@amplitude/ai 0.3.6 → 0.3.7

package/AGENTS.md

@@ -1,6 +1,6 @@
 # AGENTS.md
 
-Package: `@amplitude/ai` v0.3.6
+Package: `@amplitude/ai` v0.3.7
 
 ## Install
 
@@ -53,12 +53,12 @@ Codex auto-reads this `AGENTS.md` file for context.
 
 ## Decision Tree
 
-- Need zero-code coverage: use `patch()`.
-- Already have a provider client: use `wrap()` or provider wrappers.
-- Need user/session lineage: use `ai.agent(...).session(...)`.
+- **Default: use `ai.agent(...).session(...)` with provider wrappers** — gives you every event type, per-user analytics, session enrichment, quality scoring.
+- Already have a provider client: use `wrap()` to instrument it.
 - Multiple agents collaborating: use `session.runAs(childAgent, fn)` for automatic identity propagation.
 - Need tool telemetry: use `tool()`.
 - Need span/observability: use `observe()`.
+- Cannot modify call sites at all: use `patch()` for aggregate-only monitoring (no per-user analytics).
 - Need agent-assistant guidance: run MCP prompt `instrument_app`.
 
 ## MCP Surface
@@ -102,6 +102,8 @@ Prompt:
 
 ## CLI
 
+- `amplitude-ai` — Print instrumentation prompt for AI coding agents
+- `amplitude-ai --print-guide` — Print the full amplitude-ai.md guide to stdout
 - `amplitude-ai mcp` — Start the MCP server for AI coding agents
 - `amplitude-ai doctor [--json]` — Validate environment, deps, and event pipeline
 - `amplitude-ai status [--json]` — Show SDK version, installed providers, and env config

package/README.md

@@ -51,14 +51,19 @@ The CLI prints a prompt to paste into any AI coding agent (Cursor, Claude Code,
 
 The agent reads the guide, scans your project, discovers your agents and LLM call sites, and instruments everything — provider wrappers, session lifecycle, multi-agent delegation, tool tracking, scoring, and a verification test. You review and approve each step.
 
-### Other paths
+### Manual setup
 
-| Your situation | Recommended path | What happens |
-|---|---|---|
-| **Manual setup** | Follow the [Quick Start guide](#quick-start) | Agents + sessions + provider wrappers — the full event model |
-| **Just want to verify the SDK works** | `patch()` ([details below](#patching)) | Aggregate cost/latency monitoring only — no user analytics, no funnels |
+Whether you use a coding agent or set up manually, the goal is the same: **full instrumentation** — agents + sessions + provider wrappers. This gives you every event type, per-user analytics, and server-side enrichment.
+
+Follow the [code example above](#amplitude-ai) to get started. The pattern is:
+
+1. **Swap your LLM import** — `import { OpenAI } from '@amplitude/ai'` (or `Anthropic`, `Gemini`, etc.)
+2. **Create an agent** — `ai.agent('my-agent')` to name and track your AI component
+3. **Wrap in a session** — `agent.session({ userId, sessionId }).run(async (s) => { ... })` for per-user analytics, funnels, cohorts, and server-side enrichment
+4. **Track user messages** — `s.trackUserMessage(...)` for conversation context
+5. **Score responses** — `s.score(...)` for quality measurement
 
-> **Start with full instrumentation.** The coding agent workflow defaults to agents + sessions + provider wrappers. This gives you every event type, per-user analytics, and server-side enrichment. `patch()` exists for quick verification or legacy codebases where you can't modify call sites, but it only captures `[Agent] AI Response` without user identity — no funnels, no cohorts, no retention.
+> `patch()` exists for quick verification or legacy codebases where you can't modify call sites, but it only captures `[Agent] AI Response` without user identity — no funnels, no cohorts, no retention. Start with full instrumentation; fall back to `patch()` only if you can't modify call sites.
 
 | Property        | Value                                                                                                                                                                            |
 | --------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
@@ -100,7 +105,7 @@ The agent reads the guide, scans your project, discovers your agents and LLM cal
 - [Auto-Instrumentation CLI](#auto-instrumentation-cli)
 - [Integrations](#integrations)
 - [Data Flow](#data-flow)
-- [Which Integration Should I Use?](#which-integration-should-i-use)
+- [Integration Approaches](#integration-approaches)
 - [Integration Patterns](#integration-patterns)
 - [Serverless Environments](#serverless-environments)
 - [Error Handling and Reliability](#error-handling-and-reliability)
@@ -137,7 +142,7 @@ Install provider SDKs based on what you use (for example: `openai`, `@anthropic-
 
 1. **Install:** `npm install @amplitude/ai @amplitude/analytics-node`
 2. **Get your API key:** In Amplitude, go to **Settings > Projects** and copy the API key.
-3. **Auto-instrument:** Run `npx amplitude-ai` and paste the printed prompt into your AI coding agent — it scans your project, generates a bootstrap file, instruments your LLM call sites, and creates a verification test. Or follow the manual patterns below.
+3. **Instrument:** Run `npx amplitude-ai` and paste the printed prompt into your AI coding agent. Or follow the [manual setup](#manual-setup) steps — the goal is the same: agents + sessions + provider wrappers.
 4. **Set your API key** in the generated `.env` file and replace the placeholder `userId`/`sessionId`.
 5. **Run your app.** You should see `[Agent] User Message`, `[Agent] AI Response`, and `[Agent] Session End` within 30 seconds.
 
@@ -504,7 +509,13 @@ const session = agent.session({
 });
 ```
 
-Without this, sessions with long idle periods may be closed and evaluated prematurely. The default is 30 minutes.
+Without this, sessions with long idle periods may be closed and enrichment may run earlier than expected. The default is 30 minutes.
+
+**Session lifecycle and enrichment.** You do **not** need to call `trackSessionEnd()` for sessions to work. Amplitude's server automatically closes sessions after 30 minutes of inactivity and queues them for enrichment (topic classification, quality scoring, session evaluation) at that point. The only reason to call `trackSessionEnd()` is to **trigger enrichment sooner** — for example, if you know the conversation is over and want evaluation results immediately rather than waiting for the idle timeout.
+
+"Closed" is a server-side concept meaning "queued for enrichment" — it does **not** prevent new events from flowing into the same session. If the user resumes a conversation after session end, new messages with the same `sessionId` are still associated with that session.
+
+If you use `session.run()`, session end is tracked automatically when the callback completes. For long-lived conversations (chatbots, support agents), you can skip explicit session end entirely and let the server handle it.
 
 **Link to Session Replay**: If your frontend uses Amplitude's [Session Replay](https://www.docs.developers.amplitude.com/session-replay/), pass the browser's `deviceId` and `browserSessionId` to link AI sessions to browser recordings:
 
@@ -1643,33 +1654,20 @@ Your Application
 - The LLM Enrichment Pipeline runs asynchronously after session close (only when `contentMode: 'full'`). It produces server-side events like `[Agent] Session Evaluation` and `[Agent] Score`.
 - With `contentMode: 'customer_enriched'`, the enrichment pipeline is skipped — you provide your own enrichments via `trackSessionEnrichment()`.
 
-## Which Integration Should I Use?
-
-Start here and pick the first tier that satisfies your analytics needs:
-
-```
-                      Do you need per-user analytics
-                      (funnels, cohorts, retention)?
-                              │
-                    ┌─── No ──┴── Yes ──┐
-                    │                   │
-              Tier 0: Zero-Code     Do you need session
-              (CLI auto-patch)      grouping & enrichment?
-              Cost, latency,              │
-              tokens, errors.     ┌─ No ──┴── Yes ──┐
-                                  │                 │
-                           Tier 1: Events      Do you control
-                           Per-call tracking   the LLM call site?
-                           + userId                  │
-                                           ┌── Yes ──┴── No ──┐
-                                           │                  │
-                                    Tier 2: Sessions    Tier 3: OTEL Bridge
-                                    session.run()       Add exporter to
-                                    Full enrichment     existing OTEL pipeline
-                                    Implicit feedback   Limited to OTEL attrs
-```
-
-**Rule of thumb:** If you own the LLM call site, start with **Tier 2** (sessions). If you don't (e.g., a third-party framework exports OTEL spans), use **Tier 3** (OTEL bridge). If you just want aggregate cost monitoring without user analytics, **Tier 0** (zero-code) is ready in 60 seconds.
+## Integration Approaches
+
+**Start with full instrumentation.** Use agents + sessions + provider wrappers. This is the recommended approach for both coding agent and manual workflows — it gives you every event type, per-user analytics, and server-side enrichment.
+
+| Approach | When to use | What you get |
+|---|---|---|
+| **Full control** (recommended) | Any project, new or existing | `BoundAgent` + `session.run()` + provider wrappers — all event types, per-user funnels, cohorts, retention, quality scoring, enrichments |
+| **Express/Fastify middleware** | Web app, auto-session per request | Same as full control with automatic session lifecycle via `createAmplitudeAIMiddleware` |
+| **Swap import** | Existing codebase, incremental adoption | `new OpenAI({ amplitude: ai })` — auto-tracking per call, add sessions when ready |
+| **Wrap** | You've already created a client | `wrap(client, ai)` — instruments an existing client instance |
+| **Zero-code / `patch()`** | Verification or legacy codebases only | `patch({ amplitudeAI: ai })` — `[Agent] AI Response` only, no user identity, no funnels |
+| **OTEL Bridge** | Third-party framework exports OTEL spans | Add exporter to existing OTEL pipeline — limited to OTEL attributes |
+
+> The first four approaches all support the full event model. Choose based on how you want to integrate — the analytics capabilities are the same. **`patch()` is the exception**: it only captures aggregate `[Agent] AI Response` events without user identity, useful only for verifying the SDK works or for codebases where you can't modify call sites.
 
 ## Integration Patterns
 

package/amplitude-ai.md

@@ -29,30 +29,21 @@ Detected environment:
   Multi-agent signals: [yes/no]
   Streaming: [yes/no]
   Frontend deps: [yes/no]
-  Recommended tier: [quick_start / standard / advanced]
+  Recommended: full instrumentation
 ```
 
-**Decision point:** Ask the developer to confirm the detection and choose a tier:
-- **Quick start** — `patch({ amplitudeAI: ai })`, zero code changes, good for getting data flowing
-- **Standard** — Provider wrappers + session middleware (recommended for most apps)
-- **Advanced** — Multi-agent `runAs`, agent descriptions, scoring, tool tracking
-
-If multi-agent signals are detected, recommend Advanced.
+**Next step:** Confirm the detection with the developer, then proceed to full instrumentation. Always instrument with agents, sessions, provider wrappers, tool tracking, and scoring. If multi-agent signals are detected, also add child agents and `runAs` delegation.
 
 ---
 
 ## Phase 2: Discover Agents and Call Sites
 
-For **Quick start** tier, skip to Phase 3 — discovery is just "which providers are imported."
-
-For **Standard** and **Advanced** tiers:
-
 1. Identify files with LLM call sites (search for `chat.completions.create`, `messages.create`, `generateContent`, `streamText`, `generateText`)
 2. For each file with call sites, read the actual source and review:
    - Is it a route handler / API endpoint?
    - What provider(s) does it use?
    - Does it call other files with LLM call sites? (delegation → multi-agent)
-3. For **Advanced** tier, also identify:
+3. Identify:
    - Agent boundaries (each distinct orchestration unit = one agent)
    - Delegation patterns (parent calls child → `runAs`)
    - Feedback handlers (thumbs up/down UI components)
@@ -135,17 +126,10 @@ export const openai = new OpenAI({
 // export const anthropic = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY!, amplitude: ai });
 ```
 
-**Alternative: `wrap()` for existing clients.** If the project creates provider clients dynamically:
-
-```typescript
-import { wrap } from '@amplitude/ai';
-import OpenAI from 'openai';
-const rawClient = new OpenAI({ apiKey: process.env.OPENAI_API_KEY! });
-export const openai = wrap(rawClient, ai);
-```
-
 Add `AMPLITUDE_AI_API_KEY` to `.env.example`. Check `.gitignore` includes `.env`.
 
+> **Note:** If you cannot modify provider instantiation sites, use `wrap(existingClient, ai)` to instrument an existing client, or `patch({ amplitudeAI: ai })` for zero-code verification. These capture fewer event types — always prefer provider wrappers when possible.
+
 ### Step 3c: Swap provider imports
 
 Replace direct provider instantiation with imports from the bootstrap file:
@@ -163,7 +147,7 @@ import { openai as client } from '@/lib/amplitude';
 
 ### Step 3d: Add session context
 
-**Standard tier** — wrap route handlers with agent + session:
+Wrap route handlers with agent + session:
 
 ```typescript
 import { ai } from '@/lib/amplitude';
@@ -183,7 +167,7 @@ export async function POST(req: Request) {
 }
 ```
 
-**Advanced tier** — add multi-agent delegation with `runAs`:
+If multi-agent signals were detected, add delegation with `runAs`:
 
 ```typescript
 const orchestrator = ai.agent('shopping-agent', { description: 'Orchestrates shopping requests' });
@@ -220,7 +204,7 @@ s.trackAiMessage(completedMessage.content, 'gpt-4o', 'openai', latencyMs, {
 });
 ```
 
-### Step 3f: Add span tracking for orchestration (Advanced tier)
+### Step 3f: Add span tracking for orchestration
 
 When a parent agent delegates work to a child agent, wrap the delegation call with span tracking to measure latency and capture errors. Look for existing try/catch blocks around sub-agent execution — these are natural places to add span tracking with both success and error paths:
 
@@ -253,9 +237,9 @@ try {
 }
 ```
 
-### Step 3g: Add scoring (Advanced tier only)
+### Step 3g: Add scoring
 
-If feedback handlers were detected (thumbs up/down UI), check whether the handler receives a `messageId` — if so, target the specific message for finer-grained scoring. Otherwise fall back to session-level scoring:
+If feedback handlers were detected (thumbs up/down UI), add scoring. Check whether the handler receives a `messageId` — if so, target the specific message for finer-grained scoring. Otherwise fall back to session-level scoring:
 
 ```typescript
 const targetId = messageId ?? sessionId;

package/bin/amplitude-ai.mjs

@@ -25,9 +25,10 @@ Paste this into your AI coding agent (Cursor, Claude Code, Copilot, etc.):
   Instrument this app with @amplitude/ai. Follow node_modules/@amplitude/ai/amplitude-ai.md
 
 CLI commands:
-  mcp       Start the MCP server (optional, for advanced tooling)
-  doctor    Validate environment, deps, and event pipeline
-  status    Show SDK version, installed providers, and env config
+  mcp            Start the MCP server (optional, for advanced tooling)
+  doctor         Validate environment, deps, and event pipeline
+  status         Show SDK version, installed providers, and env config
+  --print-guide  Print the full amplitude-ai.md instrumentation guide to stdout
 `
   );
   process.exit(0);
@@ -39,6 +40,17 @@ if (command === '--version' || command === '-v') {
   process.exit(0);
 }
 
+if (command === '--print-guide') {
+  const guidePath = join(dirname(fileURLToPath(import.meta.url)), '..', 'amplitude-ai.md');
+  try {
+    process.stdout.write(readFileSync(guidePath, 'utf8'));
+  } catch {
+    process.stderr.write(`Guide not found at ${guidePath}\n`);
+    process.exit(1);
+  }
+  process.exit(0);
+}
+
 if (command in commandToBin) {
   const binPath = fileURLToPath(new URL(commandToBin[command], import.meta.url));
   const result = spawnSync(process.execPath, [binPath, ...rest], {
@@ -48,5 +60,5 @@ if (command in commandToBin) {
   process.exit(result.status ?? 1);
 }
 
-process.stderr.write(`Unknown command: ${command}\nUsage: amplitude-ai <mcp|doctor|status>\n`);
+process.stderr.write(`Unknown command: ${command}\nUsage: amplitude-ai <mcp|doctor|status|--print-guide>\n`);
 process.exit(1);

package/dist/bound-agent.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"bound-agent.d.ts","names":[],"sources":["../src/bound-agent.ts"],"sourcesContent":[],"mappings":";;;;;;AAY8B,KAAlB,eAAA,GAAkB,OAAA,CAC5B,IAD4B,CACvB,UADuB,CACZ,WADY,CAAA,kBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,SAAA,CAAA,CAAA;AAAO,KAIzB,aAAA,GAAgB,OAJS,CAKnC,IALmC,CAMjC,UANiC,CAMtB,WANsB,CAAA,gBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,SAAA,GAAA,OAAA,GAAA,UAAA,GAAA,WAAA,CAAA,CAAA;AAIzB,KAOA,YAAA,GAAe,OAPF,CAQvB,IARuB,CASrB,UATqB,CASV,WATU,CAAA,eAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,UAAA,GAAA,WAAA,GAAA,SAAA,CAAA,CAAA;AAEV,KAYH,aAAA,GAAgB,OAZb,CAab,IAba,CAcX,UAdW,CAcA,WAdA,CAAA,gBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,OAAA,GAAA,UAAA,GAAA,WAAA,CAAA,CAAA;AAAX,KAmBQ,QAAA,GAAW,OAnBnB,CAoBF,IApBE,CAoBG,UApBH,CAoBc,WApBd,CAAA,WAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,UAAA,GAAA,WAAA,CAAA,CAAA;AADF,KAwBU,cAAA,GAAiB,OAxB3B,CAyBA,UAzBA,CAyBW,WAzBX,CAAA,iBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,CAAA;AAD0B,KA6BhB,qBAAA,GAAwB,OA7BR,CA8B1B,IA9B0B,CA8BrB,UA9BqB,CA8BV,WA9BU,CAAA,wBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,aAAA,CAAA,CAAA;AAAO,KAiCvB,SAAA,GAAY,OAjCW,CAkCjC,IAlCiC,CAkC5B,UAlC4B,CAkCjB,WAlCiB,CAAA,OAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,MAAA,GAAA,OAAA,GAAA,UAAA,CAAA,CAAA;AAOvB,UA4CK,YAAA,CA5CO;EAET,MAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAAX,aAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EADF,aAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EADyB,YAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAAO,WAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAOtB,OAAA,CAAA,EA2CA,MA3Ca,CAAA,MAAA,EAAA,OAAA,CAAA,GAAA,IAAA;EAEV,GAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAAX,SAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EADF,OAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAD0B,MAAA,CAAA,EA+CjB,MA/CiB,CAAA,MAAA,EAAA,OAAA,CAAA,GAAA,IAAA;EAAO,QAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAOvB,gBAAQ,CAAA,EAAA,MAAA,GAAA,IAAA;;AACb,cA4CM,UAAA,CA5CN;EAAL,SAAA,GAAA,EA6Cc,WA7Cd;EADqB,SAAA,SAAA,EA+CD,MA/CC,CAAA,MAAA,EAAA,OAAA,CAAA;EAAO,WAAA,CAAA,EAAA,EAkDtB,WAlDsB,EAAA,IAAA,EAmDpB,YAnDoB,GAAA;IAIlB,OAAA,EAAA,MAAc;EACb,CAAA;EAAX,IAAA,OAAA,CAAA,CAAA,EAAA,MAAA;EAD2B,IAAA,EAAA,CAAA,CAAA,EAuEjB,WAvEiB;EAAO,KAAA,CAAA,OAAA,EAAA,MAAA,EAAA,SAAA,CAAA,EA2EA,YA3EA,CAAA,EA2EoB,UA3EpB;EAIxB,QAAA,MAAA;EACM,gBAAA,CAAA,OAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAwIwB,eAxIxB,CAAA,EAAA,MAAA;EAAX,cAAA,CAAA,OAAA,EAAA,MAAA,EAAA,KAAA,EAAA,MAAA,EAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAoJG,aApJH,CAAA,EAAA,MAAA;EAAL,aAAA,CAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,OAAA,EAAA,OAAA,EAAA,IAAA,CAAA,EAmKQ,YAnKR,CAAA,EAAA,MAAA;EADkC,cAAA,CAAA,KAAA,EAAA,MAAA,EAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAkL1B,aAlL0B,CAAA,EAAA,MAAA;EAAO,SAAA,CAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EA4LY,QA5LZ,CAAA,EAAA,MAAA;EAI/B,eAAS,CAAA,IAAA,CAAA,EAgMG,cAhMH,CAAA,EAAA,IAAA;EACH,sBAAA,CAAA,WAAA,EAsMD,kBAtMC,EAAA,IAAA,CAAA,EAuMR,qBAvMQ,CAAA,EAAA,IAAA;EAAX,KAAA,CAAA,IAAA,EAAA,MAAA,EAAA,KAAA,EAAA,MAAA,EAAA,QAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAmNG,SAnNH,CAAA,EAAA,IAAA;EAAL,OAAA,CAAA,IAqCM,CArCN,EAAA;IADsB,SAAA,CAAA,EAAA,MAAA,GAAA,IAAA;IAAO,kBAAA,CAAA,EAAA,MAAA,GAAA,IAAA;IAkBd,MAAA,CAAA,EAAA,MAAY,GAAA,IAAA;IAehB,QAAA,CAAU,EAAA,MAAA,GAAA,IAAA;IACP,gBAAA,CAAA,EAAA,MAAA,GAAA,IAAA;IACM,SAAA,CAAA,EAAA,OAAA;EAGd,CAAA,CAAA,EAiMH,OAjMG;EACE,KAAA,CAAA,CAAA,EAAA,OAAA;EAwBE,QAAA,CAAA,CAAA,EAAA,IAAA"}
+{"version":3,"file":"bound-agent.d.ts","names":[],"sources":["../src/bound-agent.ts"],"sourcesContent":[],"mappings":";;;;;;AAY8B,KAAlB,eAAA,GAAkB,OAAA,CAC5B,IAD4B,CACvB,UADuB,CACZ,WADY,CAAA,kBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,SAAA,CAAA,CAAA;AAAO,KAIzB,aAAA,GAAgB,OAJS,CAKnC,IALmC,CAMjC,UANiC,CAMtB,WANsB,CAAA,gBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,SAAA,GAAA,OAAA,GAAA,UAAA,GAAA,WAAA,CAAA,CAAA;AAIzB,KAOA,YAAA,GAAe,OAPF,CAQvB,IARuB,CASrB,UATqB,CASV,WATU,CAAA,eAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,UAAA,GAAA,WAAA,GAAA,SAAA,CAAA,CAAA;AAEV,KAYH,aAAA,GAAgB,OAZb,CAab,IAba,CAcX,UAdW,CAcA,WAdA,CAAA,gBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,OAAA,GAAA,UAAA,GAAA,WAAA,CAAA,CAAA;AAAX,KAmBQ,QAAA,GAAW,OAnBnB,CAoBF,IApBE,CAoBG,UApBH,CAoBc,WApBd,CAAA,WAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,UAAA,GAAA,WAAA,CAAA,CAAA;AADF,KAwBU,cAAA,GAAiB,OAxB3B,CAyBA,UAzBA,CAyBW,WAzBX,CAAA,iBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,CAAA;AAD0B,KA6BhB,qBAAA,GAAwB,OA7BR,CA8B1B,IA9B0B,CA8BrB,UA9BqB,CA8BV,WA9BU,CAAA,wBAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,aAAA,CAAA,CAAA;AAAO,KAiCvB,SAAA,GAAY,OAjCW,CAkCjC,IAlCiC,CAkC5B,UAlC4B,CAkCjB,WAlCiB,CAAA,OAAA,CAAA,CAAA,CAAA,CAAA,CAAA,EAAA,MAAA,GAAA,OAAA,GAAA,UAAA,CAAA,CAAA;AAOvB,UA4CK,YAAA,CA5CO;EAET,MAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAAX,aAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EADF,aAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EADyB,YAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAAO,WAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAOtB,OAAA,CAAA,EA2CA,MA3Ca,CAAA,MAAA,EAAA,OAAA,CAAA,GAAA,IAAA;EAEV,GAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAAX,SAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EADF,OAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAD0B,MAAA,CAAA,EA+CjB,MA/CiB,CAAA,MAAA,EAAA,OAAA,CAAA,GAAA,IAAA;EAAO,QAAA,CAAA,EAAA,MAAA,GAAA,IAAA;EAOvB,gBAAQ,CAAA,EAAA,MAAA,GAAA,IAAA;;AACb,cA4CM,UAAA,CA5CN;EAAL,SAAA,GAAA,EA6Cc,WA7Cd;EADqB,SAAA,SAAA,EA+CD,MA/CC,CAAA,MAAA,EAAA,OAAA,CAAA;EAAO,WAAA,CAAA,EAAA,EAkDtB,WAlDsB,EAAA,IAAA,EAmDpB,YAnDoB,GAAA;IAIlB,OAAA,EAAA,MAAc;EACb,CAAA;EAAX,IAAA,OAAA,CAAA,CAAA,EAAA,MAAA;EAD2B,IAAA,EAAA,CAAA,CAAA,EAuEjB,WAvEiB;EAAO,KAAA,CAAA,OAAA,EAAA,MAAA,EAAA,SAAA,CAAA,EA2EA,YA3EA,CAAA,EA2EoB,UA3EpB;EAIxB,QAAA,MAAA;EACM,gBAAA,CAAA,OAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAwIwB,eAxIxB,CAAA,EAAA,MAAA;EAAX,cAAA,CAAA,OAAA,EAAA,MAAA,EAAA,KAAA,EAAA,MAAA,EAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAoJG,aApJH,CAAA,EAAA,MAAA;EAAL,aAAA,CAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,OAAA,EAAA,OAAA,EAAA,IAAA,CAAA,EAmKQ,YAnKR,CAAA,EAAA,MAAA;EADkC,cAAA,CAAA,KAAA,EAAA,MAAA,EAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAkL1B,aAlL0B,CAAA,EAAA,MAAA;EAAO,SAAA,CAAA,QAAA,EAAA,MAAA,EAAA,SAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EA4LY,QA5LZ,CAAA,EAAA,MAAA;EAI/B,eAAS,CAAA,IAAA,CAAA,EAgMG,cAhMH,CAAA,EAAA,IAAA;EACH,sBAAA,CAAA,WAAA,EAsMD,kBAtMC,EAAA,IAAA,CAAA,EAuMR,qBAvMQ,CAAA,EAAA,IAAA;EAAX,KAAA,CAAA,IAAA,EAAA,MAAA,EAAA,KAAA,EAAA,MAAA,EAAA,QAAA,EAAA,MAAA,EAAA,IAAA,CAAA,EAmNG,SAnNH,CAAA,EAAA,IAAA;EAAL,OAAA,CAAA,IAqCM,CArCN,EAAA;IADsB,SAAA,CAAA,EAAA,MAAA,GAAA,IAAA;IAAO,kBAAA,CAAA,EAAA,MAAA,GAAA,IAAA;IAkBd,MAAA,CAAA,EAAA,MAAY,GAAA,IAMjB;IASC,QAAA,CAAU,EAAA,MAAA,GAAA,IAAA;IACP,gBAAA,CAAA,EAAA,MAAA,GAAA,IAAA;IACM,SAAA,CAAA,EAAA,OAAA;EAGd,CAAA,CAAA,EAiMH,OAjMG;EACE,KAAA,CAAA,CAAA,EAAA,OAAA;EAwBE,QAAA,CAAA,CAAA,EAAA,IAAA"}

package/dist/providers/mistral.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"mistral.d.ts","names":[],"sources":["../../src/providers/mistral.ts"],"sourcesContent":[],"mappings":";;;;;;AAaM,cADO,iBACS,EAAM,OAAA;AAI5B,cAJM,cAIyB,EAJT,MAKT,CAAA,MAAA,EAAA,OAEK,CAAA,GAAA,IAAA;AAOD,UAVA,cAAA,CAUA;EAEM,SAAA,EAXV,aAWU;EAJM,MAAA,CAAA,EAAA,MAAA;EAAc,aAAA,CAAA,EALzB,aAKyB,GAAA,IAAA;EAoC9B;EAI2B,aAAA,CAAA,EAAA,OAAA;;AAQW,cAhDtC,OAAA,SAAgB,cAAA,CAgDsB;EAoF5B,QAAA,OAAA;EAA0B,SAAA,IAAA,EAlIhC,WAkIgC;EAAO,WAAA,CAAA,OAAA,EAhIjC,cAgIiC;;;cAhG3C,WAAA;;;wCAI2B;mBAQf,0BAA0B;iBAoF5B,0BAA0B"}
+{"version":3,"file":"mistral.d.ts","names":[],"sources":["../../src/providers/mistral.ts"],"sourcesContent":[],"mappings":";;;;;;AAaM,cADO,iBACS,EAAA,OAAM;AAI5B,cAJM,cAIyB,EAJT,MAKT,CAAA,MAAA,EAAA,OAEK,CAAA,GAAA,IAAA;AAOD,UAVA,cAAA,CAUA;EAEM,SAAA,EAXV,aAWU;EAJM,MAAA,CAAA,EAAA,MAAA;EAAc,aAAA,CAAA,EALzB,aAKyB,GAAA,IAAA;EAoC9B;EAI2B,aAAA,CAAA,EAAA,OAAA;;AAQW,cAhDtC,OAAA,SAAgB,cAAA,CAgDsB;EAoF5B,QAAA,OAAA;EAA0B,SAAA,IAAA,EAlIhC,WAkIgC;EAAO,WAAA,CAAA,OAAA,EAhIjC,cAgIiC;;;cAhG3C,WAAA;;;wCAI2B;mBAQf,0BAA0B;iBAoF5B,0BAA0B"}

package/dist/utils/logger.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"logger.d.ts","names":[],"sources":["../../src/utils/logger.ts"],"sourcesContent":[],"mappings":";UAAiB,MAAA;EAAA,KAAA,CAAA,OAAM,EAAA,MAAA,CAAA,EAAA,IAAA;EAcP,KAAA,CAAA,OAAS,EAAA,MAAA,CAAA,EAAuB,IAAA;;;;iBAAhC,SAAA,uBAAgC"}
+{"version":3,"file":"logger.d.ts","names":[],"sources":["../../src/utils/logger.ts"],"sourcesContent":[],"mappings":";UAAiB,MAAA;EAAA,KAAA,CAAA,OAAM,EAAA,MAAA,CAAA,EAAA,IAAA;EAcP,KAAA,CAAA,OAAS,EAAA,MAAA,CAAA,EAAA,IAAuB;;;;iBAAhC,SAAA,uBAAgC"}

package/llms-full.txt

@@ -1,5 +1,5 @@
 # llms-full.txt
-# @amplitude/ai 0.3.6 — Complete API Reference for AI Coding Agents
+# @amplitude/ai 0.3.7 — Complete API Reference for AI Coding Agents
 #
 # This file is the definitive guide for any coding agent instrumenting
 # a JavaScript/TypeScript AI application with @amplitude/ai.
@@ -67,6 +67,7 @@ const ai = new AmplitudeAI({
 
 ### patch(options) / unpatch()
 Zero-code instrumentation. Monkey-patches all detected provider SDKs.
+Captures aggregate `[Agent] AI Response` events only — no user identity, no funnels. Use as a quick verification or for legacy codebases where you cannot modify call sites.
 ```
 import { patch, unpatch } from '@amplitude/ai';
 patch({ amplitudeAI: ai });
@@ -395,6 +396,8 @@ with code examples for every step. Read this for guided instrumentation.
 
 ## CLI
 
+- `amplitude-ai` — Print instrumentation prompt for AI coding agents
+- `amplitude-ai --print-guide` — Print the full amplitude-ai.md guide to stdout
 - `amplitude-ai mcp` — Start the MCP server for AI coding agents
 - `amplitude-ai doctor [--json] [--no-mock-check]` — Validate environment and event pipeline
 - `amplitude-ai status [--json]` — Show SDK version, installed providers, and env config

package/llms.txt

@@ -1,7 +1,7 @@
 <!-- GENERATED FILE: do not edit manually. Update scripts/generate-agent-docs.mjs instead. -->
 # llms.txt
 package=@amplitude/ai
-version=0.3.6
+version=0.3.7
 
 [mcp.tools]
 get_event_schema

package/mcp.schema.json

@@ -1,7 +1,7 @@
 {
   "generated": true,
   "package": "@amplitude/ai",
-  "version": "0.3.6",
+  "version": "0.3.7",
   "prompt": "instrument_app",
   "tools": [
     "get_event_schema",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@amplitude/ai",
-  "version": "0.3.6",
+  "version": "0.3.7",
   "private": false,
   "description": "Amplitude AI SDK - LLM usage tracking for Amplitude Analytics",
   "keywords": [