@byted-las/contextlake-openclaw 1.0.2 → 1.0.4

package/README.md

@@ -1,27 +1,32 @@
 # ContextLake OpenClaw Plugin
 
-A powerful, serverless-first Knowledge Base / Knowledge Lake (知识库/知识湖) plugin for the OpenClaw Agent framework.
+A powerful, multi-modal Knowledge Base / Knowledge Lake (知识库/知识湖) plugin for the OpenClaw Agent framework.
 
-This plugin allows OpenClaw agents to natively understand, index, and retrieve local and remote documents (PDF, Word, Markdown, Text) using vector similarity search. It features a pluggable architecture supporting both completely offline operations (via LanceDB and local GGUF models) and cloud-based operations.
+This plugin allows OpenClaw agents to natively understand, profile, index, and retrieve data from diverse data sources (Local, TOS, S3, etc.). It supports advanced multi-modal document ingestion (Text, PDF, Images, Audio, Video) powered by Large AI Services (LAS) and vector similarity search.
 
 ## Features
 
-- **Multi-modal Document Ingestion**: Automatically extracts text from PDF, DOCX, TXT, and MD files.
-- **Agentic Tools**: Exposes `contextlake-ingest`, `contextlake-retrieve`, `contextlake-list`, and `contextlake-delete` skills directly to the LLM.
-- **Slash Commands**: Quick chat actions (`/contextlake-search`, `/contextlake-ingest`) for fast human-in-the-loop operations.
-- **CLI Management**: Full command-line interface for bulk operations and CI/CD integration.
-- **Local-First & Pluggable**: 
-  - Uses `LanceDB` for local embedded vector storage.
-  - Uses `node-llama-cpp` for offline, private embedding generation (`bge-small-zh` by default).
-  - Can be easily configured to use remote embedding APIs (like OpenAI or Doubao ARK) or remote storage (TOS).
+- **Data Profiling**: Connect to heterogeneous data sources (TOS/OSS/COS/S3/Local) and auto-profile schemas and media metadata.
+- **Multi-modal Ingestion**: Automatically extracts text, converts PDFs, parses audio (ASR), understands video, and embeds images directly using LAS multi-modal models.
+- **Agentic Tools**: Exposes `contextlake-ingest`, `contextlake-retrieve`, `contextlake-list`, `contextlake-delete`, `las-data-profiler`, and `contextlake-list-datasource` skills directly to the LLM.
+- **Slash Commands**: Quick chat actions (`/contextlake-search`, `/contextlake-ingest`, `/contextlake-profiler`) for fast human-in-the-loop operations.
+- **CLI Management**: Full command-line interface for bulk operations and pipeline integration.
+- **Pluggable Architecture**: 
+  - Uses `LanceDB` for local embedded multi-modal vector storage.
+  - Pluggable storage providers (Local and TOS).
+
+## Documentation
+
+For a comprehensive guide on installation, initialization (`onboard`), connecting data sources (`connect`/`profiler`), ingestion, and searching, please refer to the [**USAGE.md**](./USAGE.md).
+
+For detailed architecture and internal workflows, please refer to [AGENTS.md](./AGENTS.md).
 
 ## Installation
 
 Assuming you are in an OpenClaw environment:
 
 ```bash
-npm install
-npm run build
+npm install @byted-las/contextlake-openclaw
 ```
 
 Then register the plugin in your OpenClaw configuration:
@@ -30,7 +35,7 @@ Then register the plugin in your OpenClaw configuration:
 {
   "plugins": [
     {
-      "package": "contextlake-openclaw",
+      "package": "@byted-las/contextlake-openclaw",
       "config": {
         "metadata_storage": {
           "type": "local",
@@ -46,19 +51,36 @@ Then register the plugin in your OpenClaw configuration:
 }
 ```
 
+## Quick Start (Slash Commands)
+
+1. **Initialize Credentials** (Run in terminal):
+   ```bash
+   openclaw contextlake onboard
+   ```
+2. **Profile a Data Source**:
+   ```text
+   /contextlake-profiler my_data local /path/to/my/files .
+   ```
+3. **Ingest the Data Source** (Processes PDF, Audio, Images, Text):
+   ```text
+   /contextlake-ingest my_data
+   ```
+4. **Search your Knowledge Base**:
+   ```text
+   /contextlake-search How do I configure the API?
+   ```
+
 ## Agent Tools (Skills)
 
 The LLM agent automatically has access to the following tools:
 
-1. **`contextlake-ingest`**: Uploads and indexes files.
-2. **`contextlake-retrieve`**: Performs vector search on the indexed documents based on a query.
-3. **`contextlake-list`**: Lists all currently indexed documents.
-4. **`contextlake-delete`**: Removes documents by ID or filter.
+1. **`las-data-profiler`**: Connect to a data source and profile its structure.
+2. **`contextlake-list-datasource`**: List all connected and profiled data sources.
+3. **`contextlake-ingest`**: Process and ingest all profiled files into the knowledge base via multi-modal embeddings.
+4. **`contextlake-retrieve`**: Performs vector search on the indexed documents based on a query.
+5. **`contextlake-list`**: Lists all currently indexed documents.
+6. **`contextlake-delete`**: Removes documents by ID or filter.
 
 You can simply talk to the agent:
-> "帮我把这个 `manual.pdf` 注入到知识库中。"
-> "在知识湖中检索关于架构设计的文档。"
-
-## Development
-
-For detailed architecture and internal workflows, please refer to [AGENTS.md](./AGENTS.md).
+> "帮我把 `my_data` 数据源注入到知识库中。"
+> "在知识湖中检索关于架构设计的文档。"

package/dist/src/client/lancedb.js

@@ -97,7 +97,7 @@ class ContextLakeLanceDBClient {
             }
             return await fallbackQuery.toArray();
         }
-        const vector = await this.embeddingProvider.generateEmbedding(query);
+        const vector = await this.embeddingProvider.generateMultimodalEmbedding([{ type: 'text', text: query }]);
         // @ts-ignore
         let search = table.vectorSearch(vector).limit(normalizedLimit);
         if (filter) {

package/dist/src/commands/cli.d.ts

@@ -1,8 +1,9 @@
 import { ContextLakeConfig } from '../utils/config';
 export declare function getCliCommands(pluginConfig: ContextLakeConfig, logger: any): {
     connectAction: (datasource_name: string, options: any) => Promise<void>;
-    ingestAction: (files: any, options: any) => Promise<void>;
+    ingestAction: (datasource_name: string) => Promise<void>;
     searchAction: (query: any, options: any) => Promise<void>;
     listAction: (options: any) => Promise<void>;
     deleteAction: (options: any) => Promise<void>;
+    onboardAction: () => Promise<void>;
 };

package/dist/src/commands/cli.js

@@ -6,6 +6,7 @@ const ingest_1 = require("../lib/actions/ingest");
 const retrieve_1 = require("../lib/actions/retrieve");
 const manage_1 = require("../lib/actions/manage");
 const profiler_1 = require("../lib/actions/profiler");
+const credentials_1 = require("../utils/credentials");
 function parseOptionalInt(value, fallback) {
     const parsed = Number.parseInt(String(value), 10);
     return Number.isFinite(parsed) ? parsed : fallback;
@@ -67,15 +68,11 @@ function getCliCommands(pluginConfig, logger) {
                 process.exitCode = 1;
             }
         },
-        ingestAction: async (files, options) => {
-            logger.info(`[${new Date().toISOString()}] [ContextLake] CLI ingest started`, { files, options });
+        ingestAction: async (datasource_name) => {
+            logger.info(`[${new Date().toISOString()}] [ContextLake] CLI ingest started`, { datasource_name });
             try {
-                const metadata = parseMetadata(options.metadata);
-                const result = await (0, ingest_1.ingestAssets)({
-                    files,
-                    metadata,
-                    chunkSize: parseOptionalInt(options.chunkSize, 500),
-                    overlap: parseOptionalInt(options.overlap, 50)
+                const result = await (0, ingest_1.ingestSource)({
+                    datasource_name
                 }, pluginConfig, logger);
                 // eslint-disable-next-line no-console
                 console.log(JSON.stringify(result, null, 2));
@@ -134,6 +131,34 @@ function getCliCommands(pluginConfig, logger) {
                 console.error('Error:', e.message);
                 logger.error(`[${new Date().toISOString()}] [ContextLake] CLI delete failed`, { error: e.message, stack: e.stack });
             }
+        },
+        onboardAction: async () => {
+            logger.info(`[${new Date().toISOString()}] [ContextLake] CLI onboard started`);
+            try {
+                const currentCreds = (0, credentials_1.loadCredentials)();
+                // eslint-disable-next-line no-console
+                console.log('Welcome to ContextLake Onboarding!');
+                // eslint-disable-next-line no-console
+                console.log('Please provide your credentials below. Press enter to keep the current value.');
+                const lasApiKey = await (0, credentials_1.promptForInput)('LAS_API_KEY', currentCreds.LAS_API_KEY);
+                const accessKey = await (0, credentials_1.promptForInput)('ACCESS_KEY', currentCreds.ACCESS_KEY || currentCreds.VOLCENGINE_ACCESS_KEY);
+                const secretKey = await (0, credentials_1.promptForInput)('SECRET_KEY', currentCreds.SECRET_KEY || currentCreds.VOLCENGINE_SECRET_KEY);
+                const region = await (0, credentials_1.promptForInput)('REGION', currentCreds.REGION || currentCreds.VOLCENGINE_REGION || 'cn-beijing');
+                const newCreds = {
+                    LAS_API_KEY: lasApiKey,
+                    ACCESS_KEY: accessKey,
+                    SECRET_KEY: secretKey,
+                    REGION: region
+                };
+                (0, credentials_1.saveCredentials)(newCreds);
+                // eslint-disable-next-line no-console
+                console.log('Credentials saved successfully!');
+                logger.info(`[${new Date().toISOString()}] [ContextLake] CLI onboard success`);
+            }
+            catch (e) {
+                console.error('Error during onboarding:', e.message);
+                logger.error(`[${new Date().toISOString()}] [ContextLake] CLI onboard failed`, { error: e.message, stack: e.stack });
+            }
         }
     };
 }

package/dist/src/commands/index.js

@@ -20,6 +20,18 @@ function registerAll(ctx, logger) {
         logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${tools.deleteTool.name}`);
         ctx.registerTool(tools.lasDataProfilerTool);
         logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${tools.lasDataProfilerTool.name}`);
+        ctx.registerTool(tools.listS3ObjectsTool);
+        logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${tools.listS3ObjectsTool.name}`);
+        ctx.registerTool(tools.readS3ObjectTool);
+        logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${tools.readS3ObjectTool.name}`);
+        ctx.registerTool(tools.writeLanceCatalogTool);
+        logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${tools.writeLanceCatalogTool.name}`);
+        ctx.registerTool(tools.listDatasourceTool);
+        logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${tools.listDatasourceTool.name}`);
+        for (const lasTool of tools.lasTools) {
+            ctx.registerTool(lasTool);
+            logger.info(`[${new Date().toISOString()}] [ContextLake] Tool registered: ${lasTool.name}`);
+        }
     }
     catch (error) {
         logger.error(`[${new Date().toISOString()}] [ContextLake] Error registering agent tools: ${error.message}${error.stack ? '\\n' + error.stack : ''}`);
@@ -46,11 +58,8 @@ function registerAll(ctx, logger) {
                 .option('--sample-rows <number>', 'Number of rows to sample per structured file', '100')
                 .action(commands.connectAction);
             // Ingest
-            contextlake.command('ingest <files...>')
-                .description('Ingest one or more files into the knowledge base')
-                .option('-c, --chunk-size <number>', 'Chunk size for text splitting', '500')
-                .option('-o, --overlap <number>', 'Chunk overlap size', '50')
-                .option('-m, --metadata <json>', 'JSON metadata to attach to the documents')
+            contextlake.command('ingest <datasource_name>')
+                .description('Process and ingest all files from a connected data source into the knowledge base')
                 .action(commands.ingestAction);
             // Search
             contextlake.command('search <query>')
@@ -70,6 +79,10 @@ function registerAll(ctx, logger) {
                 .option('--ids <ids...>', 'List of specific file IDs to delete')
                 .option('-f, --filter <string>', 'Filter string to match documents for deletion')
                 .action(commands.deleteAction);
+            // Onboard
+            contextlake.command('onboard')
+                .description('Configure credentials for ContextLake')
+                .action(commands.onboardAction);
         }, { commands: ['contextlake'] });
         logger.info(`[${new Date().toISOString()}] [ContextLake] CLI commands registered`);
     }
@@ -86,7 +99,7 @@ function registerAll(ctx, logger) {
         const slashCommands = (0, slashcmd_1.getSlashCommands)(pluginConfig, logger);
         ctx.registerCommand({
             name: 'contextlake-ingest',
-            description: 'Ingest files into the knowledge base (usage: /contextlake-ingest file1 file2)',
+            description: 'Process and ingest all files from a connected data source (usage: /contextlake-ingest <datasource_name>)',
             acceptsArgs: true,
             handler: slashCommands.ingestHandler
         });
@@ -108,6 +121,18 @@ function registerAll(ctx, logger) {
             acceptsArgs: true,
             handler: slashCommands.deleteHandler
         });
+        ctx.registerCommand({
+            name: 'contextlake-profiler',
+            description: 'Connect to a data source and profile its structure (usage: /contextlake-profiler <datasource_name> <vendor> <bucket> <prefix>)',
+            acceptsArgs: true,
+            handler: slashCommands.profilerHandler
+        });
+        ctx.registerCommand({
+            name: 'contextlake-list-datasource',
+            description: 'List all connected and profiled data sources (usage: /contextlake-list-datasource)',
+            acceptsArgs: false,
+            handler: slashCommands.listDatasourceHandler
+        });
         logger.info(`[${new Date().toISOString()}] [ContextLake] Slash commands registered`);
     }
     catch (error) {

package/dist/src/commands/slashcmd.d.ts

@@ -12,4 +12,10 @@ export declare function getSlashCommands(pluginConfig: ContextLakeConfig, logger
     deleteHandler: (commandCtx: any) => Promise<{
         text: string;
     }>;
+    profilerHandler: (commandCtx: any) => Promise<{
+        text: string;
+    }>;
+    listDatasourceHandler: (commandCtx: any) => Promise<{
+        text: string;
+    }>;
 };

package/dist/src/commands/slashcmd.js

@@ -1,9 +1,46 @@
 "use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.getSlashCommands = getSlashCommands;
 const ingest_1 = require("../lib/actions/ingest");
 const retrieve_1 = require("../lib/actions/retrieve");
 const manage_1 = require("../lib/actions/manage");
+const profiler_1 = require("../lib/actions/profiler");
+const fs = __importStar(require("fs"));
+const path = __importStar(require("path"));
+const os = __importStar(require("os"));
 function getSlashCommands(pluginConfig, logger) {
     return {
         ingestHandler: async (commandCtx) => {
@@ -12,12 +49,16 @@ function getSlashCommands(pluginConfig, logger) {
             logger.info(`[${new Date().toISOString()}] [ContextLake] Slash command ingest started`, { args });
             try {
                 if (args.length === 0) {
-                    return { text: `**Error:** Missing files. Usage: /contextlake-ingest /path/to/file1 /path/to/file2` };
+                    return { text: `**Error:** Missing datasource_name. Usage: /contextlake-ingest <datasource_name>` };
                 }
-                const result = await (0, ingest_1.ingestAssets)({
-                    files: args,
-                    metadata: {}
-                }, pluginConfig, logger);
+                const datasource_name = args[0];
+                const BASE_DIR = path.join(os.homedir(), '.openclaw', 'contextlake', 'profiler');
+                const dsDir = path.join(BASE_DIR, datasource_name);
+                const dbPath = path.join(dsDir, 'catalog_db');
+                if (!fs.existsSync(dbPath)) {
+                    return { text: `**Error:** Data source "${datasource_name}" has not been profiled yet.\n\nPlease run the profiler first using:\n\`/contextlake-profiler <datasource_name> <vendor> <bucket> <prefix> [endpoint] [ak] [sk] [region]\`` };
+                }
+                const result = await (0, ingest_1.ingestSource)({ datasource_name }, pluginConfig, logger);
                 logger.info(`[${new Date().toISOString()}] [ContextLake] Slash command ingest completed`, { resultCount: result.length });
                 return { text: `**Ingest Results (${result.length} files processed):**\n\`\`\`json\n${JSON.stringify(result, null, 2)}\n\`\`\`` };
             }
@@ -86,6 +127,49 @@ function getSlashCommands(pluginConfig, logger) {
                 logger.error(`[ContextLake] Slash delete failed`, { error: e.message });
                 return { text: `**Error executing delete:** ${e.message}` };
             }
-        }
+        },
+        profilerHandler: async (commandCtx) => {
+            const rawArgs = commandCtx.args || "";
+            const args = rawArgs.split(' ').filter((arg) => arg.trim() !== '');
+            logger.info(`[${new Date().toISOString()}] [ContextLake] Slash command profiler started`, { args });
+            try {
+                if (args.length < 4) {
+                    return { text: `**Error:** Missing arguments. Usage: /contextlake-profiler <datasource_name> <vendor> <bucket> <prefix> [endpoint] [ak] [sk] [region]` };
+                }
+                const [datasource_name, vendor, bucket, prefix, endpoint, access_key, secret_key, region] = args;
+                if (!['volcengine', 'alibaba', 'tencent', 'aws', 'local'].includes(vendor)) {
+                    return { text: `**Error:** Invalid vendor. Must be one of: volcengine, alibaba, tencent, aws, local` };
+                }
+                const params = {
+                    datasource_name,
+                    vendor: vendor,
+                    bucket,
+                    prefix,
+                    endpoint,
+                    access_key,
+                    secret_key,
+                    region,
+                };
+                const result = await (0, profiler_1.connectDataSource)(params);
+                logger.info(`[${new Date().toISOString()}] [ContextLake] Slash command profiler completed`, { result });
+                return { text: `**Profiler Results:**\n\`\`\`json\n${JSON.stringify(result, null, 2)}\n\`\`\`` };
+            }
+            catch (e) {
+                logger.error(`[ContextLake] Slash profiler failed`, { error: e.message });
+                return { text: `**Error executing profiler:** ${e.message}` };
+            }
+        },
+        listDatasourceHandler: async (commandCtx) => {
+            logger.info(`[${new Date().toISOString()}] [ContextLake] Slash command list-datasource started`);
+            try {
+                const result = await (0, profiler_1.listDataSources)();
+                logger.info(`[${new Date().toISOString()}] [ContextLake] Slash command list-datasource completed`, { result });
+                return { text: `**Data Sources:**\n\`\`\`json\n${JSON.stringify(result, null, 2)}\n\`\`\`` };
+            }
+            catch (e) {
+                logger.error(`[ContextLake] Slash list-datasource failed`, { error: e.message });
+                return { text: `**Error executing list-datasource:** ${e.message}` };
+            }
+        },
     };
 }

package/dist/src/commands/tools.d.ts

@@ -6,4 +6,9 @@ export declare function getAgentTools(pluginConfig: ContextLakeConfig, logger: a
     listTool: AnyAgentTool;
     deleteTool: AnyAgentTool;
     lasDataProfilerTool: AnyAgentTool;
+    listDatasourceTool: AnyAgentTool;
+    listS3ObjectsTool: AnyAgentTool;
+    readS3ObjectTool: AnyAgentTool;
+    writeLanceCatalogTool: AnyAgentTool;
+    lasTools: AnyAgentTool[];
 };

package/dist/src/commands/tools.js

@@ -5,44 +5,54 @@ const ingest_1 = require("../lib/actions/ingest");
 const retrieve_1 = require("../lib/actions/retrieve");
 const manage_1 = require("../lib/actions/manage");
 const profiler_1 = require("../lib/actions/profiler");
+const las_tools_1 = require("../lib/actions/las-tools");
+const s3_tools_1 = require("../lib/actions/s3-tools");
+const lance_tools_1 = require("../lib/actions/lance-tools");
 function getAgentTools(pluginConfig, logger) {
+    const lasTools = (0, las_tools_1.getLasTools)(pluginConfig, logger);
     return {
+        lasTools,
+        listDatasourceTool: {
+            name: 'contextlake-list-datasource',
+            label: 'ContextLake List Datasources',
+            description: `List all connected and profiled data sources.`,
+            parameters: {
+                type: 'object',
+                properties: {},
+                required: [],
+                additionalProperties: false
+            },
+            async execute(toolCallId, params) {
+                logger.info(`[${new Date().toISOString()}] [ContextLake] Executing list-datasource skill, toolCallId: ${toolCallId}`);
+                try {
+                    const result = await (0, profiler_1.listDataSources)();
+                    return {
+                        content: [{ type: "text", text: JSON.stringify(result) }],
+                        details: result
+                    };
+                }
+                catch (error) {
+                    logger.error(`[${new Date().toISOString()}] [ContextLake] list-datasource skill failed`, { error: error.message });
+                    return {
+                        content: [{ type: "text", text: String(error.message) }],
+                        details: { error: error.message }
+                    };
+                }
+            }
+        },
         ingestTool: {
             name: 'contextlake-ingest',
             label: 'ContextLake Ingest',
-            description: `Upload, ingest, and index documents into the ContextLake Knowledge Base (知识库) / Knowledge Lake (知识湖).
+            description: `Process and ingest all files from a connected data source into the knowledge base.
 Use this tool when the user wants to "将知识注入", "上传文件", "入库", "添加文档", "ingest files", or "add knowledge".
-Supports processing of various file types including PDF, Word, Markdown, and Text.
-Automatically handles text extraction, cleaning, chunking, embedding generation, and storage.
-
-Example User Queries:
-- "帮我把这个文档注入到知识湖中"
-- "上传这份 PDF 到知识库"
-- "Please ingest these documents into ContextLake"
-- "将 /path/to/doc.txt 添加到知识库"`,
+Supports multimodal files (text, images, audio, video, pdf) by using LAS models to understand and embed them.
+Must be called after a data source has been successfully profiled via \`las-data-profiler\`.`,
             parameters: {
                 type: 'object',
                 properties: {
-                    files: {
-                        type: 'array',
-                        items: { type: 'string' },
-                        description: 'List of file paths to ingest'
-                    },
-                    metadata: {
-                        type: 'object',
-                        description: 'Optional JSON metadata to attach to documents',
-                        additionalProperties: true
-                    },
-                    chunkSize: {
-                        type: 'integer',
-                        description: 'Chunk size for text splitting'
-                    },
-                    overlap: {
-                        type: 'integer',
-                        description: 'Overlap size for text splitting'
-                    }
+                    datasource_name: { type: 'string', description: 'Name of the data source previously profiled' }
                 },
-                required: ['files'],
+                required: ['datasource_name'],
                 additionalProperties: false
             },
             async execute(toolCallId, params) {
@@ -56,21 +66,21 @@ Example User Queries:
                         catch (e) {
                             logger.warn(`[ContextLake] Received string params, possibly toolCallId?`, { params });
                             return {
-                                content: [{ type: "text", text: `Invalid params format: received string "${params}", expected object with 'files' array.` }],
-                                details: { error: `Invalid params format: received string "${params}", expected object with 'files' array.` }
+                                content: [{ type: "text", text: `Invalid params format: received string "${params}", expected object with 'datasource_name'.` }],
+                                details: { error: `Invalid params format: received string "${params}", expected object with 'datasource_name'.` }
                             };
                         }
                     }
-                    if (!actualParams.files && actualParams.params && actualParams.params.files) {
+                    if (!actualParams.datasource_name && actualParams.params && actualParams.params.datasource_name) {
                         actualParams = actualParams.params;
                     }
-                    if (!actualParams.files || !Array.isArray(actualParams.files)) {
+                    if (!actualParams.datasource_name) {
                         return {
-                            content: [{ type: "text", text: `Invalid params: 'files' must be an array. Received keys: ${Object.keys(actualParams)}` }],
-                            details: { error: `Invalid params: 'files' must be an array. Received keys: ${Object.keys(actualParams)}` }
+                            content: [{ type: "text", text: `Invalid params: 'datasource_name' is required. Received keys: ${Object.keys(actualParams)}` }],
+                            details: { error: `Invalid params: 'datasource_name' is required. Received keys: ${Object.keys(actualParams)}` }
                         };
                     }
-                    const result = await (0, ingest_1.ingestAssets)(actualParams, pluginConfig, logger);
+                    const result = await (0, ingest_1.ingestSource)(actualParams, pluginConfig, logger);
                     logger.info(`[${new Date().toISOString()}] [ContextLake] Ingest skill completed successfully`, { resultSummary: Array.isArray(result) ? `Processed ${result.length} items` : 'Success' });
                     return {
                         content: [{ type: "text", text: JSON.stringify(result) }],
@@ -81,8 +91,7 @@ Example User Queries:
                     logger.error(`[${new Date().toISOString()}] [ContextLake] Ingest skill failed`, { error: error.message, stack: error.stack });
                     return {
                         content: [{ type: "text", text: String(error.message) }],
-                        details: { error: error.message
-                        }
+                        details: { error: error.message }
                     };
                 }
             }
@@ -274,11 +283,97 @@ Example User Queries:
                     logger.error(`[${new Date().toISOString()}] [ContextLake] las-data-profiler skill failed`, { error: error.message, stack: error.stack });
                     return {
                         content: [{ type: "text", text: String(error.message) }],
-                        details: { error: error.message
-                        }
+                        details: { error: error.message }
                     };
                 }
             }
+        },
+        listS3ObjectsTool: {
+            name: 'list-s3-objects',
+            label: 'List S3 Objects',
+            description: 'List objects in an S3-compatible bucket or local directory',
+            parameters: {
+                type: 'object',
+                properties: {
+                    vendor: { type: 'string', enum: ['volcengine', 'alibaba', 'tencent', 'aws', 'local'] },
+                    bucket: { type: 'string' },
+                    prefix: { type: 'string' },
+                    endpoint: { type: 'string' },
+                    access_key: { type: 'string' },
+                    secret_key: { type: 'string' },
+                    region: { type: 'string' },
+                    maxKeys: { type: 'integer' },
+                    continuationToken: { type: 'string' }
+                },
+                required: ['vendor', 'bucket'],
+                additionalProperties: false
+            },
+            async execute(toolCallId, params) {
+                let actualParams = params.params || params;
+                try {
+                    const result = await (0, s3_tools_1.listS3Objects)(actualParams, actualParams.prefix || '', actualParams.maxKeys, actualParams.continuationToken);
+                    return { content: [{ type: "text", text: JSON.stringify(result) }], details: result };
+                }
+                catch (e) {
+                    return { content: [{ type: "text", text: String(e.message) }], details: { error: e.message } };
+                }
+            }
+        },
+        readS3ObjectTool: {
+            name: 'read-s3-object',
+            label: 'Read S3 Object',
+            description: 'Read the contents or headers of an S3 object',
+            parameters: {
+                type: 'object',
+                properties: {
+                    vendor: { type: 'string', enum: ['volcengine', 'alibaba', 'tencent', 'aws', 'local'] },
+                    bucket: { type: 'string' },
+                    key: { type: 'string' },
+                    endpoint: { type: 'string' },
+                    access_key: { type: 'string' },
+                    secret_key: { type: 'string' },
+                    region: { type: 'string' },
+                    maxBytes: { type: 'integer' }
+                },
+                required: ['vendor', 'bucket', 'key'],
+                additionalProperties: false
+            },
+            async execute(toolCallId, params) {
+                let actualParams = params.params || params;
+                try {
+                    const buf = await (0, s3_tools_1.readS3Object)(actualParams, actualParams.key, actualParams.maxBytes);
+                    // Return as base64 string
+                    return { content: [{ type: "text", text: buf.toString('base64') }], details: { length: buf.length } };
+                }
+                catch (e) {
+                    return { content: [{ type: "text", text: String(e.message) }], details: { error: e.message } };
+                }
+            }
+        },
+        writeLanceCatalogTool: {
+            name: 'write-lance-catalog',
+            label: 'Write LanceDB Catalog',
+            description: 'Write an array of file records into a local LanceDB table',
+            parameters: {
+                type: 'object',
+                properties: {
+                    db_path: { type: 'string' },
+                    table_name: { type: 'string' },
+                    records: { type: 'array', items: { type: 'object' } }
+                },
+                required: ['db_path', 'table_name', 'records'],
+                additionalProperties: false
+            },
+            async execute(toolCallId, params) {
+                let actualParams = params.params || params;
+                try {
+                    await (0, lance_tools_1.writeLanceCatalog)(actualParams);
+                    return { content: [{ type: "text", text: "Successfully wrote records to LanceDB" }], details: { count: actualParams.records.length } };
+                }
+                catch (e) {
+                    return { content: [{ type: "text", text: String(e.message) }], details: { error: e.message } };
+                }
+            }
         }
     };
 }

package/dist/src/lib/actions/ingest-source.d.ts

@@ -0,0 +1,15 @@
+import { ContextLakeConfig } from '../../utils/config';
+export interface IngestSourceParams {
+    datasource_name: string;
+}
+export declare function ingestSource(params: IngestSourceParams, config: ContextLakeConfig, logger?: any): Promise<({
+    file: any;
+    status: string;
+    chunks: number;
+    message?: undefined;
+} | {
+    file: any;
+    status: string;
+    message: any;
+    chunks?: undefined;
+})[]>;