Skip to content
Get packs

serviette

A no-code server that answers questions from your documents and updates answers when files change.

GitHub Download source
Stars
58.7k
Forks
1.5k
Licence
MIT
Language
Python
Last update
21 days ago
Type
App

What it is

serviette points a question-answering chat and REST API at a folder of documents, so answers cite their sources. You configure it with one YAML file and a couple of commands. When a document is edited, added or deleted, the answers update within seconds. It is built on the Pathway Live Data Framework and supports several vector databases.

  • Run a chat over a folder of company policy documents with cited sources
  • Serve a /api/v1/retrieve endpoint that other tools can query for passages
  • Index PDFs, Office files and scanned images from a local folder and ask questions about them
  • Connect a Google Drive, S3 or SharePoint source so edits there reach the answers

Great for

  • Teams that want a chat over internal documents without writing code
  • Developers who want a quick local RAG prototype with DuckDB as the default store
  • People whose documents change often and need answers to stay current
  • Anyone who wants a REST retrieval endpoint with a built-in chat page

Not ideal for

  • Users who need conversation memory or user accounts, which the README says are not built in yet
  • Teams wanting a hosted service, since the README lists no hosted offering
  • Anyone unwilling to get a free Pathway license key, which the README says is required
Step-by-step guide: how to use servietteInstall, first run, examples, tips and FAQs
serviette alternativesAnythingLLM vs servietteLangChain vs servietteLlamaIndex vs serviette

How to get it

  1. 1

    Install with a package manager

    You'll need: Python · pip 23.1 or newer

    Install from PyPI

    pip install serviette

    Set Pathway license key

    export PATHWAY_LICENSE_KEY=...

    Run the demo with OpenAI embeddings

    serviette demo --embedder openai
  2. 2

    Run it

    • Get a free Pathway license key at https://pathway.com/framework/get-license and export it as PATHWAY_LICENSE_KEY
    • Run serviette demo --embedder openai and open the chat at http://localhost:8989
    • Edit or add a file in ./serviette-demo/docs and ask the question again to see the answer change
    • For your own documents, run serviette wizard to generate a config, then serviette up --config config.yaml
  3. 3

    Get the most out of it

    • Use serviette wizard to generate the YAML config instead of writing it by hand
    • Choose a vector backend such as pgvector, Qdrant or Milvus if you need a shared external database
    • Set OPENAI_API_KEY to get generated answers; omit it to run keyless, where answers quote retrieved snippets
    • Connect extra sources like Google Drive, S3 or SFTP, which the README says can be watched live together in one config
    • For production, run serviette indexer and serviette server separately, as the README describes

Downloads

Source code (.zip)main branch · straight from GitHubOfficial website & docsgithub.com/pathwaycom/serviette/blob/main/docs/README.md

Strengths

  • Runs with two commands and a single YAML config, per the README
  • Changes to documents are reflected in answers within seconds, according to the README
  • Supports eight vector database backends and defaults to embedded DuckDB with no extra service

Trade-offs

  • Requires a free Pathway license key before it runs, and the README says the default install of the local embedder is large (~0.8 GB CPU build)
  • Answers one question at a time with no conversation memory or user accounts
  • Embedding with an API sends document text to that provider; the local default keeps it on your machine

How it compares

ProjectStarsTypeLevelUpdatedBest for
serviette (this)58.7kAppintermediate21 days agoTeams that want a chat over internal documents without writing code
llama_index52.8kLibraryintermediate10 days agoPython developers building retrieval-augmented generation (RAG) apps over private documents
langchain149kFrameworkintermediate4 days agoPython developers building LLM-powered applications
anything-llm67.8kAppbeginner6 days agoPeople who want a private ChatGPT-style assistant over their own PDFs and text files
PageIndex40kLibraryintermediate7 days agoDevelopers building question-answering over long PDFs such as annual reports or manuals
LightRAG40.5kLibraryintermediate7 days agoDevelopers building question-answering over their own document collections

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.