Remove global Postgres schema registration in favor of explicit schema-aware sink factory wiring, and update weatherfeeder to register the Postgres sink explicitly. Add optional per-source HTTP timeout and response body limit overrides while keeping feedkit defaults. Remove remaining legacy source/config compatibility surfaces, including singular kind support and old source registry/type aliases, and migrate weatherfeeder sources to plural `Kinds()` metadata. Clean up related docs, tests, and sample config to match the new Postgres, HTTP, and NATS configuration model.
- add new `processors` package with canonical `Processor` interface
- add `processors.Registry` with Register/Build/BuildChain factory model
- switch `pipeline.Pipeline` to `[]processors.Processor`
- replace `normalize.Registry` + registry adapter with direct `normalize.Processor`
- remove `normalize/registry.go`
- update root docs to position normalize as one optional processing stage
- add tests for processors registry, normalize processor behavior, and pipeline flow
BREAKING CHANGE:
- `pipeline.Processor` removed; use `processors.Processor`
- `normalize.Registry` and old normalize processor adapter APIs removed
- downstream daemons must update processor wiring to new `processors.Registry`
and `normalize.NewProcessor(...)`
- Introduce explicit source interfaces: sources.PollSource and sources.StreamSource, with shared sources.Input (Name() only).
- Remove mandatory Kind() from the base source contract to support sources that emit multiple kinds.
- Add config.SourceMode (poll, stream, or omitted/auto) and SourceConfig.Kinds (plural expected kinds), while keeping legacy SourceConfig.Kind for compatibility.
- Enforce mode semantics in config validation (poll requires every, stream forbids every) and detect mode/driver mismatches in sources.Registry.
- Update docs and tests for the new source model and config behavior.
Introduce an optional normalization stage for feedkit pipelines via the new
normalize package. This adds:
- normalize.Normalizer interface with flexible Match() semantics
- normalize.Registry for ordered normalizer selection (first match wins)
- normalize.Processor adapter implementing pipeline.Processor
- Pass-through behavior when no normalizer matches (normalization is optional)
- Func helper for ergonomic normalizer definitions
Update root doc.go to fully document the normalization model, its role in the
pipeline, recommended conventions (Schema-based matching, raw vs normalized
events), and concrete wiring examples. The documentation now serves as a
complete external-facing API specification for downstream daemons such as
weatherfeeder.
This change preserves feedkit’s non-framework philosophy while enabling a
clean separation between data collection and domain normalization.
- Add logging.Logf as the canonical printf-style logger type used across feedkit.
- Update scheduler and dispatch to alias their Logger types to logging.Logf.
- Eliminates type-mismatch friction when wiring one log function through the system.
- Add dispatch.CompileRoutes(*config.Config) ([]dispatch.Route, error)
- Compiles config routes into dispatch routes with event.ParseKind normalization.
- If routes: is omitted, defaults to “all sinks receive all kinds”.
- Expand config param helpers for both SourceConfig and SinkConfig
- Add ParamBool/ParamInt/ParamDuration/ParamStringSlice (+ Default variants).
- Supports common YAML-decoded types (bool/int/float/string, []any, etc.)
- Keeps driver code cleaner and reduces repeated type assertions.
- Fix Postgres sink validation error prefix ("postgres sink", not "rabbitmq sink").
