databuddy

name	databuddy
description	Help external users integrate Databuddy into their own apps. Use for SDK setup, React/Vue/vanilla browser tracking, Node/server event tracking, custom event planning, attribution, feature flag evaluation, experiments, Databuddy DevTools, REST analytics queries, and event ingestion. Do not use for Databuddy monorepo implementation work; use databuddy-internal.

Databuddy

Use this skill for people adopting Databuddy in their own app or backend. Optimize for a working public integration or instrumentation plan: the right package, credential, endpoint, events, placement, and verification step.

Do not expose DATABUDDY_API_KEY in browser code. Browser integrations use a public website clientId; server/API integrations use an API key.

Choose The Surface

• React or Next.js browser analytics: @databuddy/sdk/react; mount <Databuddy /> near the app root.
• Vue or Nuxt browser analytics: @databuddy/sdk/vue; mount <Databuddy /> once near RouterView/NuxtPage and translate props to kebab-case in templates.
• Vanilla HTML, CMS, or GTM: CDN script https://cdn.databuddy.cc/databuddy.js with data-client-id.
• Browser custom events: track(...) from @databuddy/sdk after the tracker is installed.
• Browser helper utilities: use trackError, flush, clear, getTrackingIds, getTrackingParams, getAnonymousId, getSessionId, isTrackerAvailable, and getTracker from @databuddy/sdk when users need manual errors, navigation safety, logout reset, attribution handoff, or advanced checks.
• Server-side events: @databuddy/sdk/node; call flush() before a serverless/short-lived runtime exits.
• Feature flags: React/Vue flag helpers or createServerFlagsManager from @databuddy/sdk/node; use stable user context, handle loading/pending states, and call waitForInit() before first server read.
• DevTools: @databuddy/devtools; use for local/preview inspection of tracker status, IDs, queues, event calls, flags, overrides, diagnostics, and flag management.
• REST analytics: query API under https://api.databuddy.cc/v1.
• Raw event ingestion: POST https://basket.databuddy.cc/track; prefer SDKs unless the user explicitly wants HTTP.

Do not recommend @databuddy/sdk/ai/vercel; the public SDK currently exports only core, React, Vue, and Node entry points.

If the user says "API", determine whether they mean analytics queries, event ingestion, or feature flags before writing code.

Read public-surfaces.md when you need exact env vars, routes, scopes, snippets, or framework routing. Read frameworks.md for React, Next.js, Vue, Nuxt, and vanilla browser setup. Read server-events.md for browser helpers, Node/server events, raw /track, and attribution handoff. Read feature-flags.md for client flags, server flags, rollouts, experiments, flag metrics, and flag debugging. Read devtools.md for @databuddy/devtools setup, capabilities, flag overrides, diagnostics, and flag management. Read event-design.md when the user asks what custom events to add, which properties matter, or how to avoid noisy/high-cardinality tracking. Read instrumentation-planning.md when the user asks what to track, how to design metrics, how to wire attribution across client/server, or where to place tracking calls. Read troubleshooting.md for CSP, blockers, wrong hosts, auth, missing events, and flag/debug failures.

Credential And Endpoint Rules

Use	Host/path	Credential	Scope
Browser analytics	https://basket.databuddy.cc via SDK/CDN	clientId	none
Server events	POST https://basket.databuddy.cc/track	DATABUDDY_API_KEY	track:events
REST analytics	https://api.databuddy.cc/v1/query...	DATABUDDY_API_KEY	read:data
Feature flag evaluation	https://api.databuddy.cc/public/v1/flags...	clientId	none

• API auth accepts x-api-key: dbdy_... or Authorization: Bearer dbdy_...; prefer x-api-key for examples unless docs for that endpoint use Bearer.
• websiteId in event tracking means the public website client id, the same kind of value used by data-client-id; it is not necessarily an internal UUID.
• For server-side attribution, collect { anonId, sessionId } in the browser with getTrackingIds(), pass them to your backend, then map them to Node SDK fields anonymousId and sessionId.
• Do not invent a Databuddy identify() helper or endpoint. If an app has an identify route, treat it as the user's own backend code carrying Databuddy tracking IDs.
• Feature flag evaluation uses the main API host, not basket. It needs a website clientId, not API-key-only auth.
• Feature flag management uses manage:flags with x-api-key; do not expose that key in browser source. DevTools accepts it at runtime for local/preview flag management.
• DATABUDDY_API_URL is endpoint-specific. Do not reuse a basket root for query API or flags.
• Strict CSP sites need script-src for https://cdn.databuddy.cc and connect-src for https://basket.databuddy.cc; add https://api.databuddy.cc when using flags.

Workflow

1. Identify the user's runtime and target surface.
2. Prefer the highest-level supported SDK over raw HTTP.
3. Infer the minimum credential, or ask for only the missing one.
4. Give a short install/env/code path plus one verification step.
5. For custom instrumentation, define question -> metric -> event -> properties -> source -> placement -> verification before adding code.

Custom Event Rules

• Track user intent, product milestones, and operational outcomes, not every click.
• Use stable snake_case names and low-cardinality property keys.
• Do not track PII, secrets, raw tokens, full exception stacks, or large payloads.
• Prefer one event with useful properties over many near-duplicate event names.
• Use backend tracking for authoritative outcomes such as account creation, imports, webhooks, and jobs.
• Start with the fewest events that answer the user's question; add more only for funnel steps, important variants, or failure states.
• Do not track the same outcome in both browser and backend. Use browser events for intent and backend events for completed or failed outcomes.

Verification

• Browser SDK:
  • confirm the script loads and requests reach basket.databuddy.cc
  • confirm the expected clientId is configured
  • check CSP, ad-blockers, and registered/allowed domain settings if requests never leave the browser
• Node SDK:
  • send one test event
  • call flush() before exit
  • confirm no auth or queue errors
• REST API:
  • start with GET /v1/query/websites
  • then run the scoped query or event send
• Feature flags:
  • wait for initialization
  • test with a known flag key and user context
  • use DevTools to inspect readiness, evaluated values, variants, and overrides
• Custom instrumentation:
  • verify the event appears under the intended website
  • confirm attribution IDs are present when a server event belongs to a browser session
  • confirm the event can power the intended funnel, goal, segment, or reliability metric

Response Style

• Default to practical integration guidance and working examples
• Avoid internal monorepo details unless the user is debugging Databuddy itself.
• Keep answers short unless the user asks for a migration or full instrumentation plan.