Jiri Ondrusek created CAMEL-26341:
-------------------------------------
Summary: camel-langchain4j-ingest: answer FILTERED for a media
document whose type cannot be determined
Key: CAMEL-26341
URL: https://issues.apache.org/jira/browse/CAMEL-26341
Project: Camel
Issue Type: Improvement
Components: camel-ai
Reporter: Jiri Ondrusek
Assignee: Jiri Ondrusek
With {{modality=media}}, the producer takes the media type from the
{{contentType}} option, else from the file extension of the document id
({{MediaTypes.fromDocumentId}}). When neither gives a type, it throws
{{IllegalArgumentException}}: "cannot tell the media type of document '...'
from its extension. Set the contentType endpoint option."
A directory that also holds files of other kinds (a README, a temp file, an
editor backup) therefore fails the exchange for each such file, again on every
poll, because a failed exchange does not keep its idempotent key. Committing
the key instead would keep the file skipped even after a configuration fix, so
the exchange cannot simply succeed either.
Proposal: for a document whose type cannot be determined, answer a {{FILTERED}}
result like the {{includeId}}/{{excludeId}} filters do. It then wrote nothing
and releases its dedup claim. A hint is logged at debug level so a misspelled
extension can still be found. Documents with {{contentType}} set are not
affected.
Found while reviewing the Camel Quarkus ingest extension (CAMEL-QUARKUS pull
request 9340), which has to translate this case.
_Generated with [Claude Code|https://claude.com/claude-code] on behalf of
[~jondruse]_
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
