What does this PR do?

Fixes the credential-loss half of #23286 for Anthropic-wire providers like DeepSeek's /anthropic endpoint.

When a streaming retry is triggered after a transient stream drop, the retry cleanup path currently rebuilds the shared OpenAI client unconditionally. In api_mode == "anthropic_messages", that is the wrong transport: these providers stream through self._anthropic_client, and third-party Anthropic endpoints often leave self._client_kwargs empty. The result is a misleading Missing credentials rebuild failure during retry cleanup.

This change routes stream-retry cleanup through the Anthropic client when the active transport is anthropic_messages, so retries reset the correct client instead of trying to rebuild an unused OpenAI one.

It does not claim to solve DeepSeek's upstream ~600s server-side stream limit; it only fixes the retry-side credential loss / wrong-client rebuild.

Related Issue

Related to #23286

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• Added AIAgent._refresh_primary_stream_retry_client() in /run_agent.py to branch retry cleanup by active transport.
• Reused the existing Anthropic client rebuild path for api_mode == "anthropic_messages" instead of calling _replace_primary_openai_client().
• Updated both transient stream retry cleanup sites (stream_retry_pool_cleanup and stream_mid_tool_retry_pool_cleanup) to use the new helper.
• Added a regression test in /tests/run_agent/test_streaming.py that proves Anthropic stream retries rebuild the Anthropic client and do not try to rebuild the OpenAI client.

How to Test

1. Run:
   uv run --frozen pytest -q -o addopts='' tests/run_agent/test_streaming.py -k 'AnthropicStreamRetryCleanup or test_anthropic_stream_refreshes_activity_on_every_event'
2. Run:
   uv run --frozen ruff check run_agent.py tests/run_agent/test_streaming.py
3. Optionally reproduce the original issue on a DeepSeek Anthropic endpoint and confirm stream retry cleanup no longer logs Failed to rebuild shared OpenAI client ... Missing credentials.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Screenshots / Logs

Targeted regression test:

• 2 passed, 33 deselected

Repo smoke:

• uv run --frozen pytest -q -o addopts='' --collect-only tests/run_agent/test_streaming.py
• 35 tests collected