Make AnthropicChatModel internal methods private

[arimu1]

Thank you for taking time to contribute this pull request!
You might have already read the contributor guide, but as a reminder, please make sure to:

• Add a Signed-off-by line to each commit (git commit -s) per the DCO
• Rebase your changes on the latest main branch and squash your commits
• Add/Update unit tests as needed
• Run a build and make sure all tests pass prior to submission

For more details, please check the contributor guide.
Thank you upfront!

Description

AnthropicChatModel.internalStream (and internalCall) were left public after the model internal methods were supposed to be private (see upgrade notes). Other chat model implementations already use private.

This change restricts both methods to private, matching:

• upgrade notes: "All internalCall and internalStream methods in model classes have been changed to private"
• OpenAI, Ollama, Bedrock Converse, Mistral, DeepSeek, Google GenAI

Maintainer @guanxuc confirmed on #6700 that these should be private.

No external callers needed adjustment; both methods are only used within AnthropicChatModel (call/stream and recursive tool-call loops).

Fixes #6700

Testing

./mvnw -pl models/spring-ai-anthropic -am test -DskipITs

Results: 124 tests, 0 failures, 0 errors (JDK 17.0.19)

AI assistance

Created with AI assistance and reviewed by a human.

[kuntal1461]

Code Review — Issue #6700: AnthropicChatModel exposes internalStream and internalCall publicly

Verdict: Approve — no blocking findings.

Context

AnthropicChatModel was rewritten in March 2026 to use the official Anthropic Java SDK (b0ce25aed). That rewrite reintroduced public access on internalCall and internalStream, undoing the earlier privacy fix (da3bec4a8, Feb 2026). Every other chat model in the project (OpenAiChatModel, OllamaChatModel, MistralAiChatModel, DeepSeekChatModel, GoogleGenAiChatModel, BedrockProxyChatModel) already declares these methods private, and the upgrade notes explicitly state:

All internalCall and internalStream methods in model classes have been changed to private.

This PR is the minimal, correct fix to restore that contract for AnthropicChatModel.

Call-site analysis

A full-codebase grep confirms:

• internalCall is called only from call() (line 262, same class).
• internalStream is called only from stream() (line 283, same class).
• No external callers, no test callers, no reflection usage.
• The class is final, so no subclass extension concerns.
• Neither method appears in ChatModel or StreamingChatModel, so there is no interface override.

Testing

Ran locally:

./mvnw -pl models/spring-ai-anthropic test
Tests run: 124, Failures: 0, Errors: 0, Skipped: 0 — BUILD SUCCESS

Minor observation (pre-existing, does not block)

The Javadoc on both methods still says "This method is called recursively to support multi-turn tool calling." That description became stale when a48dc60af ("Remove internal tool execution from AnthropicChatModel") delegated recursive tool-call handling to ToolCallingManager. The PR does not need to fix this, but a follow-up Javadoc polish would improve accuracy.

Commit message

Follows the project convention correctly: uppercase verb, no fix: prefix, no PR number in the title, wrapped description, Fixes #6700, and Signed-off-by for DCO.

No blocking code-review findings. Ready to merge after maintainer triage.

[kuntal1461]

Looks right. The only caller is stream() on line 283, and since the class is final there's no subclass path either — matches what Ollama, Mistral, DeepSeek, and the rest all do.

Nit (pre-existing, not a blocker): the Javadoc still says "called recursively to support multi-turn tool calling" — that's been stale since a48dc60af moved the recursive loop into ToolCallingManager. Worth a follow-up cleanup but doesn't need to hold up this PR.

[kuntal1461]

Same story — only call() on line 262 uses this. Good to have both locked down.

Nit: same stale "called recursively" Javadoc as above.

[kuntal1461]

Looks good to me. When the Anthropic module was rewritten to use the official Java SDK (b0ce25aed), these two methods accidentally came back as public, undoing the earlier fix from da3bec4a8. Every other chat model already has them private and the upgrade notes say they should be, so this is clearly the right thing to do.

Checked both call sites — nothing outside this class calls either method, no reflection usage, no interface override. Class is final so subclassing isn't a concern either.

Ran ./mvnw -pl models/spring-ai-anthropic test locally — 124 tests, all green.

One pre-existing nit on the Javadoc: both methods still mention "called recursively" which stopped being true after a48dc60af. Not a blocker, just worth tidying up at some point.