docs: add iFlow Search LangChain JS tools

src/oss/javascript/integrations/tools/iflow_search.mdx

@@ -0,0 +1,173 @@
+---
+title: "iFlow Search integration"
+description: "Integrate with the iFlow Search tools using LangChain JavaScript."
+---
+
+[iFlow Search](https://platform.iflow.cn) is a search API that provides web search, image search, and URL fetch primitives. `@iflow-ai/search-langchain` is a community-maintained, third-party npm package that wraps those primitives as three LangChain JavaScript tools. It is **not part of LangChain core** and is currently in a pre-1.0 / prerelease state.
+
+## Overview
+
+### Integration details
+
+| Class | Package | PY support | Version |
+| :--- | :--- | :---: | :---: |
+| `createIFlowWebSearchTool` / `createIFlowImageSearchTool` / `createIFlowWebFetchTool` / `createIFlowSearchTools` | [`@iflow-ai/search-langchain`](https://www.npmjs.com/package/@iflow-ai/search-langchain) (community, third-party) | ❌ | ![NPM - Version](https://img.shields.io/npm/v/@iflow-ai/search-langchain?style=flat-square&label=%20&) |
+
+> The package is maintained outside LangChain core. See the source repository at [`zhengyanglsun/iflow-search-js`](https://github.com/zhengyanglsun/iflow-search-js) and the tracking issue [`langchain-ai/langchainjs#10931`](https://github.com/langchain-ai/langchainjs/issues/10931). Until a stable `0.1.0` ships, prefer the `@next` dist-tag.
+
+### Tool features
+
+| [Returns artifact](/oss/langchain/tools) | Native async | Return data | Pricing |
+| :---: | :---: | :--- | :--- |
+| ✅ (via `responseFormat: "content_and_artifact"`) | ✅ | Web results (titles, URLs, snippets); image results (image URLs, titles, source pages); URL fetch (title, URL, normalized content) | See [iFlow platform](https://platform.iflow.cn) |
+
+## Setup
+
+The integration lives in the `@iflow-ai/search-langchain` package, which depends on `@iflow-ai/search-core` and `@langchain/core`. Install all three explicitly:
+
+<CodeGroup>
+```bash npm
+npm install @iflow-ai/search-langchain @iflow-ai/search-core @langchain/core
+```
+```bash yarn
+yarn add @iflow-ai/search-langchain @iflow-ai/search-core @langchain/core
+```
+```bash pnpm
+pnpm add @iflow-ai/search-langchain @iflow-ai/search-core @langchain/core
+```
+</CodeGroup>
+
+To install the prerelease (`@next`) while the package is pre-1.0:
+
+<CodeGroup>
+```bash npm
+npm install @iflow-ai/search-langchain@next @iflow-ai/search-core@next @langchain/core
+```
+```bash yarn
+yarn add @iflow-ai/search-langchain@next @iflow-ai/search-core@next @langchain/core
+```
+```bash pnpm
+pnpm add @iflow-ai/search-langchain@next @iflow-ai/search-core@next @langchain/core
+```
+</CodeGroup>
+
+Peer dependencies declared by the package:
+
+- `@langchain/core`: `^0.3.0 || ^1.0.0`
+- `zod`: `^3.25.0 || ^4.0.0`
+
+### Credentials
+
+Set up an [iFlow API key](https://platform.iflow.cn) and pass it to the tool factories via an environment variable. Use a placeholder in source files; inject the real value at runtime through your shell, secret store, or `direnv` block.
+
+```typescript
+process.env.IFLOW_API_KEY = "YOUR_IFLOW_API_KEY";
+```
+
+Do not hard-code the key in source files that get committed.
+
+#### Security boundary
+
+`IFLOW_API_KEY` is read by the iFlow package only when you construct a tool, attached as the upstream `Authorization: Bearer …` header by the internal `@iflow-ai/search-core` client, and sent only to the iFlow Search API (`https://platform.iflow.cn`). The key is never written to disk, logged, or surfaced to the LangChain agent or any downstream tool consumer.
+
+It's also helpful (but not needed) to set up [LangSmith](https://smith.langchain.com/) for observability:
+
+```typescript
+process.env.LANGSMITH_TRACING = "true";
+process.env.LANGSMITH_API_KEY = "your-api-key";
+```
+
+## Instantiation
+
+The package exports three individual tool factories and one convenience factory that builds all three at once. All four accept the same options object (`IFlowLangChainOptions`).
+
+```typescript
+import {
+  createIFlowWebSearchTool,
+  createIFlowImageSearchTool,
+  createIFlowWebFetchTool,
+  createIFlowSearchTools,
+  type IFlowLangChainOptions,
+} from "@iflow-ai/search-langchain";
+
+const options: IFlowLangChainOptions = {
+  apiKey: process.env.IFLOW_API_KEY!,
+  // baseUrl?: string;     // override for testing / private deployments
+  // timeoutMs?: number;   // per-request timeout
+};
+
+// Build one tool at a time:
+const webSearch = createIFlowWebSearchTool(options);
+const imageSearch = createIFlowImageSearchTool(options);
+const webFetch = createIFlowWebFetchTool(options);
+
+// Or build all three at once:
+const tools = createIFlowSearchTools(options);
+// tools[0] → iflow_web_search
+// tools[1] → iflow_image_search
+// tools[2] → iflow_web_fetch
+```
+
+Each factory returns a standard LangChain tool created via `tool()` from `@langchain/core/tools`, configured with `responseFormat: "content_and_artifact"`. The LLM sees a compact text summary; your application code sees the full normalized result on the `ToolMessage.artifact` field.
+
+## Invocation
+
+### Direct call
+
+```typescript
+const result = await webSearch.invoke({
+  query: "LangChain JavaScript",
+  count: 3,
+});
+console.log(result);
+```
+
+The web-search and image-search tools accept `{ query: string; count?: number }`. The web-fetch tool accepts `{ url: string }` and returns the normalized readable contents of a single page.
+
+### Use with `createReactAgent`
+
+```typescript
+import { ChatOpenAI } from "@langchain/openai";
+import { createReactAgent } from "@langchain/langgraph/prebuilt";
+import {
+  createIFlowSearchTools,
+  type IFlowLangChainOptions,
+} from "@iflow-ai/search-langchain";
+
+const options: IFlowLangChainOptions = {
+  apiKey: process.env.IFLOW_API_KEY!,
+};
+
+const llm = new ChatOpenAI({ model: "gpt-4o-mini" });
+const tools = createIFlowSearchTools(options); // all three tools
+
+const agent = createReactAgent({ llm, tools });
+
+const response = await agent.invoke({
+  messages: [
+    {
+      role: "user",
+      content:
+        "Find a recent article about LangGraph multi-agent patterns, then fetch the page and summarize the key ideas.",
+    },
+  ],
+});
+```
+
+## Tool names
+
+The three tools are registered with stable names so prompts can reference them directly:
+
+- `iflow_web_search` — web search; returns titles, URLs, and snippets.
+- `iflow_image_search` — image search; returns image URLs, titles, and source pages.
+- `iflow_web_fetch` — fetch the readable contents of a single URL.
+
+Renaming any of these would be a breaking change for the package.
+
+## Related
+
+- [`@iflow-ai/search-langchain` on npm](https://www.npmjs.com/package/@iflow-ai/search-langchain)
+- Source repository: [`zhengyanglsun/iflow-search-js`](https://github.com/zhengyanglsun/iflow-search-js)
+- Tracking issue in LangChain JS: [`langchain-ai/langchainjs#10931`](https://github.com/langchain-ai/langchainjs/issues/10931)
+- Tool [conceptual guide](/oss/langchain/tools)
+- Tool [how-to guides](/oss/langchain/tools)

src/oss/javascript/integrations/tools/index.mdx

@@ -117,6 +117,13 @@ The following platforms provide access to multiple tools and services through a
     arrow="true"
     cta="View guide"
   />
+  <Card
+    title="iFlow Search"
+    icon="link"
+    href="/oss/integrations/tools/iflow_search"
+    arrow="true"
+    cta="View guide"
+  />
   <Card
     title="JigsawStack Tool"
     icon="link"