Deploy WeKnora

Tencent's RAG document Q&A: web UI, API, DocReader, ParadeDB and Redis

Deploy WeKnora

/var/lib/postgresql/data

Just deployed

/data/files

Just deployed

/data

Just deployed

Deploy on Railway

Deploy and Host WeKnora on Railway

WeKnora is Tencent's open-source document understanding and retrieval platform for LLMs. It parses PDFs, Word files, slides, images and web pages, splits and indexes them with vector and keyword search, and answers questions over your knowledge bases through a web chat, an agent mode and a REST API. This template deploys WeKnora v0.9.0.

About Hosting WeKnora

WeKnora runs as several cooperating services: an nginx web UI, a Go API server that also runs the background ingestion workers, a Python DocReader that parses documents over gRPC, PostgreSQL with the ParadeDB extensions (pg_search for BM25 keyword search, pgvector for embeddings) and Redis for task queues and chat streams. This template wires all five over Railway's private network with generated passwords and keys, keeps uploaded files and both databases on volumes, closes public sign-up, and creates your administrator account on the first boot. WeKnora does not ship models: after logging in, add an LLM and an embedding model (OpenAI-compatible APIs, DeepSeek, Qwen, a remote Ollama and others) in the settings.

Common Use Cases

  • Question answering over internal documents, manuals and policies, with cited sources
  • A private knowledge base and AI assistant for a team, with per-tenant spaces and API keys
  • A RAG backend for your own apps and chat widgets through WeKnora's REST API and embeddable widget

Dependencies for WeKnora Hosting

  • An API key for at least one LLM and one embedding model provider (entered in WeKnora's settings after deploy)
  • The five services below, all included in this template

Deployment Dependencies

  • WeKnora (public): the web UI (nginx). It serves the app and proxies /api and /files to WeKnora App. Health check /.
  • WeKnora App (private): the Go API server, database migrations and ingestion workers. Uploaded files live on its volume at /data/files. Health check /health.
  • DocReader (private): the gRPC document parser (PDF, Office, images, HTML) on port 50051.
  • Postgres (private): ParadeDB v0.25.11 (PostgreSQL 17 with pg_search and pgvector), on a volume. It stores accounts, knowledge bases, chunks, embeddings and the search indexes.
  • Redis (private): Railway's Redis, with a generated password and a volume, for task queues and chat streams.
  • Source repository: aalfath/weknora-railway-template
  • Upstream: Tencent/WeKnora (MIT)

Implementation Details

The one prompt. WEKNORA_ADMIN_EMAIL on WeKnora App is the email of your administrator account. On the first boot the App registers it on a loopback-only instance, then starts the real server with sign-up closed and promotes that account to system administrator. The password is generated: open WeKnora App, then Variables, and copy WEKNORA_ADMIN_PASSWORD. Change it in the UI after logging in.

After deploy.

  1. Open the WeKnora service's public URL and log in with the admin email and the generated password.
  2. Add your models under Settings: a chat LLM, an embedding model and, optionally, a rerank or vision model.
  3. Create a knowledge base, choose the embedding model, and upload documents.

New users join by invitation, because public sign-up is off (DISABLE_REGISTRATION=true on WeKnora App). To open sign-up, set it to false and redeploy.

Optional settings (add them to WeKnora App's Variables): OLLAMA_BASE_URL for a remote Ollama server, WEKNORA_LANGUAGE for the default language of generated answers, and S3-compatible storage (STORAGE_TYPE=s3 with the S3_* variables) instead of the local volume. Neo4j knowledge graphs, other vector stores (Qdrant, Milvus, Elasticsearch), Langfuse tracing, SearXNG web search and the Docker code sandbox are left out; see WeKnora's documentation if you want to add them.

Updates. Every service except Redis (Railway's own database image) builds from the template repository with the upstream image pinned by version and digest. When a new WeKnora version is tested, the repository's main branch is updated and Railway shows an "update available" notice on your project; you choose when to apply it. Read CHANGELOG.md first. WeKnora App runs database migrations automatically on start.

Why Deploy WeKnora on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying WeKnora on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.


Template Content

More templates in this category

View Template
Chat Chat
Chat Chat, your own unified chat and search to AI platform.

okisdev
117
View Template
Weaviate
Vector database with hybrid search and an API key-protected API

GitHub Locker
1
View Template
Möbius Personal AI OS
Self-hosted AI agent that builds and runs apps for you, with memory.

bento
1