n8n-nodes-berget-mk 0.3.0 → 0.4.0

package/README.md

@@ -1,13 +1,14 @@
 # n8n-nodes-berget-mk
 
-n8n community node for [Berget AI](https://berget.ai), packaged as a single installable module. Maintained by Micke Kring.
+n8n community nodes for [Berget AI](https://berget.ai), packaged as a single installable module. Maintained by Micke Kring.
 
-Two nodes, two purposes:
+Three nodes:
 
-- **Berget AI** — a multi-resource action node with five resources: **Chat**, **Embeddings**, **OCR**, **Rerank**, **Speech to Text**. Use this for one-shot calls.
-- **Berget AI Chat Model** — a sub-node that plugs into n8n's built-in **AI Agent**, **Basic LLM Chain**, and any other LangChain-based node. Use this to drive an agent with Berget AI as the underlying LLM. Exposes `reasoning_effort` and the full standard LLM parameter set.
+- **Berget AI** — multi-resource action node for one-shot calls: **Chat** (completions, classification), **OCR** (document text extraction), **Rerank** (document reranking), and **Speech to Text** (Swedish-tuned KB-Whisper). Can also be exposed as a tool to an AI Agent.
+- **Berget AI Chat Model** — sub-node that plugs into n8n's built-in **AI Agent**, **Basic LLM Chain**, and other LangChain-based nodes. Exposes `reasoning_effort` and the full standard LLM parameter set.
+- **Berget AI Embeddings Model** — sub-node that plugs into n8n's **Vector Store** nodes (Supabase, Qdrant, Pinecone, PGVector, etc.) and **Question and Answer Chain**.
 
-> ⚠️ **Experimental — actively developed.** This package is pre-1.0 and the shape of nodes may change between minor releases. Pin a specific version in production workflows until `1.0.0`. See [CHANGELOG.md](CHANGELOG.md) for breaking changes.
+> ⚠️ **Experimental — actively developed.** This package is pre-1.0 and may break between minor releases. Pin a specific version in production workflows until `1.0.0`. See [CHANGELOG.md](CHANGELOG.md) for breaking changes.
 
 ## Install
 
@@ -19,19 +20,31 @@ n8n-nodes-berget-mk
 
 Then add a **Berget AI API** credential with your API key from [berget.ai](https://berget.ai).
 
-## Using the Berget AI action node
+## Typical workflows
 
-1. Drop **Berget AI** onto the canvas.
-2. Select a **Resource** (Chat, Embeddings, OCR, Rerank, or Speech to Text).
-3. Fill in the resource-specific fields (model, input, options).
-4. Execute.
+### One-shot chat / classification
 
-## Using Berget with n8n's AI Agent
+1. Drop **Berget AI** onto the canvas, pick Resource = **Chat**, select a model, add a user message. Execute.
 
-1. Add n8n's built-in **AI Agent** node.
-2. Add a **Berget AI Chat Model** node and connect it to the Agent's **Chat Model** socket.
-3. Pick a Berget AI chat model and optionally configure `reasoning_effort` for reasoning-capable models (GPT-OSS, DeepSeek R1).
-4. Add Memory and Tool sub-nodes as usual — they work with Berget as the underlying LLM.
+### Agent with tools and memory
+
+1. Add n8n's built-in **AI Agent**.
+2. Add **Berget AI Chat Model** and connect it to the Agent's Chat Model socket.
+3. Add Memory and Tool sub-nodes as needed — they work with Berget as the underlying LLM.
+
+### RAG / vector search
+
+1. Add a Vector Store node (Supabase, Qdrant, etc.) or a Question and Answer Chain.
+2. Add **Berget AI Embeddings Model** and connect it to the Embedding socket.
+3. Index documents or query as usual.
+
+### Document extraction
+
+1. Drop **Berget AI** onto the canvas, pick Resource = **OCR**, provide a document URL or base64 data, and pick your output format (Markdown or JSON).
+
+### Swedish speech transcription
+
+1. Drop **Berget AI** onto the canvas, pick Resource = **Speech to Text**, pick a model (defaults to `KB-Whisper-Large`), and point at an audio file.
 
 ## Changelog
 

package/dist/nodes/BergetAi/BergetAi.node.d.ts

@@ -4,7 +4,6 @@ export declare class BergetAi implements INodeType {
     methods: {
         loadOptions: {
             getChatModels(this: ILoadOptionsFunctions): Promise<INodePropertyOptions[]>;
-            getEmbeddingsModels(this: ILoadOptionsFunctions): Promise<INodePropertyOptions[]>;
             getRerankModels(this: ILoadOptionsFunctions): Promise<INodePropertyOptions[]>;
             getSpeechModels(this: ILoadOptionsFunctions): Promise<INodePropertyOptions[]>;
         };

package/dist/nodes/BergetAi/BergetAi.node.js

@@ -4,7 +4,6 @@ exports.BergetAi = void 0;
 const n8n_workflow_1 = require("n8n-workflow");
 const n8n_workflow_2 = require("n8n-workflow");
 const chat_1 = require("./chat");
-const embeddings_1 = require("./embeddings");
 const ocr_1 = require("./ocr");
 const rerank_1 = require("./rerank");
 const speech_1 = require("./speech");
@@ -18,14 +17,15 @@ class BergetAi {
             group: ['transform'],
             version: 1,
             subtitle: '={{$parameter["resource"]}}',
-            description: 'Use Berget AI for chat completions, embeddings, document OCR, speech-to-text, and document reranking',
+            description: 'Use Berget AI for chat completions, document OCR, speech-to-text, and document reranking',
             defaults: { name: 'Berget AI' },
+            usableAsTool: true,
             codex: {
+                alias: ['Berget', 'Berget AI', 'Swedish AI', 'LLM', 'KB-Whisper'],
                 categories: ['AI'],
                 subcategories: {
-                    AI: ['Miscellaneous'],
+                    AI: ['Agents', 'Miscellaneous', 'Root Nodes'],
                 },
-                alias: ['Berget', 'Berget AI', 'Swedish AI', 'LLM'],
             },
             credentials: [
                 {
@@ -48,11 +48,6 @@ class BergetAi {
                             value: 'chat',
                             description: 'Create a chat completion',
                         },
-                        {
-                            name: 'Embeddings',
-                            value: 'embeddings',
-                            description: 'Generate vector embeddings from text',
-                        },
                         {
                             name: 'OCR',
                             value: 'ocr',
@@ -71,7 +66,6 @@ class BergetAi {
                     ],
                 },
                 ...chat_1.chatProperties,
-                ...embeddings_1.embeddingsProperties,
                 ...ocr_1.ocrProperties,
                 ...rerank_1.rerankProperties,
                 ...speech_1.speechProperties,
@@ -82,9 +76,6 @@ class BergetAi {
                 async getChatModels() {
                     return (0, shared_1.loadModelOptions)(this, (m) => m.model_type === 'text' || m.model_type === 'ocr');
                 },
-                async getEmbeddingsModels() {
-                    return (0, shared_1.loadModelOptions)(this, (m) => m.model_type === 'embedding');
-                },
                 async getRerankModels() {
                     return (0, shared_1.loadModelOptions)(this, (m) => m.model_type === 'rerank');
                 },
@@ -105,9 +96,6 @@ class BergetAi {
                     case 'chat':
                         result = await (0, chat_1.executeChat)(this, i);
                         break;
-                    case 'embeddings':
-                        result = await (0, embeddings_1.executeEmbeddings)(this, i);
-                        break;
                     case 'ocr':
                         result = await (0, ocr_1.executeOcr)(this, i);
                         break;

package/dist/nodes/BergetAiEmbeddingsModel/BergetAiEmbeddingsModel.node.d.ts

@@ -0,0 +1,10 @@
+import { type ILoadOptionsFunctions, type INodePropertyOptions, type INodeType, type INodeTypeDescription, type ISupplyDataFunctions, type SupplyData } from 'n8n-workflow';
+export declare class BergetAiEmbeddingsModel implements INodeType {
+    description: INodeTypeDescription;
+    methods: {
+        loadOptions: {
+            getEmbeddingsModels(this: ILoadOptionsFunctions): Promise<INodePropertyOptions[]>;
+        };
+    };
+    supplyData(this: ISupplyDataFunctions, itemIndex: number): Promise<SupplyData>;
+}

package/dist/nodes/BergetAiEmbeddingsModel/BergetAiEmbeddingsModel.node.js

@@ -0,0 +1,122 @@
+"use strict";
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.BergetAiEmbeddingsModel = void 0;
+const openai_1 = require("@langchain/openai");
+const axios_1 = __importDefault(require("axios"));
+const n8n_workflow_1 = require("n8n-workflow");
+const BERGET_API_BASE_URL = 'https://api.berget.ai/v1';
+class BergetAiEmbeddingsModel {
+    constructor() {
+        this.description = {
+            displayName: 'Berget AI Embeddings Model',
+            name: 'bergetAiEmbeddingsModel',
+            icon: 'file:bergetai.svg',
+            group: ['transform'],
+            version: 1,
+            description: 'Use a Berget AI embedding model with Vector Store, Question and Answer Chain, or other LangChain nodes',
+            defaults: { name: 'Berget AI Embeddings Model' },
+            codex: {
+                categories: ['AI'],
+                subcategories: {
+                    AI: ['Embeddings'],
+                },
+            },
+            credentials: [{ name: 'bergetAiApi', required: true }],
+            inputs: [],
+            outputs: [n8n_workflow_1.NodeConnectionTypes.AiEmbedding],
+            outputNames: ['Embeddings'],
+            properties: [
+                {
+                    displayName: 'This node must be connected to a Vector Store or QA Chain, not executed on its own.',
+                    name: 'notice',
+                    type: 'notice',
+                    default: '',
+                },
+                {
+                    displayName: 'Model',
+                    name: 'model',
+                    type: 'options',
+                    typeOptions: { loadOptionsMethod: 'getEmbeddingsModels' },
+                    default: '',
+                    required: true,
+                    description: 'The Berget AI embedding model to use. Fetched live from the Berget API.',
+                },
+                {
+                    displayName: 'Options',
+                    name: 'options',
+                    type: 'collection',
+                    placeholder: 'Add Option',
+                    default: {},
+                    options: [
+                        {
+                            displayName: 'Batch Size',
+                            name: 'batchSize',
+                            type: 'number',
+                            typeOptions: { minValue: 1 },
+                            default: 512,
+                            description: 'Number of documents to embed per API request',
+                        },
+                        {
+                            displayName: 'Strip New Lines',
+                            name: 'stripNewLines',
+                            type: 'boolean',
+                            default: true,
+                            description: 'Whether to strip newlines from input text before embedding',
+                        },
+                        {
+                            displayName: 'Timeout (Ms)',
+                            name: 'timeout',
+                            type: 'number',
+                            typeOptions: { minValue: 1 },
+                            default: 60000,
+                            description: 'Maximum time in milliseconds to wait for the embeddings API',
+                        },
+                    ],
+                },
+            ],
+        };
+        this.methods = {
+            loadOptions: {
+                async getEmbeddingsModels() {
+                    var _a, _b;
+                    const credentials = await this.getCredentials('bergetAiApi');
+                    const response = await axios_1.default.get(`${BERGET_API_BASE_URL}/models`, {
+                        headers: {
+                            Authorization: `Bearer ${credentials.apiKey}`,
+                            'Content-Type': 'application/json',
+                        },
+                    });
+                    const models = (_b = (_a = response.data) === null || _a === void 0 ? void 0 : _a.data) !== null && _b !== void 0 ? _b : [];
+                    return models
+                        .filter((m) => m.model_type === 'embedding')
+                        .map((m) => ({
+                        name: m.id,
+                        value: m.id,
+                    }))
+                        .sort((a, b) => a.name.localeCompare(b.name));
+                },
+            },
+        };
+    }
+    async supplyData(itemIndex) {
+        var _a, _b, _c;
+        const credentials = await this.getCredentials('bergetAiApi');
+        const modelName = this.getNodeParameter('model', itemIndex);
+        const options = this.getNodeParameter('options', itemIndex, {});
+        const embeddings = new openai_1.OpenAIEmbeddings({
+            apiKey: credentials.apiKey,
+            model: modelName,
+            configuration: {
+                baseURL: BERGET_API_BASE_URL,
+            },
+            batchSize: (_a = options.batchSize) !== null && _a !== void 0 ? _a : 512,
+            stripNewLines: (_b = options.stripNewLines) !== null && _b !== void 0 ? _b : true,
+            timeout: (_c = options.timeout) !== null && _c !== void 0 ? _c : 60000,
+        });
+        return { response: embeddings };
+    }
+}
+exports.BergetAiEmbeddingsModel = BergetAiEmbeddingsModel;

package/dist/nodes/BergetAiEmbeddingsModel/bergetai.svg

@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 463 419">
+  <path fill="#10b981" d="M208.739 17L255.261 17L446 403L398 403L313.5 255L261.5 176L233.163 96.1677L237.815 98.6522H226.185L230.837 96.1677L113 331L64.5 403L18 403L208.739 17Z"/>
+</svg>

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "n8n-nodes-berget-mk",
-  "version": "0.3.0",
-  "description": "n8n community node for Berget AI. One multi-resource action node (chat, embeddings, OCR, speech-to-text, rerank) plus a Chat Model sub-node that plugs into n8n's built-in AI Agent.",
+  "version": "0.4.0",
+  "description": "n8n community node for Berget AI. Multi-resource action node (chat, OCR, rerank, speech-to-text) plus Chat Model and Embeddings Model sub-nodes that plug into n8n's built-in AI Agent and Vector Store nodes.",
   "keywords": [
     "n8n-community-node-package",
     "n8n",
@@ -62,7 +62,8 @@
     ],
     "nodes": [
       "dist/nodes/BergetAi/BergetAi.node.js",
-      "dist/nodes/BergetAiChatModel/BergetAiChatModel.node.js"
+      "dist/nodes/BergetAiChatModel/BergetAiChatModel.node.js",
+      "dist/nodes/BergetAiEmbeddingsModel/BergetAiEmbeddingsModel.node.js"
     ]
   }
 }

package/dist/nodes/BergetAi/embeddings.d.ts

@@ -1,3 +0,0 @@
-import type { IDataObject, IExecuteFunctions, INodeProperties } from 'n8n-workflow';
-export declare const embeddingsProperties: INodeProperties[];
-export declare function executeEmbeddings(context: IExecuteFunctions, itemIndex: number): Promise<IDataObject>;

package/dist/nodes/BergetAi/embeddings.js

@@ -1,71 +0,0 @@
-"use strict";
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.embeddingsProperties = void 0;
-exports.executeEmbeddings = executeEmbeddings;
-const n8n_workflow_1 = require("n8n-workflow");
-const shared_1 = require("./shared");
-const showForEmbeddings = {
-    displayOptions: {
-        show: {
-            resource: ['embeddings'],
-        },
-    },
-};
-exports.embeddingsProperties = [
-    {
-        displayName: 'Model',
-        name: 'embeddingsModel',
-        type: 'options',
-        typeOptions: { loadOptionsMethod: 'getEmbeddingsModels' },
-        default: '',
-        required: true,
-        description: 'The Berget AI embedding model to use',
-        ...showForEmbeddings,
-    },
-    {
-        displayName: 'Input Text',
-        name: 'embeddingsInput',
-        type: 'string',
-        typeOptions: { rows: 4 },
-        default: '',
-        required: true,
-        description: 'Text to convert into an embedding vector',
-        ...showForEmbeddings,
-    },
-    {
-        displayName: 'Options',
-        name: 'embeddingsOptions',
-        type: 'collection',
-        placeholder: 'Add Option',
-        default: {},
-        options: [
-            {
-                displayName: 'Encoding Format',
-                name: 'encoding_format',
-                type: 'options',
-                options: [
-                    { name: 'Float', value: 'float' },
-                    { name: 'Base64', value: 'base64' },
-                ],
-                default: 'float',
-                description: 'Format of the returned embedding data',
-            },
-        ],
-        ...showForEmbeddings,
-    },
-];
-async function executeEmbeddings(context, itemIndex) {
-    var _a, _b;
-    const credentials = await context.getCredentials('bergetAiApi');
-    const model = context.getNodeParameter('embeddingsModel', itemIndex);
-    const input = context.getNodeParameter('embeddingsInput', itemIndex);
-    const options = context.getNodeParameter('embeddingsOptions', itemIndex, {});
-    const { status, data } = await (0, shared_1.bergetRequest)(credentials.apiKey, 'POST', '/embeddings', { model, input, ...options });
-    if (status !== 200) {
-        const message = (_b = (_a = data === null || data === void 0 ? void 0 : data.error) === null || _a === void 0 ? void 0 : _a.message) !== null && _b !== void 0 ? _b : `HTTP ${status}`;
-        throw new n8n_workflow_1.NodeOperationError(context.getNode(), `Berget AI embeddings error: ${message}`, {
-            itemIndex,
-        });
-    }
-    return data;
-}