Libreflare Worker CLI

CLI and library for Libreflare managed Worker workflows. It currently generates Wrangler config from libreflare.config.yaml.

CLI

libreflare-worker config generate

By default the command reads libreflare.config.yaml and wrangler.jsonc from the current directory, writes wrangler.generated.jsonc, and writes .wrangler/deploy/config.json so Wrangler uses the generated config.

By default runtime config is written to src/libreflare-runtime-config.generated.json. This avoids Cloudflare's per-variable size limit for projects with many routes. Pass --runtime-config vars to write the legacy LIBREFLARE_CONFIG Worker variable instead. The JSON target still writes Wrangler routes and required secret bindings to wrangler.generated.jsonc.

Before Libreflare creates libreflare.config.yaml, the command points Wrangler at the base wrangler.jsonc. In the default JSON mode, it also writes a generated JSON stub so the Worker can be bundled before managed config exists.

Managed Worker Config

version: 2
zoneName: example.com
logging:
  apiUrl: https://logs.example.com/insert/jsonline?_stream_fields=stream
  auth: true
routes:
  - prefix: /lf/pages
    domains:
      - example.com
      - www.example.com
    logging:
      sourceKey: pages
      sourceKeyRules:
        - expression: cf.verified_bot_category eq "AI Crawler"
          sourceKey: pages-ai-crawler
      varyByMonth: true
    rule: starts_with(http.request.uri.path, "/wiki/")
  - prefix: /lf/api
    domains:
      - api.example.com
    logging:
      sourceKey: api
    rule: starts_with(http.request.uri.path, "/api/")

zoneName is the Cloudflare zone where Wrangler should create Worker routes. When routes[].domains is present, it must contain one or more plain hostnames, the zone apex, *.zoneName, concrete zone subdomains, or explicit Cloudflare for SaaS custom hostnames attached to the zone. The CLI generates one Wrangler route per route domain and prefix. Omit routes[].domains when the Worker repository owns Wrangler route config for that route. Runtime route matching is path-prefix based, so route prefixes must be unique within the config.

routes[].logging.sourceKey is the required default logging stream key. Optional sourceKeyRules are evaluated in order inside the Worker before the origin fetch; the first match wins. Rule expressions use native Cloudflare Rules syntax, do not support rule fragments, and cannot reference Cloudflare lists such as $known_crawlers. Worker-visible bot fields such as cf.client.bot and cf.verified_bot_category are allowed.

For unmanaged logging, omit the top-level logging block and omit route logging blocks. The CLI still generates Wrangler routes from routes that include domains, and it still writes route/rule metadata to the selected runtime config target without logging fields. Logging variables and logging auth bindings stay owned by the Worker repository:

version: 2
zoneName: example.com
routes:
  - prefix: /lf/custom
    domains:
      - example.com
    rule: starts_with(http.request.uri.path, "/custom/")