@jaypie/mcp 0.8.70 → 0.8.72

package/dist/suites/docs/index.js

@@ -9,7 +9,7 @@ import { gt } from 'semver';
 /**
  * Docs Suite - Documentation services (skill, version, release_notes)
  */
-const BUILD_VERSION_STRING = "@jaypie/mcp@0.8.70#ec155ccd"
+const BUILD_VERSION_STRING = "@jaypie/mcp@0.8.72#c0c8b6f2"
     ;
 const __filename$1 = fileURLToPath(import.meta.url);
 const __dirname$1 = path.dirname(__filename$1);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jaypie/mcp",
-  "version": "0.8.70",
+  "version": "0.8.72",
   "description": "Jaypie MCP",
   "repository": {
     "type": "git",

package/release-notes/llm/1.2.39.md

@@ -0,0 +1,11 @@
+---
+version: 1.2.39
+date: 2026-06-16
+summary: Fix MODEL.GPT_MINI/GPT_NANO typo (gpt-5.5-* → gpt-5.4-*) to match OpenAI provider constants
+---
+
+## @jaypie/llm 1.2.39
+
+### Fixes
+
+- **Model constants**: `MODEL.GPT_MINI` and `MODEL.GPT_NANO` were pinned to nonexistent `gpt-5.5-mini`/`gpt-5.5-nano`; corrected to `gpt-5.4-mini`/`gpt-5.4-nano`, matching `PROVIDER.OPENAI.MODEL.SMALL`/`TINY` and the CI test matrix.

package/release-notes/llm/1.2.40.md

@@ -0,0 +1,14 @@
+---
+version: 1.2.40
+date: 2026-06-17
+summary: Opt-in Datadog LLM Observability span emission from operate()/stream() via DD_LLMOBS_ENABLED
+---
+
+## @jaypie/llm 1.2.40
+
+### Features
+
+- **LLM Observability**: When `DD_LLMOBS_ENABLED` is set, `operate()` and `stream()` emit Datadog `llm`/`tool` spans (annotated with input, output, and token metrics) plus an enclosing `agent`/`llm` span — no code changes required (#373).
+  - Lazy, **bundler-safe** `dd-trace` resolution (computed module specifier); `dd-trace` is not a dependency. Fully no-op when the flag is unset or `dd-trace` is absent, and instrumentation failures never break the underlying call.
+  - Parenting is AsyncLocalStorage-based: spans attach to any active enclosing LLMObs span (e.g. a consumer `llmobs.trace({ kind: "workflow" })`).
+  - `operate()` spans nest fully (model + tool under the enclosing span). `stream()` spans attach to an active enclosing span but are siblings within a single stream (the held-open streamed model span is not active when tools run).

package/release-notes/mcp/0.8.71.md

@@ -0,0 +1,10 @@
+---
+version: 0.8.71
+date: 2026-06-16
+summary: Add @jaypie/llm 1.2.39 release notes; document model-first Llm constructor
+---
+
+## Changes
+
+- Adds release notes for `@jaypie/llm` 1.2.39 (`MODEL.GPT_MINI`/`GPT_NANO` typo fix).
+- Documents that `new Llm(model)` accepts a model name as its first argument (provider auto-detected, model retained) in the `llm` skill.

package/release-notes/mcp/0.8.72.md

@@ -0,0 +1,10 @@
+---
+version: 0.8.72
+date: 2026-06-17
+summary: Add @jaypie/llm 1.2.40 release notes; document LLM Observability (DD_LLMOBS_ENABLED) in the llm skill
+---
+
+## Changes
+
+- Adds release notes for `@jaypie/llm` 1.2.40 (opt-in Datadog LLM Observability span emission).
+- Documents the `DD_LLMOBS_ENABLED` LLM Observability behavior and span parenting semantics in the `llm` skill.

package/skills/llm.md

@@ -315,6 +315,15 @@ const review1 = await llm.operate(code1);
 const review2 = await llm.operate(code2);
 ```
 
+The first constructor argument may be a provider name **or** a model name. When a model name is passed, the provider is auto-detected and the model is retained:
+
+```typescript
+import Llm, { LLM } from "@jaypie/llm";
+
+const llm = new Llm("claude-sonnet-4-6");      // -> anthropic, claude-sonnet-4-6
+const flash = new Llm(LLM.MODEL.GEMINI_FLASH); // -> google, gemini-3.5-flash
+```
+
 ## Environment Variables
 
 ```bash
@@ -327,6 +336,28 @@ XAI_API_KEY         # Required for xAI (Grok)
 
 Keys are resolved via `getEnvSecret()` which supports AWS Secrets Manager.
 
+## LLM Observability (Datadog)
+
+Set `DD_LLMOBS_ENABLED` (any truthy value except `false`/`0`) and `operate()`/`stream()` emit Datadog [LLM Observability](https://docs.datadoghq.com/llm_observability/) spans with **no code changes**:
+
+- Enclosing span per call (`agent` when tools are configured, else `llm`)
+- Child `llm` span per model request — annotated with input, output, token metrics
+- Child `tool` span per tool execution — annotated with args + result
+
+```bash
+DD_LLMOBS_ENABLED=true   # opt in; unset = fully no-op
+DD_LLMOBS_ML_APP=my-app  # ML app name (dd-trace standard)
+```
+
+Behavior:
+
+- **Opt-in and lazy** — `dd-trace` is resolved at runtime via a computed module specifier, so it is **bundler-safe** (esbuild will not bundle it) and **not** a dependency. Absence is a silent no-op; instrumentation failures never break the LLM call.
+- **Parenting is AsyncLocalStorage-based** — spans attach to whatever LLMObs span is active when created. Wrapping a call in a consumer span (e.g. `llmobs.trace({ kind: "workflow" }, () => Llm.operate(...))`) nests ours under it. The Datadog Lambda layer provides APM spans automatically, but not an enclosing *LLMObs* span around an arbitrary handler.
+- **`operate()`** spans form a full tree (model + tool nest under the enclosing span).
+- **`stream()`** spans attach to any active enclosing span, but within a single stream the `llm` and `tool` spans are **siblings** — the streamed model span is held open across `yield` boundaries, so it is not the active span when tools run.
+
+For esbuild-bundled Lambda handlers that also want dd-trace auto-instrumentation, wire the `dd-trace/esbuild` plugin with `keepNames: true`; the spans above do not require it.
+
 ## Error Handling
 
 The package auto-retries rate limits and transient errors: