Skip to content

Self-host DocSlurp

Run the whole stack on one host you control, with Docker Compose.

Self-host DocSlurp

DocSlurp turns files into page-by-page Markdown, chunks, embeddings and structured fields, and serves them through search, RAG context and a cited chat endpoint. This section covers running it yourself.

On this page

Self-hosting is for two groups. Teams that need documents to stay on their own infrastructure can run the whole stack on one host. Developers who want to try DocSlurp, or build against it, can run the same stack locally.

The current install is a single host running Docker Compose. It is suitable for a small private installation or a server behind a TLS reverse proxy. It is not a multi-host or high-availability design. There is no bundled web UI; use the REST API, CLI, SDK or MCP.

Clients: REST, SDK, CLI, MCP, OpenAI-compatible chat, embeds

Server (Hono, /v1, port 8787)

Postgres + pgvector

Redis: rate limits, ephemeral

Object storage: S3 or MinIO

LanceDB vector index under lancedb/

Temporal

Workers

Parser sandbox (bubblewrap)

Optional AI providers

Architecture explains each part.

Choose your path

Path Status
Docker Compose on a single host Available. Start with Docker Compose.
Pluggable parser sandbox providers, starting with Modal as the first remote provider Planned
Terraform stacks for Fly, AWS, GCP, Vercel and Cloudflare Planned
docslurp deploy setup wizard Planned
Agent skill for guided setup Planned

Pages

  • Docker Compose: install on one host, start, check and stop the stack, and switch between the lean and full worker.
  • Configuration: environment options, storage and vector index URIs, and AI providers.
  • Public access: TLS reverse proxy, forwarded headers and client IP handling.
  • Operations: backups and restore, upgrades, and troubleshooting.
  • Architecture: system map, ingest sequence, trust boundaries and Compose topology.
  • Requirements: host requirements, sizing and GPU.
  • Security: parser sandbox, malware scanning, secrets and network exposure.

Jump to