@crewai-ts/rag

Memory, Knowledge, vector-store, and PDF text-extraction features for CrewAI TypeScript.

This package provides Memory and MemoryScope for cross-execution recall, Knowledge source types for prompt-time retrieval, embedding functions, and Qdrant / ChromaDB client adapters for building retrieval-augmented generation workflows. PDF text extraction is included by default; vector-store integration is opt-in.

Unofficial project. This is a community port of CrewAI and is not affiliated with, endorsed by, or maintained by crewAI, Inc. See License for the upstream MIT notice.

Install

npm install @crewai-ts/core @crewai-ts/rag
# or
pnpm add @crewai-ts/core @crewai-ts/rag

Requirements:

• Node.js 22 or later
• @crewai-ts/core 0.2.0 or later

Memory

Store and recall information across agent executions. Memory is attached to a Crew via the memory option; agent-level memory is also available through new Agent({ memory }).

import { Crew, Process, Task } from "@crewai-ts/core";
import { Memory } from "@crewai-ts/rag";

const memory = new Memory();
memory.remember("CrewAI supports sequential crews", {
  scope: "/research",
  categories: ["fact"],
});

const recalled = memory.recall("sequential crews", {
  scope: "/research",
  limit: 5,
});

const crew = new Crew({
  agents: [researcher],
  tasks: [task],
  process: Process.sequential,
  memory,
});

Knowledge

Inject knowledge sources into agent and crew prompts. Each Knowledge source contributes relevant snippets to the task prompt as additional context.

StringKnowledgeSource accepts either a plain string or an options object. The string form is the most ergonomic for inline content:

import { Agent, Crew, Task } from "@crewai-ts/core";
import {
  CSVKnowledgeSource,
  JSONKnowledgeSource,
  StringKnowledgeSource,
  TextFileKnowledgeSource,
} from "@crewai-ts/rag";

const crew = new Crew({
  agents: [researcher],
  tasks: [
    new Task({
      description: "Explain the Nest integration",
      expectedOutput: "Integration guidance",
      agent: researcher,
    }),
  ],
  knowledgeSources: [
    new StringKnowledgeSource("Nest should consume crewai-ts as a normal TypeScript library."),
    new StringKnowledgeSource({
      content: "The API service uses NestJS modules, providers, and controllers.",
      metadata: { source: "architecture-notes" },
    }),
    new TextFileKnowledgeSource("docs/notes.txt"),
    new JSONKnowledgeSource("data/facts.json"),
    new CSVKnowledgeSource("data/records.csv"),
  ],
});

crew.resetMemories("knowledge");

Vector Stores

This package ships client adapters for Qdrant and ChromaDB. The clients wrap the existing EmbeddingFunction plumbing and let you perform collection operations, batching, and locked writes against a remote store.

import { ChromaDBClient, QdrantClient } from "@crewai-ts/rag";

const qdrant = new QdrantClient(existingQdrantClient, {
  embeddingFunction: myEmbeddingFunction,
  defaultLimit: 10,
  defaultScoreThreshold: 0.7,
  defaultBatchSize: 100,
  lockName: "qdrant-research",
});

const chroma = new ChromaDBClient(existingChromaClient, {
  embeddingFunction: myEmbeddingFunction,
  defaultLimit: 5,
  defaultScoreThreshold: 0.6,
  defaultBatchSize: 100,
});

QdrantClient and ChromaDBClient are thin async-friendly adapters — operations like acreate_collection, aadd, aquery, and adelete return Promises and integrate with the runNamedLock helper for process-safe writes. They expect you to bring your own @qdrant/js-client-rest or chromadb client (this package does not bundle a transport).

PDF Text Extraction

defaultPDFTextExtractorAsync extracts plain text from a PDF buffer. It uses pdf-parse internally and accepts a Buffer:

import { readFileSync } from "node:fs";
import { defaultPDFTextExtractorAsync } from "@crewai-ts/rag";

const pdfBuffer = readFileSync("docs/report.pdf");
const text = await defaultPDFTextExtractorAsync(pdfBuffer);

If you have a Uint8Array instead, wrap it with Buffer.from(bytes) first.

Exports

• Memory, MemoryScope, MemorySlice — memory storage and recall
• createMemoryTools(memory) — recall / save tools injected into crews
• Knowledge, StringKnowledgeSource, TextFileKnowledgeSource, JSONKnowledgeSource, CSVKnowledgeSource — knowledge sources
• QdrantClient, ChromaDBClient — vector-store client adapters
• QdrantEmbeddingFunctionWrapper, ChromaEmbeddingFunctionWrapper — embedding function adapters
• BaseRecord, Embedding, Embeddings, EmbeddingFunction, AsyncEmbeddingFunction, QueryEmbedding — vector store types
• defaultRagEmbeddingFunction — the default embedding function
• defaultPDFTextExtractorAsync — PDF text extraction

Related Packages

• @crewai-ts/core — agents, tasks, crews, tools, hooks, security, checkpoints
• @crewai-ts/flow — stateful Flow orchestration
• @crewai-ts/nestjs — NestJS DI integration
• @crewai-ts/cli — crewai-ts CLI

License

MIT

This project is an unofficial TypeScript port of CrewAI (Copyright © crewAI, Inc.), which is distributed under the MIT License. It is not affiliated with or endorsed by crewAI, Inc. As required by the MIT License, the original copyright and permission notice are retained in LICENSE.