anything-llm/server/utils/vectorDbProviders/weaviate/index.js

const { default: weaviate } = require("weaviate-ts-client");
const { RecursiveCharacterTextSplitter } = require("langchain/text_splitter");
const { storeVectorResult, cachedVectorInformation } = require("../../files");
const { v4: uuidv4 } = require("uuid");
const { toChunks, getLLMProvider } = require("../../helpers");
const { chatPrompt } = require("../../chats");
const { camelCase } = require("../../helpers/camelcase");

const Weaviate = {
  name: "Weaviate",
  connect: async function () {
    if (process.env.VECTOR_DB !== "weaviate")
      throw new Error("Weaviate::Invalid ENV settings");

    const weaviateUrl = new URL(process.env.WEAVIATE_ENDPOINT);
    const options = {
      scheme: weaviateUrl.protocol?.replace(":", "") || "http",
      host: weaviateUrl?.host,
      ...(process.env?.WEAVIATE_API_KEY?.length > 0
        ? { apiKey: new weaviate.ApiKey(process.env?.WEAVIATE_API_KEY) }
        : {}),
    };
    const client = weaviate.client(options);
    const isAlive = await await client.misc.liveChecker().do();
    if (!isAlive)
      throw new Error(
        "Weaviate::Invalid Alive signal received - is the service online?"
      );
    return { client };
  },
  heartbeat: async function () {
    await this.connect();
    return { heartbeat: Number(new Date()) };
  },
  totalVectors: async function () {
    const { client } = await this.connect();
    const collectionNames = await this.allNamespaces(client);
    var totalVectors = 0;
    for (const name of collectionNames) {
      totalVectors += await this.namespaceCountWithClient(client, name);
    }
    return totalVectors;
  },
  namespaceCountWithClient: async function (client, namespace) {
    try {
      const response = await client.graphql
        .aggregate()
        .withClassName(camelCase(namespace))
        .withFields("meta { count }")
        .do();
      return (
        response?.data?.Aggregate?.[camelCase(namespace)]?.[0]?.meta?.count || 0
      );
    } catch (e) {
      console.error(`Weaviate:namespaceCountWithClient`, e.message);
      return 0;
    }
  },
  namespaceCount: async function (namespace = null) {
    try {
      const { client } = await this.connect();
      const response = await client.graphql
        .aggregate()
        .withClassName(camelCase(namespace))
        .withFields("meta { count }")
        .do();

      return (
        response?.data?.Aggregate?.[camelCase(namespace)]?.[0]?.meta?.count || 0
      );
    } catch (e) {
      console.error(`Weaviate:namespaceCountWithClient`, e.message);
      return 0;
    }
  },
  similarityResponse: async function (client, namespace, queryVector) {
    const result = {
      contextTexts: [],
      sourceDocuments: [],
    };

    const weaviateClass = await this.namespace(client, namespace);
    const fields = weaviateClass.properties.map((prop) => prop.name).join(" ");
    const queryResponse = await client.graphql
      .get()
      .withClassName(camelCase(namespace))
      .withFields(`${fields} _additional { id }`)
      .withNearVector({ vector: queryVector })
      .withLimit(4)
      .do();

    const responses = queryResponse?.data?.Get?.[camelCase(namespace)];
    responses.forEach((response) => {
      // In Weaviate we have to pluck id from _additional and spread it into the rest
      // of the properties.
      const {
        _additional: { id },
        ...rest
      } = response;
      result.contextTexts.push(rest.text);
      result.sourceDocuments.push({ ...rest, id });
    });

    return result;
  },
  allNamespaces: async function (client) {
    try {
      const { classes = [] } = await client.schema.getter().do();
      return classes.map((classObj) => classObj.class);
    } catch (e) {
      console.error("Weaviate::AllNamespace", e);
      return [];
    }
  },
  namespace: async function (client, namespace = null) {
    if (!namespace) throw new Error("No namespace value provided.");
    if (!(await this.namespaceExists(client, namespace))) return null;

    const weaviateClass = await client.schema
      .classGetter()
      .withClassName(camelCase(namespace))
      .do();

    return {
      ...weaviateClass,
      vectorCount: await this.namespaceCount(namespace),
    };
  },
  addVectors: async function (client, vectors = []) {
    const response = { success: true, errors: new Set([]) };
    const results = await client.batch
      .objectsBatcher()
      .withObjects(...vectors)
      .do();

    results.forEach((res) => {
      const { status, errors = [] } = res.result;
      if (status === "SUCCESS" || errors.length === 0) return;
      response.success = false;
      response.errors.add(errors.error?.[0]?.message || null);
    });

    response.errors = [...response.errors];
    return response;
  },
  hasNamespace: async function (namespace = null) {
    if (!namespace) return false;
    const { client } = await this.connect();
    const weaviateClasses = await this.allNamespaces(client);
    return weaviateClasses.includes(camelCase(namespace));
  },
  namespaceExists: async function (client, namespace = null) {
    if (!namespace) throw new Error("No namespace value provided.");
    const weaviateClasses = await this.allNamespaces(client);
    return weaviateClasses.includes(camelCase(namespace));
  },
  deleteVectorsInNamespace: async function (client, namespace = null) {
    await client.schema.classDeleter().withClassName(camelCase(namespace)).do();
    return true;
  },
  addDocumentToNamespace: async function (
    namespace,
    documentData = {},
    fullFilePath = null
  ) {
    const { DocumentVectors } = require("../../../models/vectors");
    try {
      const {
        pageContent,
        docId,
        id: _id, // Weaviate will abort if `id` is present in properties
        ...metadata
      } = documentData;
      if (!pageContent || pageContent.length == 0) return false;

      console.log("Adding new vectorized document into namespace", namespace);
      const cacheResult = await cachedVectorInformation(fullFilePath);
      if (cacheResult.exists) {
        const { client } = await this.connect();
        const weaviateClassExits = await this.hasNamespace(namespace);
        if (!weaviateClassExits) {
          await client.schema
            .classCreator()
            .withClass({
              class: camelCase(namespace),
              description: `Class created by AnythingLLM named ${camelCase(
                namespace
              )}`,
              vectorizer: "none",
            })
            .do();
        }

        const { chunks } = cacheResult;
        const documentVectors = [];
        const vectors = [];

        for (const chunk of chunks) {
          // Before sending to Weaviate and saving the records to our db
          // we need to assign the id of each chunk that is stored in the cached file.
          chunk.forEach((chunk) => {
            const id = uuidv4();
            const flattenedMetadata = this.flattenObjectForWeaviate(
              chunk.properties
            );
            documentVectors.push({ docId, vectorId: id });
            const vectorRecord = {
              id,
              class: camelCase(namespace),
              vector: chunk.vector || chunk.values || [],
              properties: { ...flattenedMetadata },
            };
            vectors.push(vectorRecord);
          });

          const { success: additionResult, errors = [] } =
            await this.addVectors(client, vectors);
          if (!additionResult) {
            console.error("Weaviate::addVectors failed to insert", errors);
            throw new Error("Error embedding into Weaviate");
          }
        }

        await DocumentVectors.bulkInsert(documentVectors);
        return true;
      }

      // If we are here then we are going to embed and store a novel document.
      // We have to do this manually as opposed to using LangChains `Chroma.fromDocuments`
      // because we then cannot atomically control our namespace to granularly find/remove documents
      // from vectordb.
      const textSplitter = new RecursiveCharacterTextSplitter({
        chunkSize: 1000,
        chunkOverlap: 20,
      });
      const textChunks = await textSplitter.splitText(pageContent);

      console.log("Chunks created from document:", textChunks.length);
      const LLMConnector = getLLMProvider();
      const documentVectors = [];
      const vectors = [];
      const vectorValues = await LLMConnector.embedChunks(textChunks);
      const submission = {
        ids: [],
        vectors: [],
        properties: [],
      };

      if (!!vectorValues && vectorValues.length > 0) {
        for (const [i, vector] of vectorValues.entries()) {
          const flattenedMetadata = this.flattenObjectForWeaviate(metadata);
          const vectorRecord = {
            class: camelCase(namespace),
            id: uuidv4(),
            vector: vector,
            // [DO NOT REMOVE]
            // LangChain will be unable to find your text if you embed manually and dont include the `text` key.
            // https://github.com/hwchase17/langchainjs/blob/5485c4af50c063e257ad54f4393fa79e0aff6462/langchain/src/vectorstores/weaviate.ts#L133
            properties: { ...flattenedMetadata, text: textChunks[i] },
          };

          submission.ids.push(vectorRecord.id);
          submission.vectors.push(vectorRecord.values);
          submission.properties.push(metadata);

          vectors.push(vectorRecord);
          documentVectors.push({ docId, vectorId: vectorRecord.id });
        }
      } else {
        throw new Error(
          "Could not embed document chunks! This document will not be recorded."
        );
      }

      const { client } = await this.connect();
      const weaviateClassExits = await this.hasNamespace(namespace);
      if (!weaviateClassExits) {
        await client.schema
          .classCreator()
          .withClass({
            class: camelCase(namespace),
            description: `Class created by AnythingLLM named ${camelCase(
              namespace
            )}`,
            vectorizer: "none",
          })
          .do();
      }

      if (vectors.length > 0) {
        const chunks = [];
        for (const chunk of toChunks(vectors, 500)) chunks.push(chunk);

        console.log("Inserting vectorized chunks into Weaviate collection.");
        const { success: additionResult, errors = [] } = await this.addVectors(
          client,
          vectors
        );
        if (!additionResult) {
          console.error("Weaviate::addVectors failed to insert", errors);
          throw new Error("Error embedding into Weaviate");
        }
        await storeVectorResult(chunks, fullFilePath);
      }

      await DocumentVectors.bulkInsert(documentVectors);
      return true;
    } catch (e) {
      console.error(e);
      console.error("addDocumentToNamespace", e.message);
      return false;
    }
  },
  deleteDocumentFromNamespace: async function (namespace, docId) {
    const { DocumentVectors } = require("../../../models/vectors");
    const { client } = await this.connect();
    if (!(await this.namespaceExists(client, namespace))) return;

    const knownDocuments = await DocumentVectors.where({ docId });
    if (knownDocuments.length === 0) return;

    for (const doc of knownDocuments) {
      await client.data
        .deleter()
        .withClassName(camelCase(namespace))
        .withId(doc.vectorId)
        .do();
    }

    const indexes = knownDocuments.map((doc) => doc.id);
    await DocumentVectors.deleteIds(indexes);
    return true;
  },
  query: async function (reqBody = {}) {
    const { namespace = null, input, workspace = {} } = reqBody;
    if (!namespace || !input) throw new Error("Invalid request body");

    const { client } = await this.connect();
    if (!(await this.namespaceExists(client, namespace))) {
      return {
        response: null,
        sources: [],
        message: "Invalid query - no documents found for workspace!",
      };
    }

    const LLMConnector = getLLMProvider();
    const queryVector = await LLMConnector.embedTextInput(input);
    const { contextTexts, sourceDocuments } = await this.similarityResponse(
      client,
      namespace,
      queryVector
    );

    const prompt = {
      role: "system",
      content: `${chatPrompt(workspace)}
    Context:
    ${contextTexts
      .map((text, i) => {
        return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`;
      })
      .join("")}`,
    };
    const memory = [prompt, { role: "user", content: input }];
    const responseText = await LLMConnector.getChatCompletion(memory, {
      temperature: workspace?.openAiTemp ?? 0.7,
    });

    return {
      response: responseText,
      sources: this.curateSources(sourceDocuments),
      message: false,
    };
  },
  // This implementation of chat uses the chat history and modifies the system prompt at execution
  // this is improved over the regular langchain implementation so that chats do not directly modify embeddings
  // because then multi-user support will have all conversations mutating the base vector collection to which then
  // the only solution is replicating entire vector databases per user - which will very quickly consume space on VectorDbs
  chat: async function (reqBody = {}) {
    const {
      namespace = null,
      input,
      workspace = {},
      chatHistory = [],
    } = reqBody;
    if (!namespace || !input) throw new Error("Invalid request body");

    const { client } = await this.connect();
    if (!(await this.namespaceExists(client, namespace))) {
      return {
        response: null,
        sources: [],
        message: "Invalid query - no documents found for workspace!",
      };
    }

    const LLMConnector = getLLMProvider();
    const queryVector = await LLMConnector.embedTextInput(input);
    const { contextTexts, sourceDocuments } = await this.similarityResponse(
      client,
      namespace,
      queryVector
    );
    const prompt = {
      role: "system",
      content: `${chatPrompt(workspace)}
    Context:
    ${contextTexts
      .map((text, i) => {
        return `[CONTEXT ${i}]:\n${text}\n[END CONTEXT ${i}]\n\n`;
      })
      .join("")}`,
    };
    const memory = [prompt, ...chatHistory, { role: "user", content: input }];
    const responseText = await LLMConnector.getChatCompletion(memory, {
      temperature: workspace?.openAiTemp ?? 0.7,
    });

    return {
      response: responseText,
      sources: this.curateSources(sourceDocuments),
      message: false,
    };
  },
  "namespace-stats": async function (reqBody = {}) {
    const { namespace = null } = reqBody;
    if (!namespace) throw new Error("namespace required");
    const { client } = await this.connect();
    const stats = await this.namespace(client, namespace);
    return stats
      ? stats
      : { message: "No stats were able to be fetched from DB for namespace" };
  },
  "delete-namespace": async function (reqBody = {}) {
    const { namespace = null } = reqBody;
    const { client } = await this.connect();
    const details = await this.namespace(client, namespace);
    await this.deleteVectorsInNamespace(client, namespace);
    return {
      message: `Namespace ${camelCase(namespace)} was deleted along with ${
        details?.vectorCount
      } vectors.`,
    };
  },
  reset: async function () {
    const { client } = await this.connect();
    const weaviateClasses = await this.allNamespaces(client);
    for (const weaviateClass of weaviateClasses) {
      await client.schema.classDeleter().withClassName(weaviateClass).do();
    }
    return { reset: true };
  },
  curateSources: function (sources = []) {
    const documents = [];
    for (const source of sources) {
      if (Object.keys(source).length > 0) {
        documents.push(source);
      }
    }

    return documents;
  },
  flattenObjectForWeaviate: function (obj = {}) {
    // Note this function is not generic, it is designed specifically for Weaviate
    // https://weaviate.io/developers/weaviate/config-refs/datatypes#introduction
    // Credit to LangchainJS
    // https://github.com/hwchase17/langchainjs/blob/5485c4af50c063e257ad54f4393fa79e0aff6462/langchain/src/vectorstores/weaviate.ts#L11C1-L50C3
    const flattenedObject = {};

    for (const key in obj) {
      if (!Object.hasOwn(obj, key)) {
        continue;
      }
      const value = obj[key];
      if (typeof obj[key] === "object" && !Array.isArray(value)) {
        const recursiveResult = this.flattenObjectForWeaviate(value);

        for (const deepKey in recursiveResult) {
          if (Object.hasOwn(obj, key)) {
            flattenedObject[`${key}_${deepKey}`] = recursiveResult[deepKey];
          }
        }
      } else if (Array.isArray(value)) {
        if (
          value.length > 0 &&
          typeof value[0] !== "object" &&
          // eslint-disable-next-line @typescript-eslint/no-explicit-any
          value.every((el) => typeof el === typeof value[0])
        ) {
          // Weaviate only supports arrays of primitive types,
          // where all elements are of the same type
          flattenedObject[key] = value;
        }
      } else {
        flattenedObject[key] = value;
      }
    }

    return flattenedObject;
  },
};

module.exports.Weaviate = Weaviate;