_PrefabInstance: {fileID: 0}
+  m_PrefabAsset: {fileID: 0}
+  m_GameObject: {fileID: 0}
+  m_Enabled: 1
+  m_EditorHideFlags: 0
+  m_Script: {fileID: 11500000, guid: bc964192a1404b2458c3a69e06450221, type: 3}
+  m_Name: CoreAISettings_OpusApi
+  m_EditorClassIdentifier: CoreAI.Source::CoreAI.Infrastructure.Llm.CoreAISettingsAsset
+  backendType: 2
+  executionMode: 0
+  autoPriority: 0
+  apiBaseUrl: http://localhost:8801/v1
+  apiKey: 
+  modelName: opus
+  visionSupport: 0
+  enableTemperatureOverriding: 0
+  temperature: 0.1
+  maxTokens: 20000
+  maxClientLimitedRequestsPerSession: 0
+  maxClientLimitedPromptChars: 0
+  requestTimeoutSeconds: 240
+  enableFallbackBackend: 0
+  secondaryApiBaseUrl: 
+  secondaryApiKey: 
+  secondaryModelName: 
+  llmUnityAgentName: 
+  llmUnityAutoCreateRuntimeHost: 1
+  llmUnityRuntimeHostObjectName: 
+  llmUnityAutostartLocalServer: 1
+  ggufModelPath: Qwen3.5-4B-Q4_K_M.gguf
+  llmUnityDontDestroyOnLoad: 1
+  llmUnityStartupTimeoutSeconds: 120
+  llmUnityStartupDelaySeconds: 1
+  llmUnityKeepAlive: 0
+  reasoningMode: 0
+  thinkingBudgetTokens: 0
+  extraBodyJson: 
+  llmUnityMaxConcurrentChats: 1
+  llmUnityNumGPULayers: 99
+  llmUnityServerPort: 13333
+  universalSystemPromptPrefix: Respond concisely and to the point. Avoid unnecessary
+    verbosity.
+  toolContractAdditionalInstructions: 
+  maxLuaRepairRetries: 3
+  maxToolCallRetries: 3
+  allowDuplicateToolCalls: 0
+  maxLlmRequestRetries: 2
+  maxContextOverflowRetries: 3
+  contextWindowTokens: 128000
+  maxToolResultChars: 8000
+  defaultToolTimeoutMs: 30000
+  maxResponseChars: 0
+  maxToolCallRoundtrips: 10
+  maxToolCallHistoryMessages: 0
+  maxParallelToolCalls: 4
+  allowWorldPrimitives: 1
+  enableStreaming: 1
+  webGlNativeStreaming: 1
+  enableLuaOnWebGl: 1
+  sameOriginCredentials: 0
+  enableConversationHistorySummarization: 1
+  conversationHistoryRecentTokenBudgetOverride: 0
+  conversationRolledSummaryMaxTokens: 0
+  conversationCompactionTriggerRatio: 0.8
+  enableContextPruning: 1
+  maxRetainedToolResultMessages: 3
+  enableLlmContextCompaction: 0
+  enableTokenCalibration: 1
+  offlineUseCustomResponse: 0
+  offlineCustomResponse: 'Offline mode: LLM unavailable'
+  offlineCustomResponseRoles: '*'
+  enableMeaiDebugLogging: 1
+  enableHttpDebugLogging: 1
+  logLlmInput: 1
+  logLlmOutput: 1
+  logTokenUsage: 1
+  logLlmLatency: 1
+  logLlmConnectionErrors: 1
+  logToolCalls: 1
+  logToolCallArguments: 1
+  logToolCallResults: 1
+  logMeaiToolCallingSteps: 1
+  llmRequestTimeoutSeconds: 240
+  maxConcurrentOrchestrations: 2
+  logOrchestrationMetrics: 1
+  inputTokenPricePer1KUsd: 0
+  outputTokenPricePer1KUsd: 0

Assets/Resources/CoreAIPresets/README.md

@@ -0,0 +1,17 @@
+# CoreAI settings presets
+
+Ready-made `CoreAISettingsAsset` instances. To use one, either point your bootstrap at it,
+or copy its serialized values over `Assets/Resources/CoreAISettings.asset`
+(the Hub's AI Settings tab edits the live asset the same way).
+
+## CoreAISettings_OpusApi
+
+Claude Opus 4.8 through the local `cli-agents` OpenAI-compatible bridge:
+
+```bash
+# start the bridge (keeps running; one process = one model)
+bash ~/.claude/skills/cli-agents/agent.sh openai-server -e claude -m opus
+```
+
+The preset targets `http://localhost:8801/v1` (the bridge's default port) with
+`backendType = OpenAiHttp`. No API key needed — the bridge uses your local CLI login.