tescmd 0.1.2__tar.gz → 0.3.1__tar.gz

tescmd-0.3.1/.env.example

@@ -0,0 +1,75 @@
+# tescmd — Environment Variables
+#
+# Copy this file to .env and uncomment the variables you need.
+# The setup wizard (tescmd setup) will create ~/.config/tescmd/.env
+# for you automatically.
+#
+# Resolution order:
+#   1. Shell environment variables (highest priority)
+#   2. .env in the current working directory
+#   3. ~/.config/tescmd/.env (created by tescmd setup)
+
+# ─── Tesla Fleet API Credentials ─────────────────────────────────
+# Required. Obtain from https://developer.tesla.com
+# TESLA_CLIENT_ID=
+# TESLA_CLIENT_SECRET=
+
+# ─── Vehicle ──────────────────────────────────────────────────────
+# Default vehicle VIN (avoids passing --vin on every command)
+# TESLA_VIN=
+
+# API region: na (North America), eu (Europe), cn (China)
+# TESLA_REGION=na
+
+# ─── Token Override / Storage ─────────────────────────────────────
+# Direct token override (bypasses keyring / token file)
+# TESLA_ACCESS_TOKEN=
+# TESLA_REFRESH_TOKEN=
+
+# File path for token storage (skips OS keyring)
+# Useful for Docker, headless Linux, CI environments
+# TESLA_TOKEN_FILE=~/.config/tescmd/tokens.json
+
+# ─── Config & Cache ──────────────────────────────────────────────
+# TESLA_CONFIG_DIR=~/.config/tescmd
+# TESLA_CACHE_DIR=~/.cache/tescmd
+# TESLA_CACHE_TTL=60
+# TESLA_CACHE_ENABLED=true
+
+# ─── Output Format & Display Units ───────────────────────────────
+# Force output format: rich (TTY default), json (piped default), quiet
+# TESLA_OUTPUT_FORMAT=
+
+# Display unit preferences
+# TESLA_TEMP_UNIT=F           # F or C
+# TESLA_DISTANCE_UNIT=mi      # mi or km
+# TESLA_PRESSURE_UNIT=psi     # psi or bar
+
+# ─── Command Protocol & Key Hosting ──────────────────────────────
+# Command signing: auto (default), signed, unsigned
+# TESLA_COMMAND_PROTOCOL=auto
+
+# Domain for key hosting (set by tescmd setup)
+# TESLA_DOMAIN=
+
+# Key hosting method: github or tailscale
+# TESLA_HOSTING_METHOD=
+
+# GitHub repo for key deployment (e.g. username/username.github.io)
+# TESLA_GITHUB_REPO=
+
+# Setup tier: readonly or full
+# TESLA_SETUP_TIER=
+
+# Config profile name
+# TESLA_PROFILE=default
+
+# ─── MCP Server ──────────────────────────────────────────────────
+# Credentials for MCP OAuth 2.1 authentication
+# Required for tescmd serve and tescmd mcp serve
+# TESCMD_MCP_CLIENT_ID=
+# TESCMD_MCP_CLIENT_SECRET=
+
+# ─── OpenClaw ────────────────────────────────────────────────────
+# Gateway authentication token
+# OPENCLAW_GATEWAY_TOKEN=

tescmd-0.3.1/.github/workflows/publish.yml

@@ -0,0 +1,19 @@
+name: Publish to PyPI
+
+on:
+  release:
+    types: [published]
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write  # trusted publishing
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: pip install build
+      - run: python -m build
+      - uses: pypa/gh-action-pypi-publish@release/v1

tescmd-0.3.1/.github/workflows/test.yml

@@ -0,0 +1,23 @@
+name: Test
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.11", "3.12", "3.13"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - run: pip install -e ".[dev,telemetry]"
+      - run: ruff check src/ tests/
+      - run: ruff format --check src/ tests/
+      - run: mypy src/
+      - run: pytest

{tescmd-0.1.2 → tescmd-0.3.1}/.gitignore

@@ -5,8 +5,9 @@ dist/
 build/
 .venv/
 .env
+.env.bak
 *.pem
 .mypy_cache/
 .pytest_cache/
 .ruff_cache/
-.coverage
+.coverage

tescmd-0.3.1/CHANGELOG.md

@@ -0,0 +1,158 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.3.1] - 2026-02-02
+
+### Added
+
+- **README overhaul** — header banner, logo, new "What It Does" summary section, expanded Prerequisites table with Python 3.11+, pip, Tesla account, and helpful links
+- **Tailscale Funnel auto-detection in auth setup** — `_interactive_setup` now detects Tailscale and offers to start Funnel so Tesla can verify the origin URL during Developer Portal configuration; cleans up Funnel on exit
+- **Tailscale hostname passthrough** — setup wizard forwards detected Tailscale hostname to auth flow, showing a concrete "Also add" origin URL instead of the generic placeholder
+- **Comprehensive agent skill documentation** — expanded skill covering all command groups, triggers, and OpenClaw dispatch
+
+### Fixed
+
+- **Cross-platform file permissions** — OpenClaw gateway key file now uses `secure_file()` from `_internal.permissions` instead of raw `chmod(0o600)`, adding proper Windows support via `icacls`
+- **12-factor app compliance** — config, disposability, and concurrency improvements across the codebase
+- **Documentation accuracy** — corrected `media_adjust_volume` tool name in MCP docs; emphasized MCP/OpenClaw over direct CLI for cost savings in agent skill
+
+## [0.3.0] - 2026-02-01
+
+### Added
+
+- **Unified `tescmd serve`** — single command combining MCP server, Fleet Telemetry streaming, cache warming, and optional OpenClaw bridge; full-screen TUI dashboard shows live telemetry, MCP status, tunnel URL, sink count, and connection health
+- **TUI dashboard** — 8-panel Textual layout with telemetry field table, server info sidebar, activity log, request log, and filter status; command palette (`ctrl+p`), keybindings (`q` to quit, `f` to toggle filters), clean graceful shutdown
+- **OpenClaw Bridge** — `tescmd openclaw bridge [VIN]` streams Fleet Telemetry to an OpenClaw Gateway with configurable delta+throttle filtering per field; supports `--dry-run` for JSONL output without a gateway connection
+- **OpenClaw node role** — bidirectional command dispatch over the gateway WebSocket; bots send commands via the gateway that are forwarded as `node.invoke.request` events, routed through `CommandDispatcher` with read/write separation, tier enforcement, and VCSEC signing guards
+- **Trigger subscription system** — `trigger.create`, `trigger.delete`, `trigger.list`, `trigger.poll` commands let bots register conditions on any telemetry field; supports operators `lt`, `gt`, `lte`, `gte`, `eq`, `neq`, `changed`, `enter`, `leave`; one-shot and persistent modes with configurable cooldown; max 100 triggers, 500 pending notifications
+- **Geofence triggers** — `enter`/`leave` operators on Location field detect boundary crossings using haversine distance; fires only on actual crossing (not "already inside"), requires previous position for comparison
+- **Trigger convenience aliases** — `cabin_temp.trigger`, `outside_temp.trigger`, `battery.trigger`, `location.trigger` pre-fill the field name so bots don't need to know raw telemetry field names
+- **Trigger notification delivery** — dual-channel: OpenClaw push events (`trigger.fired`) for connected bots, and MCP polling via `trigger_poll` tool for agent frameworks
+- **`system.run` meta-dispatch** — allows bots to invoke any registered handler by name with alias mapping (e.g., `door_lock` → `door.lock`, `auto_conditioning_start` → `climate.on`); guards against recursive self-dispatch
+- **MCP Server** — `tescmd mcp serve` exposes all tescmd commands as MCP tools for Claude Desktop/Code and other agent frameworks; supports stdio and streamable-http transports with OAuth 2.1 authentication
+- **MCP custom tool registry** — `MCPServer.register_custom_tool()` allows runtime registration of non-CLI tools (used by trigger system); custom tools appear alongside CLI-backed tools with proper schemas
+- **Agent skill** — `skills/tescmd/SKILL.md` teaches AI agents how to use every tescmd command group with examples, parameter types, and common patterns
+- **Reusable telemetry session** — extracted shared telemetry lifecycle (server → tunnel → partner registration → fleet config → cleanup) into `telemetry/setup.py` for use by serve, stream, and bridge commands
+- **Dual-gate telemetry filter** — `openclaw/filters.py` combines delta threshold (value change) and throttle interval (minimum time between emissions) to reduce noise in telemetry streams; includes haversine distance for location fields
+- **OpenClaw event emitter** — maps telemetry fields to OpenClaw `req:agent` event payloads (location, battery, temperature, speed, charge state transitions, security changes)
+- **Gateway WebSocket client** — implements the OpenClaw operator protocol (challenge → connect → hello-ok handshake) with Ed25519 device key signing and exponential backoff reconnection
+- **Bridge configuration** — `BridgeConfig` pydantic model with per-field filter settings, loadable from JSON file or CLI flags
+- **CSV telemetry logging** — wide-format CSV log with one row per frame and one column per subscribed field; written to `~/.config/tescmd/logs/` by default, disable with `--no-log`
+- **Cache sink** — telemetry-driven cache warming keeps read-command cache fresh while telemetry is active, making agent reads free
+- **Frame fanout** — `FrameFanout` distributes decoded telemetry frames to multiple sinks (dashboard, CSV, cache, OpenClaw bridge, triggers) in parallel
+- **Telemetry field mapper** — `telemetry/mapper.py` maps protobuf field names to tescmd model field names with unit conversion
+- **Command guards in dispatcher** — extracted `check_command_guards()` (tier check + VCSEC signing requirement) into a shared function called by both CLI and OpenClaw dispatcher paths
+- **Standard dependencies** — `websockets>=14.0`, `mcp>=1.0`, and `textual>=1.0` now included in default install
+
+### Changed
+
+- Refactored `_cmd_telemetry_stream` in `cli/vehicle.py` to use the shared `telemetry_session()` context manager (no behavioral change)
+- Proto-aligned telemetry field definitions with proper `interval_seconds` for delta fields
+- Separated JSON serialization errors from WebSocket connection errors in gateway send path for clearer diagnostics
+- Raised trigger push notification and lifecycle event failures from debug to warning level for observability
+
+### Fixed
+
+- Fixed prompt for re-authentication when refresh token is expired or revoked (was failing silently)
+- Fixed delta fields requiring `interval_seconds` configuration
+- Fixed log file paths not shown on quit
+- Added `exc_info` to MCP custom tool error logging for full stack traces
+- Added JSON parse guards and explicit parameter validation in MCP tool wrappers
+- Added `system.run` recursive self-dispatch guard
+
+## [0.2.0] - 2026-01-31
+
+### Added
+
+- **Fleet Telemetry Streaming** — `tescmd vehicle telemetry stream [VIN]` starts a local WebSocket server, exposes it via Tailscale Funnel, configures the vehicle to push real-time telemetry, and displays it in an interactive Rich Live dashboard (TTY) or JSONL stream (piped)
+- **Telemetry dashboard** — Rich Live TUI with field name, value, and last-update columns; unit conversion (°F/°C, mi/km, psi/bar); connection status; frame counter; uptime display
+- **Protobuf telemetry decoder** — official Tesla protobuf definitions (`vehicle_data`, `vehicle_alert`, `vehicle_error`, `vehicle_metric`, `vehicle_connectivity`) for fully typed telemetry message parsing
+- **FlatBuffer telemetry support** — `flatbuf.py` parser for Tesla's FlatBuffer-encoded telemetry payloads alongside protobuf
+- **Field presets** — `--fields` option accepts preset names (`default`, `driving`, `charging`, `climate`, `all`) or comma-separated field names with 120+ registered telemetry fields
+- **Interval override** — `--interval` option overrides the polling interval for all fields
+- **Tailscale Funnel integration** — automatic Funnel start/stop with cert retrieval for Fleet Telemetry HTTPS requirement
+- **JSONL output** — piped mode emits one JSON line per telemetry frame for scripting and log ingestion
+- **TunnelError hierarchy** — `TunnelError` parent with `TailscaleError` subtype; actionable install/setup guidance
+- **websockets dependency** — `websockets>=14.0` now included in default install
+- **Tailscale key hosting** — `tescmd key deploy --method tailscale` hosts the public key via Tailscale Funnel at `https://<machine>.tailnet.ts.net/.well-known/appspecific/com.tesla.3p.public-key.pem`; auto-detected as second priority after GitHub Pages
+- **Key hosting priority chain** — setup wizard and `key deploy` auto-detect the best hosting method: GitHub Pages → Tailscale Funnel → manual; `--method` flag overrides auto-detection
+- **`TESLA_HOSTING_METHOD` setting** — persists the chosen key hosting method (`github`, `tailscale`) across sessions
+- **Schnorr signature support** — `crypto/schnorr.py` for Schnorr-based authentication challenges used in telemetry server handshake
+- **`auth import` command** — `tescmd auth import < tokens.json` imports tokens from a JSON file for headless/CI environments
+- **Setup guide** — `docs/setup.md` with step-by-step walkthrough of all 7 setup phases
+- **FAQ** — `docs/faq.md` covering common questions about tescmd, costs, hosting, and configuration
+- **CI/CD workflows** — GitHub Actions for test-on-push (Python 3.11–3.13) and publish-to-PyPI-on-release via trusted publishing
+- **README badges** — PyPI version, Python versions, CI build status, license, and GitHub release badges
+- **E2E smoke tests** — `tests/cli/test_e2e_smoke.py` provides 179 pytest-based end-to-end tests covering every CLI command against the live Fleet API, with JSON envelope validation and save/restore for write commands (`pytest -m e2e`)
+
+### Fixed
+
+- Fixed telemetry dashboard uptime counter not incrementing
+- Improved tunnel start/stop success messages for clarity
+
+## [0.1.2] - 2025-01-31
+
+### Added
+
+- **Universal read-command caching** — every read command is now transparently cached with tiered TTLs (STATIC 1h, SLOW 5m, DEFAULT 1m, FAST 30s); bots can call tescmd as often as needed — within the TTL window, responses are instant and free
+- **Generic cache key scheme** — `generic_cache_key(scope, identifier, endpoint, params)` generates scope-aware keys (`vin`, `site`, `account`, `partner`) for any API endpoint
+- **`cached_api_call()` helper** — unified async helper that handles cache lookup, fetch, serialisation (Pydantic/dict/list/scalar), and storage for all non-vehicle-state read commands
+- **Site-scoped cache invalidation** — `invalidate_cache_for_site()` clears energy site entries after write commands; `invalidate_cache_for_vin()` now also clears generic vin-scoped keys
+- **`cache clear` options** — `--site SITE_ID` and `--scope {account,partner}` flags for targeted cache clearing alongside existing `--vin`
+- **Partner endpoints** — `partner public-key`, `partner telemetry-error-vins`, `partner telemetry-errors` for partner account data (require client credentials)
+- **Billing endpoints** — `billing history`, `billing sessions`, `billing invoice` for Supercharger charging data
+- **Cross-platform file permissions** — `_internal/permissions.py` provides `secure_file()` using `chmod 0600` on Unix and `icacls` on Windows
+- **Token store file backend** — `_FileBackend` with atomic writes and restricted permissions as fallback when keyring is unavailable
+- **Spec-driven Fleet API validation** — `scripts/validate_fleet_api.py` validates implementation against `spec/fleet_api_spec.json` using AST introspection
+- **6 missing Fleet API commands** — added `managed_charging_set_amps`, `managed_charging_set_location`, `managed_charging_set_schedule`, `add_charge_schedule`, `remove_charge_schedule`, `clear_charge_schedules`
+- **Configurable display units** — `--units metric` flag switches all display values to °C/km/bar; individual env vars (`TESLA_TEMP_UNIT`, `TESLA_DISTANCE_UNIT`, `TESLA_PRESSURE_UNIT`) for granular control
+
+### Fixed
+
+- Aligned schedule/departure command parameters with Tesla Go SDK (correct param names and types)
+- Fixed energy endpoint paths to match Fleet API spec
+- Fixed Rich markup escaping bug in command output
+- Aligned command parameters (3 param gaps) with Go SDK specs
+
+### Changed
+
+- Response cache documentation in CLAUDE.md expanded to cover universal caching, TTL tiers, and generic cache key scheme
+
+## [0.1.1]
+
+### Added
+
+- **`status` command** — `tescmd status` shows current configuration, auth, cache, and key status at a glance
+- **Retry option in wake prompt** — when a vehicle is asleep, the interactive prompt now offers `[R] Retry` alongside `[W] Wake via API` and `[C] Cancel`, allowing users to wake the vehicle for free via the Tesla app and retry without restarting the command
+- **Key enrollment** — `tescmd key enroll <VIN>` sends the public key to the vehicle and guides the user through Tesla app approval with interactive [C]heck/[R]esend/[Q]uit prompt, `--wait` auto-polling, and JSON mode support
+- **Tier enforcement** — readonly tier now blocks write commands with a clear error and upgrade guidance (`tescmd setup`)
+- **Vehicle Command Protocol** — ECDH session management, HMAC-SHA256 command signing, and protobuf RoutableMessage encoding for the `signed_command` endpoint; commands are automatically signed when keys are available (`command_protocol=auto`)
+- **SignedCommandAPI** — composition wrapper that transparently routes signed commands through the Vehicle Command Protocol while falling back to unsigned REST for `wake_up` and unknown commands
+- **`command_protocol` setting** — `auto` (default), `signed`, or `unsigned` to control command routing; configurable via `TESLA_COMMAND_PROTOCOL` env var
+- **Enrollment step in setup wizard** — full-tier setup now offers to enroll the key on a vehicle after key generation
+- **Friendly command output** — all vehicle commands now display descriptive success messages (e.g. "Climate control turned on.", "Doors locked.") instead of bare "OK"
+- **E2E smoke tests** — `tests/cli/test_e2e_smoke.py` provides 179 pytest-based end-to-end tests covering every CLI command against the live Fleet API, with JSON envelope validation and save/restore for write commands (`pytest -m e2e`)
+
+## [0.1.0]
+
+### Added
+
+- OAuth2 PKCE authentication with browser-based login flow
+- Vehicle state queries: battery, charge, climate, drive, location, doors, windows, trunks, tire pressure
+- Vehicle commands: charge start/stop/limit/schedule, climate on/off/set/seats/wheel, lock/unlock, sentry, trunk/frunk, windows, media, navigation, software updates, HomeLink, speed limits, PIN management
+- Energy products: Powerwall live status, site info, backup reserve, operation mode, storm mode, TOU settings, charging history, calendar history, grid config
+- User account: profile info, region, orders, feature config
+- Vehicle sharing: add/remove drivers, create/redeem/revoke invites
+- Rich terminal output with tables, panels, and status indicators
+- JSON output mode for scripting and agent integration
+- Configurable display units (F/C, mi/km, PSI/bar)
+- Response caching with configurable TTL for API cost reduction
+- Cost-aware wake confirmation (interactive prompt or `--wake` flag)
+- Multi-profile configuration support
+- EC key generation and Tesla Developer Portal registration
+- Raw API access (`raw get`, `raw post`) for uncovered endpoints
+- First-run setup wizard with Fleet Telemetry cost guidance

tescmd-0.3.1/CLAUDE.md

@@ -0,0 +1,131 @@
+# CLAUDE.md — Project Context for Claude Code
+
+## Project Overview
+
+**tescmd** is a Python 3.11+ CLI for querying and controlling Tesla vehicles via the [Tesla Fleet API](https://developer.tesla.com/docs/fleet-api). It covers auth, vehicle queries/commands, energy products, Supercharger billing, Fleet Telemetry streaming, OpenClaw bridge, and an MCP server for agent integration.
+
+## Tech Stack
+
+- **Python 3.11+**, **pydantic v2**, **click**, **httpx** (async), **rich**, **cryptography**, **protobuf**, **keyring**, **python-dotenv**
+- **websockets** (telemetry + OpenClaw), **mcp** (MCP server) — core dependencies
+- Optional: **bleak** (`[ble]` extra — BLE key enrollment)
+
+## Project Structure
+
+```
+src/tescmd/
+├── cli/                   # Click CLI layer
+│   ├── main.py            # Root group, AppContext, _register_commands()
+│   ├── _options.py        # Shared Click options/decorators (@global_options)
+│   ├── _client.py         # API client builders, auto_wake, cached_vehicle_data, cached_api_call, TTL tiers
+│   ├── auth.py, cache.py, charge.py, billing.py, climate.py, security.py
+│   ├── status.py, trunk.py, vehicle.py, media.py, nav.py, partner.py
+│   ├── software.py, energy.py, user.py, sharing.py, raw.py, key.py
+│   ├── setup.py           # Interactive first-run wizard
+│   ├── serve.py           # Unified MCP + telemetry + OpenClaw command
+│   ├── openclaw.py        # Standalone openclaw bridge command
+│   └── mcp_cmd.py         # mcp serve command
+├── api/                   # HTTP client + domain APIs (composition pattern)
+│   ├── client.py          # TeslaFleetClient (base HTTP, auth headers, retries)
+│   ├── vehicle.py, command.py, signed_command.py, energy.py
+│   ├── charging.py, partner.py, sharing.py, user.py
+│   └── errors.py          # AuthError, VehicleAsleepError, TierError, TunnelError, etc.
+├── models/                # Pydantic v2 models (vehicle, energy, user, auth, command, config)
+├── auth/                  # OAuth2 PKCE, token_store (keyring + file fallback), callback server
+├── protocol/              # Vehicle Command Protocol (ECDH sessions, HMAC signing, protobuf)
+├── crypto/                # EC key gen, ECDH, Schnorr signatures
+├── cache/                 # File-based JSON cache with tiered TTLs
+├── output/                # OutputFormatter, RichOutput (DisplayUnits), JsonOutput
+├── telemetry/             # Fleet Telemetry streaming
+│   ├── setup.py           # Reusable telemetry_session() context manager
+│   ├── server.py, decoder.py, fields.py, dashboard.py, tailscale.py
+├── openclaw/              # OpenClaw bridge
+│   ├── config.py          # BridgeConfig, FieldFilter, NodeCapabilities (pydantic)
+│   ├── filters.py         # DualGateFilter (delta + throttle), haversine()
+│   ├── emitter.py         # EventEmitter (telemetry → OpenClaw events)
+│   ├── gateway.py         # GatewayClient (WebSocket, node protocol, Ed25519 auth)
+│   ├── bridge.py          # TelemetryBridge orchestrator, build_openclaw_pipeline()
+│   ├── dispatcher.py      # CommandDispatcher (reads, writes, triggers, system.run)
+│   └── telemetry_store.py # In-memory latest-value cache for telemetry fields
+├── triggers/              # Trigger subscription system
+│   ├── models.py          # TriggerOperator, TriggerCondition, TriggerDefinition, TriggerNotification
+│   └── manager.py         # TriggerManager (evaluation, cooldown, delivery, geofencing)
+├── mcp/                   # MCP server
+│   └── server.py          # MCPServer (FastMCP, CliRunner + custom callable tools)
+├── deploy/                # Key hosting (GitHub Pages, Tailscale Funnel)
+└── _internal/             # vin.py, async_utils.py, permissions.py
+```
+
+## Coding Conventions
+
+- **Type hints everywhere** — all function signatures, all variables where non-obvious
+- **async/await** — all API calls are async; CLI entry points use `run_async()` helper
+- **Pydantic models** — all API request/response payloads; all configuration
+- **src layout** — code in `src/tescmd/`, tests in `tests/`
+- **No star imports** — explicit imports only
+- **Single responsibility** — CLI modules handle args + output, API modules handle HTTP
+- **Composition over inheritance** — API classes wrap `TeslaFleetClient`, don't extend it
+- **Error stream routing** — JSON/piped mode writes errors to stderr; Rich/TTY mode uses stdout
+
+## Key Patterns
+
+**CLI command registration:** New command groups are registered in `cli/main.py:_register_commands()`. Each CLI module defines a Click group and is imported there.
+
+**Global options propagation:** Use `@global_options` decorator from `cli/_options.py` on commands that need VIN, format, wake, cache flags. Options flow through `AppContext` via `@click.pass_obj`.
+
+**API client construction:** `cli/_client.py` provides `get_vehicle_api(app_ctx)` → returns `(client, api)` tuple. Also `get_command_api()` for write commands (handles signed vs unsigned routing).
+
+**Cache:** All read commands use `cached_api_call()` with scope-aware TTLs (STATIC 1h, SLOW 5m, DEFAULT 1m, FAST 30s). Write commands invalidate via `invalidate_cache_for_vin()` / `invalidate_cache_for_site()`.
+
+**Output:** `OutputFormatter` auto-detects TTY → Rich, piped → JSON, `--quiet` → stderr only. JSON output uses a consistent `{ok, command, data, error, timestamp}` envelope.
+
+**Telemetry session lifecycle:** `telemetry/setup.py:telemetry_session()` is an async context manager handling: server start → Tailscale tunnel → partner domain re-registration → fleet config → yield → cleanup. Used by both `cli/vehicle.py` (stream) and `cli/openclaw.py` (bridge).
+
+**OpenClaw pipeline:** `TelemetryServer.on_frame` → `DualGateFilter.should_emit()` → `EventEmitter.to_event()` → `GatewayClient.send_event()`. The `TelemetryBridge` class wires these together. `build_openclaw_pipeline()` is the shared factory used by both `cli/openclaw.py` (standalone) and `cli/serve.py` (combined mode). Inbound commands flow: gateway `node.invoke.request` → `CommandDispatcher.dispatch()` → handler → `node.invoke.result`.
+
+**Trigger system:** `TriggerManager` evaluates conditions on every telemetry frame (independent of the dual-gate filter). Supports numeric comparison (`lt`, `gt`, `eq`, etc.), `changed` detection, and geofence `enter`/`leave` with haversine distance. One-shot and persistent (with cooldown) firing modes. Notifications delivered via OpenClaw push and MCP polling (`trigger.poll`).
+
+**MCP tools:** `mcp/server.py` maps tool names to CLI arg lists via `_CliToolDef` dataclasses. `invoke_tool()` uses `CliRunner.invoke(cli, ["--format", "json", "--wake", *args], env=os.environ.copy())`. Custom tools (e.g. trigger CRUD) use `_CustomToolDef` with direct callable handlers registered via `register_custom_tool()`. Read/write tools are separated for `readOnlyHint` annotations. HTTP transport uses OAuth 2.1 via `_InMemoryOAuthProvider` (auto-approve authorization, dynamic client registration, in-memory token storage) with `_PermissiveClient` wrappers that accept any redirect URI. DNS rebinding protection is configured to allow the Tailscale Funnel hostname when `public_url` is set.
+
+## Build & Test
+
+```bash
+# Build
+python -m build                    # hatchling via pyproject.toml
+
+# Test (parallel by default via pytest-xdist)
+pytest                             # all ~1600 tests
+pytest tests/openclaw/ -x -v       # openclaw tests
+pytest tests/triggers/ -x -v       # trigger tests
+pytest tests/mcp/ -x -v            # mcp tests
+pytest -m e2e                      # live API smoke tests (needs TESLA_ACCESS_TOKEN)
+
+# Lint
+ruff check src/ tests/
+ruff format src/ tests/
+mypy src/
+```
+
+## Environment Variables
+
+| Variable | Description |
+|---|---|
+| `TESLA_CLIENT_ID` / `TESLA_CLIENT_SECRET` | OAuth2 app credentials |
+| `TESLA_VIN` | Default vehicle VIN |
+| `TESLA_REGION` | API region: `na`, `eu`, `cn` |
+| `TESLA_ACCESS_TOKEN` / `TESLA_REFRESH_TOKEN` | Direct token override |
+| `TESLA_TOKEN_FILE` | File path for token storage (skips keyring) |
+| `TESLA_CONFIG_DIR` | Config directory (default: `~/.config/tescmd`) |
+| `TESLA_CACHE_DIR` / `TESLA_CACHE_TTL` / `TESLA_CACHE_ENABLED` | Cache settings |
+| `TESLA_OUTPUT_FORMAT` | Force format: `rich`, `json`, `quiet` |
+| `TESLA_COMMAND_PROTOCOL` | `auto` (default), `signed`, `unsigned` |
+| `TESLA_TEMP_UNIT` / `TESLA_DISTANCE_UNIT` / `TESLA_PRESSURE_UNIT` | Display units |
+| `TESLA_DOMAIN` / `TESLA_HOSTING_METHOD` / `TESLA_GITHUB_REPO` | Key hosting |
+| `TESLA_SETUP_TIER` | `readonly` or `full` |
+| `TESLA_PROFILE` | Active config profile |
+| `OPENCLAW_GATEWAY_TOKEN` | OpenClaw gateway auth token |
+
+
+## Additional Resources
+
+Tesla's published vehicle-command proto omits some VehicleAction fields (e.g. navigation commands). The Teslemetry project maintains a more complete proto: https://github.com/Teslemetry/python-tesla-fleet-api/blob/main/proto/car_server.proto