Quickstart

obsei 1.0 is in pre-release. Install it with an explicit version specifier, which keeps dependencies on stable releases:

uv tool install "obsei[mcp]>=1.0.0a1"     # or: pip install "obsei[mcp]>=1.0.0a1"

Create a project with a ten-language sample dataset:

mkdir voc && cd voc
obsei init
export OBSEI_DB_KEY="$(openssl rand -hex 24)"          # encrypts the database
export OBSEI_PSEUDONYM_SALT="$(openssl rand -hex 24)"  # pseudonymises authors
obsei try      # preview redacted records; nothing stored or sent
obsei run      # fetch, redact, enrich, deliver, store

Keep both secrets in your secret manager: without the key the database cannot be read, and a new salt produces new pseudonyms.

Add a model

Start Ollama with a model such as qwen3:8b, then uncomment the classify enricher in obsei.yaml. Every record gets sentiment, intent, language and a confidence score. See Models for hosted and enterprise endpoints.

Ask your agent

claude mcp add obsei -- obsei mcp

Then ask: “What are the top complaints this month, by country?” See MCP and agents.

Run with Docker

Run inside the project directory. The image’s entrypoint is obsei, so arguments are subcommands:

docker run --rm -v "$PWD:/data" -w /data --user "$(id -u):$(id -g)" \
  -e OBSEI_DB_KEY -e OBSEI_PSEUDONYM_SALT \
  ghcr.io/obsei/obsei:1.0.0-alpha.1 run

Images are tagged by release (1.0.0-alpha.1); there is no latest tag yet. --user keeps the database and outputs owned by you.