[Bug]: ChromaServerError: The server encountered an error while handling the request.

[galer7]

What happened?

Running Chroma 0.5.15 via official docker image.

Error from client code:

ChromaServerError: The server encountered an error while handling the request.
    at parseServerError (/Users/galer7/p/<project>/node_modules/.pnpm/chromadb@1.9.2_cohere-ai@7.13.0_encoding@0.1.13__encoding@0.1.13_openai@4.68.0_encoding@0.1.13_zod@3.23.8_/node_modules/chromadb/src/ChromaFetch.ts:35:10)
    at chromaFetch (/Users/galer7/p/<project>/node_modules/.pnpm/chromadb@1.9.2_cohere-ai@7.13.0_encoding@0.1.13__encoding@0.1.13_openai@4.68.0_encoding@0.1.13_zod@3.23.8_/node_modules/chromadb/src/ChromaFetch.ts:75:17)
    at processTicksAndRejections (node:internal/process/task_queues:105:5)
    at async Collection.add (/Users/galer7/p/<project>/node_modules/.pnpm/chromadb@1.9.2_cohere-ai@7.13.0_encoding@0.1.13__encoding@0.1.13_openai@4.68.0_encoding@0.1.13_zod@3.23.8_/node_modules/chromadb/src/Collection.ts:73:5)
    at async main (/Users/galer7/p/<project>/server/scripts/ingest-notion.ts:49:5) {
  cause: undefined
}

Error from the server in docker:

INFO:     [19-10-2024 00:04:36] 192.168.215.0:59783 - "POST /api/v1/collections/27c16ee6-d4bd-4afd-b75b-499ec3d25394/add HTTP/1.1" 500

It only happens when I add data. My first instinct was the default embedding function.

No other error messages...

Versions

Chroma 0.5.15, Docker, Mac 14.6.1

Relevant log output

No response

[galer7]

this is my ChromaService, I think it's pretty basic

import { ChromaClient, IncludeEnum } from "chromadb";
import { log } from "../decorators/logger";
export interface Document {
  text: string;
  embedding: number[];
  metadata: Record<string, any>;
}

type Collection = ReturnType<ChromaClient["getOrCreateCollection"]>;

export class ChromaService {
  private client: ChromaClient;

  constructor(
    private collectionName: string = process.env.CHROMA_COLLECTION_NAME!
  ) {
    this.client = new ChromaClient({
      path: process.env.CHROMA_URL!,
    });
  }

  @log("ChromaService")
  async getCollection(): Promise<Collection> {
    return this.client.getOrCreateCollection({
      name: this.collectionName,
    });
  }

  @log("ChromaService")
  async addDocuments(documents: Document[]): Promise<void> {
    const ids = documents.map((_, index) => `doc_${index}`);
    const embeddings = documents.map((doc) => doc.embedding);
    const metadatas = documents.map((doc) => doc.metadata);
    const texts = documents.map((doc) => doc.text);

    const collection = await this.getCollection();

    return collection.add({
      ids,
      embeddings,
      metadatas,
      documents: texts,
    });
  }

  @log("ChromaService")
  async deleteDocuments(ids: string[]): Promise<string[]> {
    const collection = await this.getCollection();

    return collection.delete({ ids });
  }

  @log("ChromaService")
  async queryByVector(
    vector: number[],
    limit: number = 1
  ): Promise<Document[]> {
    const collection = await this.getCollection();
    const results = await collection.query({
      queryEmbeddings: [vector],
      nResults: limit,
      include: [
        "metadatas" as IncludeEnum,
        "documents" as IncludeEnum,
        "distances" as IncludeEnum,
      ],
    });

    return results.documents[0].map((text, index) => ({
      text: text!,
      embedding: vector, // Note: Chroma doesn't return the original embedding
      metadata: results.metadatas[index],
    }));
  }

  @log("ChromaService")
  async queryByText(text: string, limit: number = 1): Promise<Document[]> {
    const collection = await this.getCollection();
    const results = await collection.query({
      queryTexts: [text],
      nResults: limit,
      include: [
        "metadatas" as IncludeEnum,
        "documents" as IncludeEnum,
        "distances" as IncludeEnum,
      ],
    });

    return results.documents[0].map((text, index) => ({
      text: text!,
      embedding: [], // Note: Chroma doesn't return the embedding
      metadata: results.metadatas[0][index]!,
    }));
  }

  @log("ChromaService")
  async queryByMetadata(metadata: Record<string, any>): Promise<Document[]> {
    const collection = await this.getCollection();
    const results = await collection.get({
      where: metadata,
      include: ["metadatas" as IncludeEnum, "documents" as IncludeEnum],
    });

    return results.documents.map((text, index) => ({
      text: text!,
      embedding: [], // Note: Chroma doesn't return the embedding by default
      metadata: results.metadatas[index]!,
    }));
  }
}

[tazarov]

@galer7, thanks for raising this. Are these the only server logs you were able to find? Any stack trace in the docker container?