@newrelic/preflight 0.0.1-pre.1 → 1.0.0

package/README.md

@@ -0,0 +1,498 @@
+# NR AI Coding Observability: Preflight
+
+**Observability for AI coding assistants.** Captures every action your AI coding tool takes — file reads, edits, commands, searches — and sends the data to New Relic so you can see exactly what's happening, how much it costs, and where it's wasting time.
+
+Think of it like Google Analytics for your AI pair programmer.
+
+## What It Does
+
+- **Tracks every action** — sees every file the AI reads, every command it runs, every edit it makes
+- **Tracks costs** — calculates USD spend per session, day, and week, broken down by model
+- **Detects waste** — catches inefficiencies like re-reading the same file repeatedly, making edits without reading first, or running the same failing command in a loop
+- **Measures efficiency** — computes a 0-100 score per task based on how directly the AI worked toward the goal
+- **Sends to New Relic** — all data lands in your NR account as queryable events and metrics, ready for dashboards and alerts
+
+---
+
+## Before You Start
+
+Two things are always required. A New Relic account is optional — the tool works in [local mode](#local-mode) without one.
+
+### 1. An AI coding tool
+
+This works with **Claude Code**, Cursor, Windsurf, GitHub Copilot, Zed, Continue.dev, or Amazon Q Developer. The examples below use Claude Code, which has the deepest integration.
+
+### 2. Node.js v22 or higher (v24 recommended)
+
+Open a terminal and run:
+
+```bash
+node --version
+```
+
+If it shows `v22.x.x` or higher, you're set. v24 is recommended (and what the project uses for development). If you need to upgrade, install it from [nodejs.org](https://nodejs.org) or via nvm:
+
+```bash
+nvm install 24 && nvm use 24
+```
+
+### 3. A New Relic account (optional)
+
+> Skip this section if you plan to use [local mode](#local-mode) — no NR account needed.
+
+To send telemetry to New Relic, you need two keys:
+
+| Key              | What it does                            | Where to find it                                                                                                              |
+| ---------------- | --------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- |
+| **License key**  | Sends telemetry data _into_ NR (ingest) | NR One → top-right menu → API keys → create a key of type **License**. Long hex string ending in `NRAL`. Not `NRAK-`.         |
+| **User API key** | Deploys dashboards and alerts _into_ NR | NR One → top-right menu → API keys → create a key of type **User**. Starts with `NRAK-`. Only needed for the deploy commands. |
+
+You'll also need your **Account ID** — a number visible in the URL when you're logged into NR One: `https://one.newrelic.com/nr1-core?account=`**`12345`**.
+
+**Data ingest:** this tool is free and open source, but the telemetry it sends counts against your NR data ingest. Standard ingest rates apply on paid plans. Monitor your usage in **NR One → Data Management → Data Ingestion**.
+
+---
+
+## Quick Start
+
+**Step 1 — Install**
+
+```bash
+npm install -g @newrelic/preflight
+```
+
+**Step 2 — Run the interactive setup wizard**
+
+```bash
+preflight setup
+```
+
+The wizard first asks for a mode: choose **cloud** to send telemetry to New Relic, or **local** for offline dashboard-only use. In cloud mode it then asks for your license key, account ID, region (US, EU, FedRAMP), and optionally a user API key. It validates keys live before continuing and pre-fills your developer name from the email on the API key. Most people are running in under 5 minutes.
+
+If `NEW_RELIC_LICENSE_KEY`, `NEW_RELIC_ACCOUNT_ID`, or `NEW_RELIC_API_KEY` are already set in your shell environment, the wizard detects them and lets you press Enter to accept — no copy-paste needed.
+
+Prefer non-interactive? Skip the wizard and run:
+
+```bash
+preflight install \
+  --license-key YOUR_LICENSE_KEY \
+  --account-id YOUR_ACCOUNT_ID
+```
+
+This installs the hook scripts in your Claude Code settings and registers the MCP server in `~/.mcp.json`, so every tool call is captured and the `nr_observe_*` tools are available in Claude Code. You only run this once.
+
+**Step 3 — Deploy dashboards** _(optional but recommended)_
+
+Replace `NRAK-...` with your user API key and `12345` with your account ID:
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-dashboards --all
+```
+
+This creates 7 dashboards in your NR account. Find them under **Dashboards** → search "AI Coding". The deploy commands ship with the package — both `preflight deploy-dashboards` and `preflight deploy-alerts` are available immediately after `npm install -g`.
+
+> **Region:** Use your account's license key and account ID for your environment. Add `--eu` for EU region accounts or `--gov` for FedRAMP accounts. Omit both flags for standard production (`one.newrelic.com`).
+
+**Step 4 — Restart Claude Code and verify**
+
+Restart Claude Code, then verify the MCP server connected:
+
+> _Can you call the `nr_observe_health` tool and show me the result?_
+
+You should see `"status": "ok"` with a version and uptime. Then make any request to Claude Code (triggering at least one tool call), and check session tracking:
+
+> _Can you call the `nr_observe_get_session_stats` tool and show me the result?_
+
+If you get back tool call counts and timing data, it's working.
+
+---
+
+## From source / contributing
+
+This is also the current install path until the npm package is published. Use it to develop the project, deploy dashboards/alerts, or stay on the latest unreleased changes:
+
+```bash
+git clone https://github.com/newrelic-experimental/preflight
+cd preflight
+nvm use          # Switch to the right Node version (24+)
+npm install      # Install all dependencies
+npm run build    # Compile TypeScript
+npm link         # Register preflight binary on PATH (required for hooks)
+```
+
+Then run `preflight setup` exactly as in the Quick Start.
+
+> **`npm link` permission error?** If you see `EACCES: permission denied` pointing at `/usr/local/lib/node_modules`, your system Node.js is installed in a root-owned directory. Pick one fix:
+>
+> _Quick fix — set a user-writable npm prefix (keeps your existing Node.js):_
+>
+> ```bash
+> npm config set prefix ~/.npm-global
+> export PATH="$HOME/.npm-global/bin:$PATH"   # also add to ~/.zshrc or ~/.bash_profile
+> npm link
+> ```
+>
+> _Recommended — use nvm (better if you switch Node versions):_
+>
+> ```bash
+> curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash
+> # restart your shell, then:
+> nvm install 24 && nvm use 24
+> npm install && npm run build && npm link
+> ```
+>
+> Do not use `sudo npm link` — it creates root-owned files that break future `npm install` runs.
+
+---
+
+## Talking to Preflight
+
+Once installed, Claude Code can query live session data on your behalf. Just ask it in plain English — or use the tool names directly:
+
+| What to ask                                                           | What you get back                                                                           |
+| --------------------------------------------------------------------- | ------------------------------------------------------------------------------------------- |
+| _"Show me my session stats"_ → `nr_observe_get_session_stats`         | Tool call counts, success rate, total duration                                              |
+| _"What's my efficiency score?"_ → `nr_observe_get_efficiency_score`   | A 0-100 score with a breakdown of where points were lost                                    |
+| _"How much has this session cost?"_ → `nr_observe_get_cost_breakdown` | USD cost broken down by tool type and AI model                                              |
+| _"Any budget warnings?"_ → `nr_observe_get_budget_status`             | Current spend vs. your configured caps (if set)                                             |
+| _"Any wasteful patterns?"_ → `nr_observe_get_anti_patterns`           | Detected inefficiencies — repeated reads, blind edits, stuck loops                          |
+| _"Any recommendations?"_ → `nr_observe_get_recommendations`           | Personalized suggestions for this session                                                   |
+| _"How am I doing this week?"_ → `nr_observe_get_personal_insights`    | A narrative coaching report vs. your own historical baseline (requires 2+ weeks of history) |
+
+Everything also flows into your New Relic dashboards automatically — you don't have to ask Claude to see it there.
+
+---
+
+## Dashboards
+
+After deploying, you'll have seven dashboards in NR One:
+
+| Dashboard               | What it shows                                                   |
+| ----------------------- | --------------------------------------------------------------- |
+| **Overview**            | Session stats, efficiency score, cost summary, top tools        |
+| **Session Detail**      | Every tool call in a specific session, in order                 |
+| **Personal**            | 30-day self-reflection view scoped to one developer             |
+| **Team View**           | Aggregated cost and efficiency across multiple developers       |
+| **Manager View**        | Team-level cost by developer with no tool-call content visible  |
+| **Platform Comparison** | Side-by-side metrics across Claude Code, Cursor, Windsurf, etc. |
+| **Security Audit**      | Audit trail of sensitive file access and destructive commands   |
+
+### Personal dashboard
+
+Deploy a dashboard pre-filtered to your name (it opens already showing your data):
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-dashboards \
+  ai-coding-assistant-personal.json --developer your-name
+```
+
+### Updating or removing dashboards
+
+To replace existing dashboards in place after pulling new fixes (preserves the dashboard's GUID and URL), add `--update`:
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-dashboards --all --update
+```
+
+To delete the deployed dashboards, add `--teardown`. Dashboards are matched by name; missing ones are skipped:
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-dashboards --all --teardown
+```
+
+### Terraform (IaC alternative)
+
+A Terraform module in `terraform/` deploys all 7 dashboards and the full alert policy as an alternative to the deploy scripts — useful for GitOps workflows or when you want state tracking. See [ADVANCED.md — Terraform Deployment](./docs/ADVANCED.md#terraform-deployment) for usage.
+
+---
+
+## Alert Conditions
+
+Optional: get notified in NR when something goes wrong.
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-alerts
+```
+
+Add `--eu` for EU region accounts or `--gov` for FedRAMP accounts. This creates five alert conditions: daily cost spike, low efficiency score, stuck loop rate, anti-pattern rate, and session cost budget. To remove them, add `--teardown`.
+
+To apply changes to alert JSONs without losing the existing policy, add `--update`. This syncs conditions in place (matched by name): updates existing ones, creates new ones, and deletes any that have been removed locally:
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-alerts --update
+```
+
+For personal alerts scoped to your developer name:
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-alerts --developer your-name
+```
+
+---
+
+## Configuration
+
+The easiest way to configure is through the setup wizard (`preflight setup`). To edit manually, open `~/.newrelic-preflight/config.json`:
+
+```json
+{
+  "licenseKey": "aabbccdd...",
+  "accountId": "12345",
+  "developer": "your-name",
+  "sessionBudgetUsd": 1.0,
+  "dailyBudgetUsd": 5.0,
+  "weeklyBudgetUsd": 20.0
+}
+```
+
+### Key settings
+
+| Setting              | What it does                                                                                                                                                                                                                                                            | Default      |
+| -------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------ |
+| `developer`          | Your identifier on all NR events. Automatically normalized to lowercase with underscores — e.g., "John Doe" → "john_doe". Falls back to `$USER` or your git name if not set.                                                                                            | Inferred     |
+| `collectorHost`      | Region override: `null` = US (default), `'eu'` = EU, `'gov'` = FedRAMP/GovCloud. The wizard auto-detects from your license key prefix and lets you confirm.                                                                                                             | `null` (US)  |
+| `nrApiKey`           | User API key (`NRAK-...`) for NerdGraph queries (team summaries, dashboard/alert deploy). The wizard prompts for it and validates it live.                                                                                                                              | Not set      |
+| `sessionBudgetUsd`   | Emits a warning event at 50%, 80%, 100% of this amount per session                                                                                                                                                                                                      | No limit     |
+| `dailyBudgetUsd`     | Emits a warning event at 50%, 80%, 100% of this amount per day                                                                                                                                                                                                          | No limit     |
+| `weeklyBudgetUsd`    | Emits a warning event at 50%, 80%, 100% of this amount per week                                                                                                                                                                                                         | No limit     |
+| `retainSessionsDays` | Auto-deletes local session files older than N days                                                                                                                                                                                                                      | Keep forever |
+| `teamId`             | A label **you define** (e.g. `"platform-eng"`, `"my-team"`) stamped on all NR events as `team_id`, enabling cross-developer queries like `WHERE team_id = 'platform-eng'`. This is **not** your NR account ID — it's a free-form slug you choose to identify your team. | Not set      |
+| `projectId`          | Tags all events with a project name (auto-derived from your git remote URL if not set)                                                                                                                                                                                  | Auto-derived |
+| `digestWebhookUrl`   | Slack webhook URL for weekly cost and efficiency summaries                                                                                                                                                                                                              | Not set      |
+
+All settings can also be set via environment variables — see [example.config.js](./example.config.js) for the full annotated reference.
+
+### Validating your config
+
+If the MCP server fails to connect, run:
+
+```bash
+preflight validate
+```
+
+This checks your config file for JSON syntax errors, invalid field types, and misspelled or unknown keys — and suggests corrections:
+
+```
+✗ Error: mode: Invalid enum value. Expected 'cloud' | 'local' | 'both', received 'clod'
+⚠ Warning: Unknown key "licensekey" — did you mean "licenseKey"?
+```
+
+Pass `--config <path>` to check a file at a non-default location.
+
+### OTLP Transport
+
+To export telemetry to other OpenTelemetry-compatible backends (Datadog, Grafana Cloud, Honeycomb, or New Relic's OTLP endpoint), or to enable an inbound OTLP receiver in proxy mode, see [ADVANCED.md](./docs/ADVANCED.md#otlp-transport).
+
+---
+
+## Updating
+
+To pull the latest changes and rebuild in one step:
+
+```bash
+preflight update
+```
+
+This runs `git pull` followed by `npm run build` in the repo directory. Restart Claude Code afterwards to pick up the new version.
+
+---
+
+## Uninstalling
+
+To remove the Preflight hooks and MCP server from Claude Code:
+
+```bash
+preflight uninstall
+```
+
+This removes the hooks from your user-level Claude Code settings and deregisters the MCP server. A timestamped backup of your settings is saved automatically before any changes are made.
+
+If you installed at the project level, add `--project`:
+
+```bash
+preflight uninstall --project
+```
+
+Restart Claude Code after uninstalling for the changes to take effect.
+
+### Removing dashboards and alerts
+
+If you deployed dashboards or alerts, tear them down separately:
+
+```bash
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-dashboards --all --teardown
+
+NEW_RELIC_API_KEY=NRAK-... NEW_RELIC_ACCOUNT_ID=12345 \
+  preflight deploy-alerts --teardown
+```
+
+### Removing local data
+
+Session history and configuration are stored in `~/.newrelic-preflight/`. To remove everything:
+
+```bash
+rm -rf ~/.newrelic-preflight
+```
+
+### Unlinking the binary
+
+If you registered the CLI globally via `npm link`, remove it with:
+
+```bash
+npm unlink -g @newrelic/preflight
+```
+
+---
+
+## Local mode
+
+If you'd rather not ship telemetry to New Relic, set `mode: 'local'` in your config:
+
+```json
+{
+  "mode": "local"
+}
+```
+
+In local mode:
+
+- The MCP server does **not** construct `NrIngestManager` and never makes outbound HTTP calls to NR.
+- An embedded dashboard boots at **http://127.0.0.1:7777** (configurable via `dashboard.port` or `NR_AI_DASHBOARD_PORT`).
+- All telemetry stays in `~/.newrelic-preflight/` on your machine.
+- `licenseKey` and `accountId` are not required.
+
+**With Claude Code** (default): the server runs via the MCP connection (`--stdio`). You don't launch it manually — Claude Code starts it automatically when you open a session, because `preflight install` registered it as an MCP server. The dashboard stays alive as long as your Claude Code session is open.
+
+**Standalone** (no Claude Code required): pass `--local` to run the dashboard server directly, without an MCP transport. Use this to browse the dashboard when Claude Code isn't running, or to observe non-Claude-Code sources that hit the hooks (e.g. scripts using the Anthropic SDK). If the per-session MCP is also installed, only one process owns the dashboard at a time — whichever started first — and the other runs headless.
+
+```bash
+npm run build          # build once
+node dist/index.js --local
+# or:
+npm run dev            # shortcut: assumes dist/ already built
+npm run dev:all        # build + start in one step
+npm run start:local    # alias for npm run dev
+```
+
+Open `http://127.0.0.1:7777` in your browser. The process stays alive until you press Ctrl+C.
+
+The dashboard has six views:
+
+- **Today** — live KPIs, sparkline of tool latencies, recent calls, anti-pattern alerts.
+- **Sessions** — list of past sessions with a per-session timeline of every tool call.
+- **History** — weekly efficiency and daily spend trends.
+- **Audit** — every classified tool call (sensitive file access, destructive commands, external network), with a JSONL export button.
+- **Settings** — edit developer name, team ID, budget caps, and session retention from the browser (no config file editing required).
+- **Alerts** — live budget spend vs. caps, editable personal alert thresholds, and Slack digest configuration.
+
+Run `preflight setup` to choose a mode interactively.
+
+---
+
+## Local Alerts
+
+Local-mode users get threshold alerting evaluated in-process — no New Relic dependency. Rules live at `~/.newrelic-preflight/alerts/rules.json`; a starter set is copied into place by the setup wizard.
+
+For the full list of rule types, channel options, alert log configuration, and live reload behavior, see [ADVANCED.md](./docs/ADVANCED.md#local-alerts).
+
+---
+
+## Weekly Digest
+
+Register a Slack webhook to receive a weekly summary every Monday morning:
+
+In Claude Code, ask: _"Call `nr_observe_subscribe_digest` with this webhook URL: `https://hooks.slack.com/services/...`"_
+
+Or set it in your config file as `digestWebhookUrl`, or configure it directly from the **Alerts** tab in the local dashboard.
+
+---
+
+## Supported Platforms
+
+| Platform           | How to enable                                          |
+| ------------------ | ------------------------------------------------------ |
+| Claude Code        | `preflight install` (automatic)                        |
+| Cursor             | Set `NEW_RELIC_AI_PLATFORM=cursor` in your environment |
+| Windsurf           | Set `NEW_RELIC_AI_PLATFORM=windsurf`                   |
+| GitHub Copilot     | Set `NEW_RELIC_AI_PLATFORM=copilot`                    |
+| Zed                | Set `NEW_RELIC_AI_PLATFORM=zed`                        |
+| Continue.dev       | Set `NEW_RELIC_AI_PLATFORM=continue`                   |
+| Amazon Q Developer | Set `NEW_RELIC_AI_PLATFORM=amazonq`                    |
+
+---
+
+## Glossary
+
+**MCP (Model Context Protocol)** — A standard that lets AI assistants like Claude Code discover and call external tools. Preflight registers itself as an MCP server so Claude Code can call it directly.
+
+**License key** — A NR credential for _sending_ data into New Relic. Looks like a long hex string ending in `NRAL`. Found under API Keys in NR One.
+
+**User API key** — A NR credential for _reading_ data and managing resources (dashboards, alerts). Starts with `NRAK-`. Create one under API Keys in NR One.
+
+**Anti-pattern** — A detected waste pattern. Examples: re-reading the same file multiple times without making changes between reads (the AI lost context and is reloading it), making edits to a file without reading it first (blind edit), running the same failing command in a loop (stuck loop).
+
+**Efficiency score** — A 0-100 number per task. High means the AI worked directly toward the goal. Low means wasted tool calls — repeated reads, blind edits, unnecessary backtracking.
+
+**Token** — The unit AI models use to measure text length for billing. Roughly 3-4 characters per token. One page of text ≈ 500 tokens.
+
+**Hook** — A script that Claude Code calls automatically before and after every tool call. Preflight uses this to capture tool call data without interrupting your workflow.
+
+---
+
+## Requirements
+
+- **Node.js**: v22 or higher (`.nvmrc` pins v24 for development)
+- **New Relic account**: free tier works; you need a license key (a user API key is optional — only needed for deploying dashboards and alerts)
+- **An AI coding tool**: Claude Code, Cursor, Windsurf, GitHub Copilot, Zed, Continue.dev, or Amazon Q
+
+---
+
+## Documentation
+
+- **[CONTRIBUTING.md](./CONTRIBUTING.md)** — Development setup, architecture, conventions, testing, and end-to-end verification
+- **[ADVANCED.md](./docs/ADVANCED.md)** — OTLP export, local alerts, per-developer alerts, session backfill, Terraform deployment
+- **[COMMANDS_TABLE.md](./docs/COMMANDS_TABLE.md)** — All MCP tools with parameters and return values
+- **[METRICS_TABLE.md](./docs/METRICS_TABLE.md)** — Every event and metric sent to New Relic
+- **[SECURITY.md](./docs/SECURITY.md)** — Security practices and audit trail
+
+---
+
+## For Contributors
+
+### Development setup
+
+```bash
+nvm install && nvm use
+npm install
+npm run build
+npm test
+```
+
+### Common tasks
+
+| Command                | Purpose                                                    |
+| ---------------------- | ---------------------------------------------------------- |
+| `npm run build`        | Build TypeScript server + Vite web dashboard               |
+| `npm run build:server` | Build only the TypeScript server (`tsc --build`)           |
+| `npm run build:web`    | Build only the Vite web dashboard (output: `dist/web/`)    |
+| `npm test`             | Run all tests                                              |
+| `npm run lint`         | Check code style                                           |
+| `npm run format`       | Auto-format code                                           |
+| `npm run dev`          | Start local dashboard (assumes pre-built `dist/`)          |
+| `npm run dev:all`      | Build then start local dashboard                           |
+| `npm run dev:full`     | Build backend, then run backend + Vite dev server together |
+| `npm run start:local`  | Alias for `npm run dev`                                    |
+
+See [CONTRIBUTING.md](./CONTRIBUTING.md) for the full development guide, conventions, and architecture.
+
+---
+
+**Questions?** Start with [CONTRIBUTING.md](./CONTRIBUTING.md) or [open an issue](https://github.com/newrelic-experimental/preflight/issues).

package/dist/alerts/alert-log.d.ts

@@ -0,0 +1,24 @@
+import type { AlertEvent } from '../dashboard/live-event-bus.js';
+export interface AlertLogOptions {
+    readonly path: string;
+    readonly maxBytes?: number;
+}
+/**
+ * Append-only JSONL log for alert events. One file with one rotation
+ * (`<path>.1`). All filesystem I/O is deferred to first append() so the
+ * constructor stays cheap and side-effect-free for testing.
+ */
+export declare class AlertLog {
+    private readonly path;
+    private readonly maxBytes;
+    private initialized;
+    private pending;
+    constructor(opts: AlertLogOptions);
+    append(event: AlertEvent): Promise<void>;
+    readRecent(limit: number): Promise<AlertEvent[]>;
+    private doReadRecent;
+    rotateIfNeeded(): Promise<void>;
+    close(): Promise<void>;
+    private ensureInitialized;
+}
+//# sourceMappingURL=alert-log.d.ts.map

package/dist/alerts/alert-log.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"alert-log.d.ts","sourceRoot":"","sources":["../../src/alerts/alert-log.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,gCAAgC,CAAC;AAoBjE,MAAM,WAAW,eAAe;IAC9B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,CAAC;CAC5B;AAED;;;;GAIG;AACH,qBAAa,QAAQ;IACnB,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAS;IAC9B,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAS;IAClC,OAAO,CAAC,WAAW,CAAS;IAC5B,OAAO,CAAC,OAAO,CAAoC;gBAEvC,IAAI,EAAE,eAAe;IAK3B,MAAM,CAAC,KAAK,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC;IAuBxC,UAAU,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAAC,UAAU,EAAE,CAAC;YAexC,YAAY;IA4CpB,cAAc,IAAI,OAAO,CAAC,IAAI,CAAC;IAuB/B,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;YAKd,iBAAiB;CAahC"}