Point Prologue at your system

Turn on the capture sources your system actually has, wire the Extractor in as a sidecar, and confirm captures are landing — before you worry about interpretation.

Prerequisites

• The system you’re capturing is reachable over HTTP, or has a SQL Server / PostgreSQL database you can connect to (or both).
• Docker, to run cratis/prologue-extractor.
• For SQL Server capture: a connection string with permission to enable Change Data Capture.
• For PostgreSQL capture: a connection string with permission to create a replication slot and publication, and wal_level = logical on the server.

Pick your capture sources

You don’t need all four. A system with no database access still gets useful HTTP captures; a batch system with no HTTP surface still gets database captures. Decide which of sqlServer, postgres, openTelemetry, and the HTTP reverse proxy apply, and leave the rest at their empty/disabled defaults. See Reference — Capture sources for what each one needs.

Write cratis-prologue.json

cratis-prologue.json

{
    "prologue": {
        "output": {
            "kind": "Json",
            "json": { "directory": "/captures" }
        },
        "correlation": { "windowMilliseconds": 2000 },
        "sqlServer": [
            { "name": "main", "connectionString": "Server=...;Database=...;", "tables": ["Orders", "OrderLines"] }
        ],
        "postgres": [],
        "openTelemetry": { "enabled": false }
    },
    "reverseProxy": {
        "routes": {
            "monitored": { "clusterId": "monitored", "match": { "path": "{**catch-all}" } }
        },
        "clusters": {
            "monitored": { "destinations": { "primary": { "address": "http://your-system:8081/" } } }
        }
    }
}

• Leave tables out (or empty) to capture every user table’s changes instead of an allowlist.
• Omit the reverseProxy block entirely if you’re not capturing HTTP — the Extractor just won’t proxy anything.
• Switch output.kind to "Api" with an output.api.endpoint instead of "Json" to post captures straight to a Receiver rather than writing files.

The full property list, with every default, is in Reference — Configuration.

Run the Extractor as a sidecar

docker run -d --name prologue-extractor \
  -p 8080:8080 -p 4317:4317 -p 4318:4318 \
  -v "$(pwd)/cratis-prologue.json:/config/cratis-prologue.json:ro" \
  -v "$(pwd)/captures:/captures" \
  cratis/prologue-extractor

Port 8080 is the reverse proxy your system’s traffic should go through instead of hitting your system directly; 4317/4318 are the OTLP gRPC/HTTP endpoints if you’re forwarding telemetry through the Extractor. Point your own reverse proxy, load balancer, or client traffic at 8080 — anything that reaches your system without going through it produces no HTTP captures.

If you’d rather not mount a config file, every property has a double-underscore environment variable equivalent (Prologue__SqlServer__0__ConnectionString, ReverseProxy__Clusters__monitored__Destinations__primary__Address, and so on) — env vars win over the file, which is handy for injecting secrets from your orchestrator instead of writing them into cratis-prologue.json.

Confirm captures are landing

Exercise the system — a real request, or your own smoke test — through port 8080, then check the mounted captures folder for new .jsonl files (or your Receiver’s MongoDB collection, if that’s where you pointed output). No files after a state-changing request usually means the reverse proxy destination is wrong, or the request didn’t use POST/PUT/DELETE — see Troubleshooting.

Next

• Running the Interpreter — turn the folder of captures into an event model.
• Reference — Capture sources — the full property list per source.