Reference & FAQ / Configuration

Configuration

The LingXizhi kernel reads its configuration from lxz.config.json (JSON format) at startup. All configuration keys, descriptions and defaults are listed below by module.

!

Before you edit: server.auth_token is generated automatically by the kernel — do not modify it by hand or commit it to version control. privacy.exclude_files and safety.blocked_paths protect your code and credentials — be sure of the consequences before clearing them.

Agent & LLM

KeyDescriptionDefault
agent.auto_executeAgent executes automatically without step-by-step confirmationfalse
agent.confirmation_timeout_secConfirmation timeout for dangerous operations (seconds)300
agent.max_context_tokensAgent context limit (tokens)128000
agent.max_iterationsMax iterations per task (hard limit)50
agent.parallel_max_workersMax parallel Coder agents in RUSH4
agent.subagent_max_iterationsMax iterations for subagents20
llm.max_tokensMax output tokens per request4096
llm.temperatureSampling temperature0.7
llm.timeout_secRequest timeout (seconds)60
llm.retry_count / llm.retry_delay_msRetry count / interval on failure (milliseconds)3 / 1000
llm.stream_chunk_sizeStreaming chunk size64
model_config.sourceSource of model configuration"go"
model_config.server_urlModel configuration service URLhttp://localhost:9527/api/v1
model_config.cache_ttl_secondsModel configuration cache TTL (seconds)300
model_config.offline_fallback_enabledOffline fallback switchfalse

About RUSH: iteration behavior is governed by agent.max_iterations (hard limit) together with a built-in convergence threshold of 0.95 — the current route terminates and re-plans when the same error occurs 3 times in a row or the limit is reached; parallelism is controlled by agent.parallel_max_workers. See the RUSH Guide.

Cache & Memory

KeyDescriptionDefault
cache.enabledMaster switch for the three-layer cachetrue
cache.l1_capacityL1 exact-cache capacity (entries)256
cache.l2_similarity_thresholdL2 similarity-cache hit threshold0.92
cache.l3_max_templatesL3 template cache limit100
cache.preheat_on_startup / cache.preheat_top_filesStartup preheating switch / number of files to preheattrue / 20
cache.ttl_secondsCache expiry (seconds)3600
memory.enabledPersistent memory switchtrue
memory.auto_learnAutomatically learn project patternstrue
memory.db_pathMemory database path.lxz/memory.db
memory.max_entriesMemory entry limit10000
memory.context_assembly_max_tokensMemory injection context limit4000
memory.semantic_index_enabled / memory.wal_modeSemantic retrieval index / SQLite WAL modetrue / true

IDE UI & Language

KeyDescriptionDefault
ide.themeTheme (dark / light)"dark"
ide.layoutUI layout"trae_cn"
ide.chat_position / ide.sidebar_positionAI chat / sidebar position"right_panel" / "right"
ide.model_selector_styleModel picker style"trae_cup"
ide.completion_debounce_msCompletion trigger debounce (milliseconds)150
ide.inline_edit_context_linesContext lines carried by inline edit5
ide.diagnostic_refresh_on_saveRefresh diagnostics on savetrue
i18n.language / i18n.fallback_languageUI language / fallback language"zh" / "en"

LSP Language Services

KeyDescriptionDefault
lsp.enabledLSP master switchtrue
lsp.modeRun mode (built-in first, fallback on failure)"builtin_with_fallback"
lsp.builtin_servers_dirBuilt-in language servers directory"lsp-servers"
lsp.max_concurrent_processes / lsp.max_per_workspaceGlobal concurrent process limit / per-workspace limit10 / 5
lsp.cpu_limit_percentCPU usage limit (%)50
lsp.memory_limit_mbMemory limit (MB)1024
lsp.idle_timeout_minutesIdle reclaim time (minutes)10
lsp.initialize_timeout_msInitialize timeout (milliseconds)30000
lsp.fallback_timeout_msFallback timeout (milliseconds)2000
lsp.disabled_languagesLanguages with LSP disabled[]
lsp.disabled_for_remoteDisabled in remote workspacestrue
lsp.language_overridesPer-language server config overrides{}
lsp.sandbox_modeRun language servers in sandbox modefalse

Security & Privacy

KeyDescriptionDefault
safety.enabledSafety protection master switchtrue
safety.snapshot_before_editAutomatic snapshot before editstrue
safety.auto_recover_on_failureAutomatic recovery on failuretrue
safety.dangerous_commands_require_confirmationDangerous commands require confirmationtrue
safety.max_file_size_kbMax single-file processing size (KB)512
safety.blocked_pathsPaths the AI is forbidden to read/writenode_modules, .git, target, __pycache__
privacy.modePrivacy mode"standard"
privacy.exclude_filesSensitive files never uploaded.env, .env.local, credentials.json, secrets.yaml
privacy.redact_patternsRegexes redacted before transmission (passwords, keys, Bearer, card numbers, emails)5 built-in rules

Sandbox, Cost & Benchmark

KeyDescriptionDefault
sandbox.enabledLocal sandbox switchfalse
sandbox.typeSandbox type"process"
sandbox.docker_imageDocker sandbox image"lingxizhi/sandbox:latest"
sandbox.max_cpu_percent / sandbox.max_memory_mbCPU (%) / memory (MB) limits80 / 512
sandbox.network_enabledNetworking inside the sandboxfalse
sandbox.timeout_secExecution timeout (seconds)300
cost.budget_usd_daily / cost.budget_usd_monthlyDaily / monthly cost budget (USD, self-managed under BYOK)10.0 / 200.0
cost.track_by_modelTrack costs per modeltrue
cost.warning_threshold_percentBudget warning threshold (%)80
benchmark.categoriesBenchmark categoriesfilesystem, cache, memory
benchmark.output_dirBenchmark output directory.lxz/benchmarks
benchmark.run_on_startupRun benchmarks automatically at startupfalse

Integrations, Plugins & Hooks

KeyDescriptionDefault
integrations.git_platformDefault Git platform"gitee"
integrations.gitee_access_token / github_access_token / gitlab_access_tokenAccess tokens for each platform""
integrations.notificationsNotification channel configuration[]
plugins.enabledPlugin system switchtrue
plugins.auto_updateAutomatic plugin updatesfalse
plugins.marketplace_urlPlugin marketplace URLhttps://marketplace.lingxizhi.dev/api/v1
plugins.search_dirsPlugin search directories.lxz/extensions
hooks.enabledHooks switchtrue
hooks.pre_edit / post_editPre-edit / post-edit hook commands[]
hooks.pre_commit / post_commitPre-commit / post-commit hook commands[]
hooks.on_errorOn-error hook command[]

Kernel Server

KeyDescriptionDefault
server.hostListen address (local loopback only)127.0.0.1
server.portListen port (auto-increments when occupied, up to 10)3721
server.auth_tokenLocal call token (auto-generated)Generated at startup
server.cors_originsAllowed CORS originshttp://localhost:*, vscode-webview://*
server.max_body_size_mbRequest body limit (MB)50
server.request_timeout_secRequest timeout (seconds)120

Next Steps

  • How to call each kernel endpoint: Kernel API.
  • Configuration not taking effect or kernel issues: FAQ.