[ 
https://issues.apache.org/jira/browse/CAMEL-26963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18144950#comment-18144950
 ] 

Federico Mariani commented on CAMEL-26963:
------------------------------------------

Agreed. Checked on main:
* {{camel_ai_pipeline_scaffold}} refers to six beans it never defines 
({{#buildSummarizationPrompt}}, {{#buildExtractionPrompt}}, 
{{#buildRagPrompt}}, {{#buildClassificationPrompt}}, 
{{#mergeDoclingAndTextract}}, {{#bedrockEmbedding}}), and its default model is 
Claude Sonnet 4.
* {{camel_route_cost_estimate}} prices every Bedrock call as Claude Sonnet 4 
($3/$15 per MTok).
* {{camel_route_test_scaffold}}: I would drop it rather than write a Citrus 
scaffold. {{camel test init}} already creates a Citrus test, the example 
catalog has examples with Citrus tests, and {{camel_write_file}} checks a 
Citrus test and says which actions it takes.
* {{camel_dependency_security_audit}} folds well into 
{{camel_dependency_check}}: both take the same arguments ({{pomContent}}, 
{{routes}}, {{runtime}}, {{camelVersion}}, {{platformBom}}), and the advisories 
are catalog data, so the check gains a {{vulnerabilities}} section without a 
network call.

One question on {{camel_error_diagnose}}. It is not only a camel-jbang-mcp 
tool: it is one of the shared authoring tools ({{AuthoringTools.java}}, in the 
core set), which {{camel mcp}} and {{camel tui --mcp}} both list and the TUI 
agent uses, and you extended its table from the local-model benchmark 
(CAMEL-24882, CAMEL-26316, CAMEL-25982). Does the benchmark show that it helps? 
If yes, it stays on both servers so the shared set remains the same everywhere; 
if not, it leaves both. The first PR leaves it as it is.

I am working on the PR for the three removals, the fold and the upgrade-guide 
entry.

_Claude Code on behalf of Croway_

> camel-jbang-mcp - remove or fold the niche and stale tools that the 
> consolidation does not cover yet
> ----------------------------------------------------------------------------------------------------
>
>                 Key: CAMEL-26963
>                 URL: https://issues.apache.org/jira/browse/CAMEL-26963
>             Project: Camel
>          Issue Type: Sub-task
>          Components: camel-jbang
>            Reporter: Claus Ibsen
>            Assignee: Federico Mariani
>            Priority: Major
>             Fix For: 4.25.0
>
>
> This adds to the consolidation of the parent issue: tools it does not cover 
> yet, which are niche, stale or out of favor. Every tool is listed to every 
> client (CAMEL-25537), and in the evaluation of CAMEL-25537 (4.22.1, 1,729 
> calls) none of these tools was called.
> h3. Remove
> || Tool || Why ||
> | {{camel_ai_pipeline_scaffold}} | A template for one stack: Docling or 
> Textract, then Bedrock. The route it returns refers to beans it never defines 
> ({{#buildSummarizationPrompt}}, {{#buildRagPrompt}}, 
> {{#mergeDoclingAndTextract}}, {{#bedrockEmbedding}} and others), so it does 
> not run as its description says. Its default model is Claude Sonnet 4 
> ({{anthropic.claude-sonnet-4-20250514-v1:0}}), retired on the Claude API on 
> Jun 15 2026 (Amazon Bedrock sets its own dates), and its examples name Claude 
> Opus 4, also retired. An agent can build such a route with the catalog tools. 
> |
> | {{camel_route_cost_estimate}} | Prices written into the code, and prices 
> change faster than Camel releases. The parent issue found that it prices 
> Bedrock as Claude Sonnet whatever the model. |
> h3. Replace or remove
> || Tool || Why ||
> | {{camel_route_test_scaffold}} | It writes a JUnit 5 test in Java 
> ({{CamelTestSupport}} or {{@CamelSpringBootTest}}), and the parent issue 
> found that it tests a new route, not the real one. The tests that the CLI 
> runs are Citrus tests in YAML ({{camel test run}}), as the examples use them. 
> A scaffold of a Citrus YAML test for a given route would fit; otherwise drop 
> the tool. |
> h3. Fold or drop
> || Tool || Into || Why ||
> | {{camel_dependency_security_audit}} | {{camel_dependency_check}} | Both 
> read the POM. The audit matches the dependencies against the same advisory 
> data that {{camel_security_advisories}} returns. |
> | {{camel_error_diagnose}} | drop, unless a benchmark shows that it helps | A 
> fixed list of 19 exceptions with fixed causes. {{camel_get_errors}} returns 
> the real error of a running integration, and the catalog tools answer the 
> rest. |
> Keep {{camel_security_advisories}}: it returns the published CVE data, which 
> an agent cannot get otherwise. With the parent issue folding 
> {{camel_route_harden_context}} into {{camel_security_scan}}, two security 
> tools remain: the advisories and the scan.
> cc [~acosentino]: most of these tools came with CAMEL-23979 and CAMEL-24219 
> to CAMEL-24223, for 4.22.0. Please say if one of them has users we do not 
> know about.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to