Menu
Complete catalogue of environment variables consumed by the LingTai kernel, wrapper, bundled MCPs, adapters, tests, and daemon composition surfaces.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Read-only health diagnostics for LingTai agents and bots. Use when an agent appears offline or unreachable, when a machine migration may have left stale MCP/addon command paths, when heartbeat/status/process/notification/log surfaces disagree, or before deciding whether to mail, refresh, CPR, or edit persistent configuration. Includes a bundled doctor.py script for layered local checks without exposing secrets.
Second-layer router for LingTai's progressive-disclosure operating manuals. Read this when resident substrate/procedures are too compact and you need the right lower reference. Route here for the expanded substrate/runtime model, lifecycle and `system` tool actions, the init.json composition and preset runtime model route, action/procedure discipline and skill routing, tool-result summarization, SQLite/`log.sqlite` trace inspection and trajectory mining, runtime/kernel update checks and nudges, environment variables, goal notifications, molt/memory questions, MCP/addon ownership, collaboration topology, and resident prompt design. The nested catalog below names each reference and its exact trigger.
Mandatory repository-local development guide for LingTai's Python kernel and runtime. Use this before changing code, architecture documents, tests, packaging, capabilities, adapters, or developer documentation in lingtai-kernel. Routes each task through the exact baseline, the distributed ANATOMY/CONTRACT systems, focused validation, and pull-request safety gates.
Operational guide for LingTai's built-in file tools: read, write, edit, glob, and grep. Use when working with local text files, deciding whether to use file tools versus bash, handling large files, avoiding binary/image misuse, or reading non-UTF-8 text via explicit bash/Python/iconv instead of complicating the core read tool. Covers UTF-8 policy, safe write/edit discipline, search workflows, and examples for GBK/Shift-JIS/Latin-1 conversion.
Router for LingTai's distributed code navigation system. Use this before navigating, creating, or maintaining kernel ANATOMY.md files. The normative anatomy-of-anatomy now lives at the repository-root ANATOMY.md; this skill explains how an agent enters and descends that graph, how it pairs with the distributed CONTRACT.md interface-definition graph, and what to do when code and navigation disagree.
Thin routing manual for NoKV-controlled workbenches. Use when an agent is asked to persist task inputs, scripts, outputs, logs, provenance, or run manifests through the `workbench_*` MCP tools instead of ordinary local file writes. Covers MCP registration, directory layout, append and edit discipline, conditional reads, cross-workbench queries, commit discipline, and leased checkpoint snapshots with naming, renewal, discovery, and point-in-time history reads.
| name | environment-variables |
| description | Complete catalogue of environment variables consumed by the LingTai kernel, wrapper, bundled MCPs, adapters, tests, and daemon composition surfaces. |
| version | 1.0.0 |
| last_changed_at | 2026-07-16T00:00:00-07:00 |
| related_files | ["ANATOMY.md","src/lingtai/intrinsic_skills/system-manual/SKILL.md","src/lingtai/prompts/substrate/substrate.md","src/lingtai/prompts/procedures/procedures.md","src/lingtai/kernel/nudge/__init__.py","src/lingtai/kernel/nudge/ANATOMY.md","src/lingtai/kernel/config_resolve.py","src/lingtai/kernel/refresh_watcher/MANUAL.md","src/lingtai/ANATOMY.md","src/lingtai/adapters/posix/ANATOMY.md","src/lingtai/adapters/windows/ANATOMY.md","src/lingtai/auth/ANATOMY.md","src/lingtai/kernel/ANATOMY.md","src/lingtai/kernel/base_agent/ANATOMY.md","src/lingtai/llm/openai/ANATOMY.md","src/lingtai/mcp_servers/ANATOMY.md","src/lingtai/services/ANATOMY.md","src/lingtai/tools/daemon/ANATOMY.md","src/lingtai/tools/soul/ANATOMY.md"] |
| maintenance | Keep this catalogue complete for every LINGTAI_* variable read by production source or intentionally exposed test/daemon composition. For each new read, add name, purpose, default, accepted values, scope, read point, reload/restart behavior, invalid-value behavior, implementation anchor, and security note. Keep the router/progressive-disclosure links reciprocal; do not paste this catalogue into resident substrate or procedures. |
This is the deep reference for environment configuration. Resident prompt layers
only explain the two global Nudge axes. Values are read by the component named in
the table; changing a variable does not imply that an already-running component
reloads it. Never put secrets directly in an environment value when a supported
*_env/env_file configuration path can be used.
| Variable | Purpose / default | Accepted values and scope | Read/reload and invalid behavior | Anchor / security |
|---|---|---|---|---|
LINGTAI_NUDGE_ENABLED | Global Nudge publication switch; default on. | on/off (also true/false/1/0), all Nudge kinds for one process/workdir. | Read at each Nudge operation/heartbeat; no restart. Invalid falls back to on and emits a bounded diagnostic. | src/lingtai/kernel/nudge/__init__.py; disabling reminders is not an authorization boundary. |
LINGTAI_NUDGE_REPEAT_INTERVAL | Post-dismiss repeat interval for the same unresolved finding; default 24h. | Positive duration with s, m, h, or d, for example 30m, 24h, 2d; global. | Read at each Nudge operation; no restart. Invalid/zero/negative falls back to 24h and emits a bounded diagnostic. | src/lingtai/kernel/nudge/__init__.py; values affect reminder timing only, not resolution. |
LINGTAI_ACTIVE_STUCK_THRESHOLD_S | ACTIVE no-progress watchdog threshold; default 600 seconds. | Numeric seconds; kernel clamps values below 30 to 30. | Read when the watchdog evaluates a turn; next evaluation sees a changed value. Invalid falls back to 600. | src/lingtai/kernel/base_agent/lifecycle.py; tuning a watchdog can change availability, so operators should record it. |
LINGTAI_REFRESH_ENV_OVERWRITE | One refresh handoff lets edited env_file values replace inherited process values; default unset/off. | 1 enables the one-shot overwrite. | Read on boot/refresh setup; consumed/removed after use. Other values are treated as off. | src/lingtai/cli.py, src/lingtai/agent.py; do not log env-file contents. |
LINGTAI_RUNTIME_PYTHON | Interpreter used for runtime/self-check and host-tool routing; default unset (caller supplies sys.executable). | Executable path. | Read by consumers when invoked; restart/relaunch the consumer to change its process interpreter. Invalid/missing is a caller/configuration error, not a fallback to a remote runtime. | src/lingtai/cli.py and runtime checks; path is not a credential. |
LINGTAI_RUNTIME_VENV | Active runtime virtualenv hint; default unset. | Local directory path. | Read by host tools when invoked; change takes effect for a new tool/process. Missing is tolerated where the caller has another interpreter. | src/lingtai/cli.py; never infer package freshness from an unrelated shell Python. |
LINGTAI_AGENT_DIR | Workdir for an out-of-process MCP/client; default unset and normally injected by the launcher. | Existing local directory. | Read at MCP process start; restart the MCP process after changing. Invalid path fails the MCP/client operation. | src/lingtai/mcp_servers/_config.py; local path only, do not expose private workdir contents. |
LINGTAI_MCP_NAME | Name of the active MCP server; default unset. | Registered MCP name. | Read at MCP process start; restart after changing. Invalid/missing fails closed rather than selecting an arbitrary server. | src/lingtai/mcp_servers/_config.py; keeps one process bound to one server identity. |
LINGTAI_ENV_VERSION | Environment/config schema marker used by environment setup; default unset. | Version string. | Read by the setup caller; relaunch/setup after changing. Invalid is reported by the setup path. | src/lingtai/venv_resolve.py; do not treat a marker as proof of package trust. |
LINGTAI_TUI_DIR | TUI-managed project/config root hint; default unset. | Local directory. | Read by TUI-facing callers at invocation; restart caller after changing. Invalid path fails the caller's lookup. | src/lingtai/venv_resolve.py/TUI integration; path may reveal local layout. |
LINGTAI_ORIGINATOR | Non-secret origin label for launched work; default unset. | Short string. | Read at process construction; restart/relaunch to change. Invalid is treated as absent. | src/lingtai/kernel launch surfaces; do not use it for authorization. |
LINGTAI_FILE_IO_BACKEND | Selects file-I/O backend; default implementation backend. | Backend name recognized by the wrapper. | Read when file-I/O service is constructed; rebuild/restart service. Unknown value fails closed. | src/lingtai/services/file_io.py (where available); backend selection is not a sandbox. |
LINGTAI_FILE_IO_SIDECAR | File-I/O sidecar executable/path hint; default unset. | Local executable/path. | Read at service construction; restart service. Invalid/unavailable sidecar fails the selected path. | File-I/O service adapter; do not grant sidecar broader filesystem authority than intended. |
LINGTAI_SEARCH_SIDECAR | Search sidecar executable/path hint; default unset (pure-Python fallback where supported). | Local executable/path. | Read when search service is constructed; restart service. Invalid sidecar uses the documented fallback or fails the operation; it never downloads one. | Search service adapter; validate executable ownership before use. |
LINGTAI_SKIP_RUST_BUILD | Developer/package-build switch to skip the Rust search-sidecar build; default unset/off. | 1 enables skip in build/test setup. | Read by build setup at build invocation; rerun the build after changing. Invalid values are treated as off. | setup.py / wheel tests; do not ship a production wheel claiming a sidecar it did not build. |
LINGTAI_REQUIRE_RUST_BUILD | Developer/package-build switch requiring the Rust search-sidecar build; default unset/off. | 1 enables requirement in build/test setup. | Read by build setup at build invocation; rerun the build after changing. Invalid values fail the required-build path. | setup.py / wheel tests; build policy is not runtime authorization. |
LINGTAI_SOUL_FLOW_ENABLED | Enables optional soul-flow capability; default disabled unless the host enables it. | 1/0 (implementation also accepts the component's documented boolean forms). | Read when soul capability is bootstrapped; refresh/restart capability after changing. Invalid is treated as disabled. | src/lingtai/tools/soul; not a command-execution or approval switch. |
LINGTAI_TASK_CARD_MAX_TOOL_ROWS | Bounds rendered Telegram task-card tool rows; default component limit. | Positive integer. | Read by the Telegram manager when rendering a card; next render sees valid changes. Invalid falls back to the component default. | src/lingtai/mcp_servers/telegram; bounds output size, not message authority. |
LINGTAI_PROFILE_MIME | Optional MIME/profile hint for bundled skill/profile transport; default unset. | MIME string. | Read when the profile resource is loaded; reload resource after changing. Invalid is ignored/fails the resource lookup. | src/lingtai profile loader; never use it to bypass content validation. |
LINGTAI_SKILL_MIME | Optional MIME hint for skill transport; default unset. | MIME string. | Read at skill-resource load; reload resource after changing. Invalid is ignored/fails lookup. | src/lingtai skill loader; content still requires normal skill validation. |
LINGTAI_CLOUD_MAIL_CONFIG | Cloud-mail configuration path/name; default unset. | Local config reference. | Read when cloud-mail integration starts; restart integration after changing. Invalid config fails closed. | Cloud mail adapter; config may contain credentials, so keep file permissions private. |
LINGTAI_IMAP_CONFIG | IMAP MCP config path/name; default unset. | Local config reference. | Read when IMAP MCP starts; restart MCP after changing. Invalid config fails closed. | src/lingtai/mcp_servers/imap; do not print config or password. |
LINGTAI_FEISHU_CONFIG | Feishu MCP config path/name; default unset. | Local config reference. | Read when Feishu MCP starts; restart MCP after changing. Invalid config fails closed. | src/lingtai/mcp_servers/feishu; keep app secrets out of logs. |
LINGTAI_TELEGRAM_CONFIG | Telegram MCP config path/name; default unset. | Local config reference. | Read when Telegram MCP starts; restart MCP after changing. Invalid config fails closed. | src/lingtai/mcp_servers/telegram; protect bot tokens and chat routing. |
LINGTAI_WECHAT_CONFIG | WeChat MCP config path/name; default unset. | Local config reference. | Read when WeChat MCP starts; restart MCP after changing. Invalid config fails closed. | src/lingtai/mcp_servers/wechat; protect credentials and contact state. |
LINGTAI_WHATSAPP_CONFIG | WhatsApp MCP config path/name; default unset. | Local config reference. | Read when WhatsApp MCP starts; restart MCP after changing. Invalid config fails closed. | src/lingtai/mcp_servers/whatsapp; protect credentials and recipient state. |
These are adapter/provider controls. The kernel does not reinterpret them as Nudge policy. Defaults below mean “unset” unless the provider adapter documents a provider-specific default.
| Variable | Purpose / default | Accepted values and scope | Read/reload and invalid behavior | Anchor / security |
|---|---|---|---|---|
LINGTAI_CODEX_TRANSPORT | Legacy/diagnostic transport hint; normal runtime remains the adapter's default REST path. | Adapter-recognized diagnostic value; default unset. | Read by the provider adapter at session construction; restart session. Unsupported values do not silently switch to an unsafe transport. | src/lingtai/llm; do not put bearer tokens in transport hints. |
LINGTAI_CODEX_WS | Legacy WebSocket experiment/diagnostic hint; default unset/off. | Adapter-recognized value; normal production behavior is unchanged when set by itself. | Read at provider-session construction; restart session. Unsupported values are ignored/rejected by the adapter. | src/lingtai/llm; transport selection is not a trust boundary. |
LINGTAI_CODEX_WS_EPOCH_RESET_TURNS | Diagnostic WebSocket epoch reset count; default adapter default. | Positive integer. | Read at session construction; restart session. Invalid falls back to adapter default. | Codex adapter; avoid values that cause unexpected session churn. |
LINGTAI_CODEX_RESPONSES_TRACE | Enables bounded Responses-wire tracing for diagnostics; default off. | Adapter-recognized boolean. | Read at session construction; restart session. Invalid is treated as off. | Codex adapter; traces must remain redacted and local. |
LINGTAI_CODEX_RESPONSES_TRACE_PATH | Local destination for Responses-wire diagnostics; default unset/adapter default. | Local path. | Read at session construction; restart session. Invalid/unwritable path fails closed or disables tracing. | Codex adapter; trace files can contain sensitive prompts, so restrict permissions. |
LINGTAI_CLAUDE_MANAGED_ROOT | Claude managed installation root; default host-specific/ unset. | Local directory. | Read by Claude launcher at launch; relaunch after changing. Invalid path fails closed. | Claude adapter; never widen root based on untrusted model text. |
LINGTAI_CLAUDE_INTERACTIVE_FIFO | Claude interactive FIFO path; default unset. | Local FIFO/path. | Read at interactive launch; relaunch after changing. Invalid type/permissions fail closed. | Claude adapter; protect FIFO from other users/processes. |
The following variables are intentionally not user-facing product configuration. They are injected by daemon supervisors, fake backends, or focused tests. They are catalogued so an audit can distinguish a test hook from a runtime control. Do not set them in a production agent environment.
| Variable | Purpose / default | Accepted values and scope | Read/reload and invalid behavior | Anchor / security |
|---|---|---|---|---|
LINGTAI_DAEMON_CAPSULE_FD | Supervisor-provided capsule file descriptor; default unset. | Integer FD supplied by the supervisor. | Read once at daemon start; restart daemon. Invalid/missing fails the supervised path. | src/lingtai/kernel/daemon_supervisor; never accept an arbitrary inherited FD without supervisor validation. |
LINGTAI_DAEMON_CAPSULE_HANDLE | Windows capsule pipe HANDLE number; default unset. | Integer OS handle inherited via the spawn handle_list. | Read/consumed once by the Windows daemon entrypoints, which convert it to the FD wire above; restart daemon. Carries only the handle number, never capsule content. | src/lingtai/adapters/windows/daemon_supervisor.py; never accept an arbitrary inherited handle without supervisor validation. |
LINGTAI_DAEMON_COMPLETION_FILE | Daemon completion marker path; default unset. | Supervisor-owned local path. | Read once at daemon start; restart daemon. Invalid path fails completion reporting. | Daemon completion MCP/runner; keep path within the assigned run directory. |
LINGTAI_DAEMON_CREDENTIALS_RESTORED | Internal marker that credential restoration finished; default unset. | Supervisor boolean marker. | Read during daemon startup; restart daemon. Invalid/missing means credentials are not considered restored. | Daemon supervisor; never log restored credential contents. |
LINGTAI_DAEMON_DETACHED_SUPERVISOR | Internal detached-supervisor mode; default unset/off. | Supervisor boolean marker. | Read once at daemon startup; restart daemon. Invalid is treated as off/fails the detached path. | Daemon supervisor; does not authorize external side effects. |
LINGTAI_DAEMON_RUN_ID | Correlates one daemon run; default unset. | Opaque run identifier. | Read at daemon startup and event/token writes; restart daemon to change. Invalid/missing is a run-integrity error where required. | Daemon run directory; do not use as a secret or expose unrelated run IDs. |
LINGTAI_DAEMON_SUPERVISOR_TEST_FAKE_LLM | Test fake-LLM switch; default off. | Test-only boolean. | Read at test/supervisor construction; restart child. Invalid is off. | Test/supervisor code; never enable in production. |
LINGTAI_DAEMON_SUPERVISOR_TEST_FAKE_LLM_FINISH | Test fake-LLM finish mode; default unset. | Test scenario string. | Read at child construction; restart child. Invalid scenario fails the test. | Test/supervisor code; no production effect. |
LINGTAI_DAEMON_SUPERVISOR_TEST_FAKE_LLM_SCENARIO | Selects a fake-LLM test scenario; default unset. | Registered test scenario. | Read at child construction; restart child. Unknown scenario fails closed. | Test/supervisor code; never use to bypass real provider checks. |
LINGTAI_DAEMON_SUPERVISOR_TEST_FAKE_LLM_SLEEP | Fake-LLM test delay; default zero/unset. | Non-negative numeric seconds. | Read at fake backend construction; restart child. Invalid falls back to zero or fails the test. | Test/supervisor code; not a production timeout control. |
LINGTAI_FAKE_CLI_REPORT | Test fake CLI report path/content selector; default unset. | Test-only path/selector. | Read by fake CLI at invocation; rerun test after changing. Invalid fails the test. | tests/_fake_*; do not set in production. |
LINGTAI_TEST_CONFIG | Test fixture configuration selector; default unset. | Test-only string/path. | Read when fixture is built; rerun test after changing. Invalid fails fixture setup. | tests/; no production behavior. |
LINGTAI_TEST_FAKE_CLAUDE_SIGNAL_RECORD | Test fake Claude signal record path; default unset. | Test-only local path. | Read by fake launcher at invocation; rerun test after changing. Invalid fails the test. | tests/; keep artifacts in the test temp directory. |
init.json; this catalogue
describes fallback/failed-read behavior, while the real init reader reports
parse/materialize/validate/resolve outcomes and leaves user input untouched.