@axonpush/wizard 0.0.4 → 0.0.6

package/skills/ts-langgraph/SKILL.md

@@ -0,0 +1,46 @@
+---
+name: ts-langgraph
+description: Integrate AxonPush tracing into a TypeScript LangGraph project
+---
+
+# AxonPush + LangGraph (TypeScript) Integration
+
+Integrate AxonPush tracing into a TypeScript LangGraph project.
+
+## What gets added
+
+- `AxonPushLangGraphHandler` that extends `AxonPushCallbackHandler` with graph node-level tracing
+- Events: `chain.start`, `chain.end`, `graph.node.start`, `graph.node.end`, `llm.start`, `llm.end`, `tool.*.start`, `tool.end`
+
+## Reference Code
+
+```typescript
+import { AxonPush } from "@axonpush/sdk";
+import { AxonPushLangGraphHandler } from "@axonpush/sdk/integrations/langgraph";
+
+const axonpush = new AxonPush({
+  apiKey: process.env.AXONPUSH_API_KEY!,
+  tenantId: process.env.AXONPUSH_TENANT_ID!,
+  baseUrl: process.env.AXONPUSH_BASE_URL,
+});
+
+const handler = new AxonPushLangGraphHandler({
+  client: axonpush,
+  channelId: Number(process.env.AXONPUSH_CHANNEL_ID),
+  agentId: "langgraph-agent",
+});
+
+// const result = await graph.invoke(input, { callbacks: [handler] });
+```
+
+## Steps
+
+1. Install `@axonpush/sdk` using the project's package manager
+2. Add AXONPUSH_API_KEY, AXONPUSH_TENANT_ID, AXONPUSH_BASE_URL, AXONPUSH_CHANNEL_ID to .env
+3. Find the main file where the LangGraph graph is invoked
+4. Add the imports and client initialization (as module-level code)
+5. Add `{ callbacks: [handler] }` to `.invoke()` calls
+
+## Fail-Open
+
+The SDK is fail-open by default (`failOpen: true`). If AxonPush is unreachable, tracing callbacks are silently suppressed.

package/skills/ts-llamaindex/SKILL.md

@@ -0,0 +1,52 @@
+---
+name: ts-llamaindex
+description: Integrate AxonPush tracing into a TypeScript LlamaIndex project
+---
+
+# AxonPush + LlamaIndex (TypeScript) Integration
+
+Integrate AxonPush tracing into a TypeScript LlamaIndex project.
+
+## What gets added
+
+- `AxonPushLlamaIndexHandler` with LLM, embedding, retriever, and query lifecycle methods
+- Events: `llm.start`, `llm.end`, `llm.token`, `embedding.start`, `embedding.end`, `retriever.query`, `retriever.result`, `query.start`, `query.end`
+
+## Reference Code
+
+```typescript
+import { AxonPush } from "@axonpush/sdk";
+import { AxonPushLlamaIndexHandler } from "@axonpush/sdk/integrations/llamaindex";
+
+const axonpush = new AxonPush({
+  apiKey: process.env.AXONPUSH_API_KEY!,
+  tenantId: process.env.AXONPUSH_TENANT_ID!,
+  baseUrl: process.env.AXONPUSH_BASE_URL,
+});
+
+const handler = new AxonPushLlamaIndexHandler({
+  client: axonpush,
+  channelId: Number(process.env.AXONPUSH_CHANNEL_ID),
+  agentId: "llamaindex",
+});
+
+// Call handler methods at the appropriate points:
+// handler.onQueryStart("What is...");
+// handler.onLLMStart("gpt-4", 1);
+// handler.onLLMEnd(output);
+// handler.onRetrieverStart("What is...");
+// handler.onRetrieverEnd(5);
+// handler.onQueryEnd(response);
+```
+
+## Steps
+
+1. Install `@axonpush/sdk` using the project's package manager
+2. Add AXONPUSH_API_KEY, AXONPUSH_TENANT_ID, AXONPUSH_BASE_URL, AXONPUSH_CHANNEL_ID to .env
+3. Find the main query engine or retrieval pipeline
+4. Add imports and create the handler
+5. Call the appropriate handler methods at each lifecycle point
+
+## Fail-Open
+
+The SDK is fail-open by default (`failOpen: true`). If AxonPush is unreachable, handler calls are silently suppressed.

package/skills/ts-mastra/SKILL.md

@@ -0,0 +1,51 @@
+---
+name: ts-mastra
+description: Integrate AxonPush tracing into a Mastra project
+---
+
+# AxonPush + Mastra Integration
+
+Integrate AxonPush tracing into a project using Mastra.
+
+## What gets added
+
+- `AxonPushMastraHooks` with tool and workflow lifecycle methods
+- Events: `tool.*.start`, `tool.*.end`, `workflow.start`, `workflow.end`, `workflow.error`
+
+## Reference Code
+
+```typescript
+import { AxonPush } from "@axonpush/sdk";
+import { AxonPushMastraHooks } from "@axonpush/sdk/integrations/mastra";
+
+const axonpush = new AxonPush({
+  apiKey: process.env.AXONPUSH_API_KEY!,
+  tenantId: process.env.AXONPUSH_TENANT_ID!,
+  baseUrl: process.env.AXONPUSH_BASE_URL,
+});
+
+const hooks = new AxonPushMastraHooks({
+  client: axonpush,
+  channelId: Number(process.env.AXONPUSH_CHANNEL_ID),
+  agentId: "mastra",
+});
+
+// In your workflow or agent:
+// hooks.onWorkflowStart("my-workflow", input);
+// hooks.beforeToolUse("tool-name", input);
+// hooks.afterToolUse("tool-name", output);
+// hooks.onWorkflowEnd("my-workflow", output);
+// hooks.onWorkflowError("my-workflow", error);
+```
+
+## Steps
+
+1. Install `@axonpush/sdk` using the project's package manager
+2. Add AXONPUSH_API_KEY, AXONPUSH_TENANT_ID, AXONPUSH_BASE_URL, AXONPUSH_CHANNEL_ID to .env
+3. Find the main workflow or agent entry point
+4. Add imports and create the hooks instance
+5. Call the appropriate hook methods at workflow/tool lifecycle points
+
+## Fail-Open
+
+The SDK is fail-open by default (`failOpen: true`). If AxonPush is unreachable, hook calls are silently suppressed.

package/skills/ts-openai-agents/SKILL.md

@@ -0,0 +1,45 @@
+---
+name: ts-openai-agents
+description: Integrate AxonPush tracing into a TypeScript OpenAI Agents SDK project
+---
+
+# AxonPush + OpenAI Agents SDK (TypeScript) Integration
+
+Integrate AxonPush tracing into a TypeScript project using the OpenAI Agents SDK.
+
+## What gets added
+
+- `AxonPushRunHooks` that traces agent runs, tool calls, and handoffs
+- Events: `agent.run.start`, `agent.run.end`, `tool.*.start`, `tool.*.end`, `agent.handoff`
+
+## Reference Code
+
+```typescript
+import { AxonPush } from "@axonpush/sdk";
+import { AxonPushRunHooks } from "@axonpush/sdk/integrations/openai-agents";
+
+const axonpush = new AxonPush({
+  apiKey: process.env.AXONPUSH_API_KEY!,
+  tenantId: process.env.AXONPUSH_TENANT_ID!,
+  baseUrl: process.env.AXONPUSH_BASE_URL,
+});
+
+const hooks = new AxonPushRunHooks({
+  client: axonpush,
+  channelId: Number(process.env.AXONPUSH_CHANNEL_ID),
+});
+
+// const result = await Runner.run(agent, input, { hooks });
+```
+
+## Steps
+
+1. Install `@axonpush/sdk` using the project's package manager
+2. Add AXONPUSH_API_KEY, AXONPUSH_TENANT_ID, AXONPUSH_BASE_URL, AXONPUSH_CHANNEL_ID to .env
+3. Find the main file where Runner.run() is called
+4. Add the imports and AxonPush client initialization
+5. Pass `hooks` to `Runner.run()`
+
+## Fail-Open
+
+The SDK is fail-open by default (`failOpen: true`). If AxonPush is unreachable, tracing hooks are silently suppressed.

package/skills/ts-vercel-ai/SKILL.md

@@ -0,0 +1,52 @@
+---
+name: ts-vercel-ai
+description: Integrate AxonPush tracing into a Vercel AI SDK project
+---
+
+# AxonPush + Vercel AI SDK Integration
+
+Integrate AxonPush tracing into a project using the Vercel AI SDK.
+
+## What gets added
+
+- `axonPushMiddleware` that wraps generateText/streamText with tracing
+- Events: `llm.start`, `llm.end`, `llm.token` (streaming)
+
+## Reference Code
+
+```typescript
+import { AxonPush } from "@axonpush/sdk";
+import { axonPushMiddleware } from "@axonpush/sdk/integrations/vercel-ai";
+import { generateText, streamText, wrapLanguageModel } from "ai";
+
+const axonpush = new AxonPush({
+  apiKey: process.env.AXONPUSH_API_KEY!,
+  tenantId: process.env.AXONPUSH_TENANT_ID!,
+  baseUrl: process.env.AXONPUSH_BASE_URL,
+});
+
+const middleware = axonPushMiddleware({
+  client: axonpush,
+  channelId: Number(process.env.AXONPUSH_CHANNEL_ID),
+  agentId: "vercel-ai",
+});
+
+// Option A: wrap the model
+// const tracedModel = wrapLanguageModel({ model: openai("gpt-4o"), middleware });
+// const result = await generateText({ model: tracedModel, prompt: "..." });
+
+// Option B: pass middleware directly (if supported)
+// const result = await generateText({ model, prompt: "...", experimental_telemetry: { middleware } });
+```
+
+## Steps
+
+1. Install `@axonpush/sdk` using the project's package manager
+2. Add AXONPUSH_API_KEY, AXONPUSH_TENANT_ID, AXONPUSH_BASE_URL, AXONPUSH_CHANNEL_ID to .env
+3. Find files that call `generateText()` or `streamText()`
+4. Add imports and create the middleware
+5. Wrap the language model using `wrapLanguageModel({ model, middleware })`
+
+## Fail-Open
+
+The SDK is fail-open by default (`failOpen: true`). If AxonPush is unreachable, the middleware passes through without tracing.