davsclaus opened a new pull request, #3154:
URL: https://github.com/apache/camel-kamelets/pull/3154
_Claude Code on behalf of Claus Ibsen_
A batch of Kamelets rewritten to the documentation standard proposed in
#3140, `docs/modules/ROOT/pages/documenting.adoc`, like #3141–#3153.
This batch has 27 Kamelets and 8 partial pages:
- **search:** Elasticsearch, OpenSearch and Solr;
- **cache:** `caffeine-action`, Infinispan and Redis;
- **documents and AI:** `docling-convert-action`, `langchain4j-ingest-sink`,
`langchain4j-ingest-file-source` and `tika-extract-text-action`;
- **FHIR;**
- **other services:** `sap-netweaver-sink`, `graphql-sink` and
`splunk-hec-sink`;
- **geo:** `nominatim-geocode-action` and `ogcapi-features-action`;
- **PQC:** `pqc-signature-action` and `pqc-kem-action`;
- **the rest:** `snmp-source`, `bitcoin-source` and `exec-sink`.
## How they were checked
**Every one was run**, in one of these ways:
- against a local service from `camel infra`: Redis, Infinispan, OpenSearch,
Solr, FHIR and Docling;
- against an Elasticsearch container;
- against a local stand-in, for SAP NetWeaver, GraphQL and Splunk HEC;
- against a public service, for Nominatim, the OGC API demo and the Binance
ticker.
## Facts that were missing or wrong before
- **`exec-sink`:**
- the body is the command's input, and the output is stdout, plus the exit
value and stderr as headers;
- `executable` cannot take arguments;
- a non-zero exit does not fail the message;
- `Good to know` says it runs a program by design.
- **The PQC actions:**
- the key pair is created when the integration starts and cannot be
supplied, so a signature can be verified only in the same running integration;
- the KEM output is Java key objects.
- **`tika-extract-text-action`:**
- a PDF gives empty text unless the Tika PDF parser module is added;
- all non-Camel headers are removed.
- **`docling-convert-action`:**
- outputs Markdown text;
- `doclingServeUrl` alone still runs the local CLI, so
`useDoclingServe=true` is also needed.
- **The langchain4j Kamelets** are rewritten for a beginner:
- the outcomes are INGESTED, EMPTY, SKIPPED and FILTERED;
- re-ingesting the same ID adds pieces and does not replace the old ones;
- the file source re-reads everything after a restart.
- **Redis:**
- SET, GET, RPUSH, PUBLISH and (P)SUBSCRIBE checked;
- there is no password and no TLS;
- the `redis-source` partial page described options that do not exist.
- **Infinispan:**
- the cache must exist;
- `secure` means login, not TLS;
- the default `DIGEST-MD5` is refused by current Infinispan servers, which
need `SCRAM-SHA-512`.
- **Elasticsearch and OpenSearch:**
- the sink returns the document ID;
- the search sources return all hits again on every poll, every second.
- **The Solr source** gives a Map-like object, not JSON text.
- **`fhir-sink`:** only create, update, delete and validate work.
`fhir-source` gives a Bundle as JSON text and polls about every 0.5 s.
- **`sap-netweaver-sink`** is a GET reader that never sends the body.
- **`graphql-sink`:** the body is the query, and it does not do
subscriptions.
- **`bitcoin-source`** reads the Binance BTC/USDT ticker; it is not a
Bitcoin node.
- **`snmp-source`** gives XML text.
- **Partial pages:** several were placeholder text or claimed wrong options
(Redis, OpenSearch, Solr, Infinispan, Elasticsearch, GraphQL, FHIR).
## Left out, and found on the way
- **`djl-image-to-text-action` is not in this PR,** because it fails on
every image tried (`RasterFormatException: negative or zero width`, on 4.23 and
4.24). It will be described with its fix.
- **The Elasticsearch and OpenSearch search sources give unusable document
fields.** Gson marshals Jackson `ObjectNode`s, which gives
`{"_children":{…},"_nodeFactory":…}`.
- **`ogcapi-features-action`:** `split=true` has no effect.
- **`redis-sink`:**
- `channels` is never used, and PUBLISH needs a `channel` header;
- the default command of the sink is `GET`.
- **`pqc-kem-action`:** several algorithms in its enum fail on Camel 4.24,
because `symmetricKeyLength` is not exposed.
- **Types, defaults and required lists:**
- `infinispan-source` `eventTypes` has an enum, but takes a
comma-separated value, and its default is not one of the enum values;
- `fhir-sink` `apiName` is needed but not `required`;
- `fhir-source` `username` and `password` are required even for servers
without login;
- `splunk-hec-sink` `time` is a `string`.
- **Commonly needed component options these Kamelets do not expose:**
- arguments and a timeout for `exec-sink`;
- expiry and a maximum size for `caffeine-action`;
- a password or TLS for Redis;
- paging for the search sources;
- HTTPS for Solr;
- the poll delay and OAuth for FHIR;
- your own keys for PQC.
These are for a follow-up.
## Notes for review
- No template, type, default or required list was changed. The `dataTypes`
blocks are `headers` only, with no `default`.
- `KameletsCatalogTest.testSupportedHeaders`: the catalog now returns the
headers these Kamelets declare instead of those of their components. The
changed counts:
- `elasticsearch-index-sink` 10 → 4
- `exec-sink` 0 → 2
- `infinispan-source` 6 → 4
- `infinispan-sink` 14 → 2
- `redis-sink` 29 → 4
- `redis-source` 28 → 1
- This PR is based on `main`; if #3140 merges first, I will rebase it and
regenerate the index.
`./mvnw clean install` from the root passes.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]