perf(agentgateway): parallelize MCP fragment and A2A agent card fetching

[cassiofariasmachado]

Disclaimer: Do not include SAP-internal or customer-specific information in this PR (e.g. internal system URLs, customer names, tenant IDs, or confidential configurations). This is a public repository.

Description

get_mcp_tools_lob and get_agent_cards_lob previously fetched each fragment sequentially with await inside a for loop. With N fragments each taking 3–4 s, total load time scaled as O(N × latency) — ~95 s for 22 fragments.

This PR replaces both sequential loops with asyncio.gather(return_exceptions=True), fetching all fragments concurrently. Wall-clock time drops to the latency of the single slowest fragment (~4 s regardless of fragment count).

Changes:

• get_mcp_tools_lob: two-pass approach — validate URLs into a pending list up front, then asyncio.gather over all of them. Uses isinstance(result, BaseException) (not Exception) to correctly catch CancelledError and other BaseException subclasses returned by gather(return_exceptions=True). The final log now reports len(pending) (fragments that had a valid URL) rather than len(fragments) (all fragments including URL-less ones). Timing is logged: Loaded N MCP tool(s) from M fragment(s) in X.XXs.
• get_agent_cards_lob: same pattern applied — was missed by prior art (issue Parallelize fragment connections inside list_mcp_tools so all fragments are handled concurrently #301 / PR perf(agentgateway): parallelize LoB MCP tool listing across fragments #322). pending_cards collects (fragment_name, url, ord_id) tuples; asyncio.gather fetches all concurrently; errors are isolated per-fragment.

Related Issue

Closes #301

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update
• Code refactoring
• Dependency update

Actually: performance improvement (non-breaking)

How to Test

python -m pytest tests/agentgateway/unit/test_lob.py -s --log-cli-level=INFO

New tests verify:

1. All fragments are dispatched concurrently (test_fragments_fetched_concurrently) — uses asyncio.sleep to confirm all N tasks start before any finish
2. Per-fragment failure isolation — one failing fragment does not abort the others
3. Fragment count in log excludes URL-less fragments

Checklist

• I have read the Contributing Guidelines
• I have verified that my changes solve the issue
• I have added/updated automated tests to cover my changes
• All tests pass locally
• I have verified that my code follows the Code Guidelines
• I have updated documentation (if applicable)
• I have added type hints for all public APIs
• My code does not contain sensitive information (credentials, tokens, etc.)
• I have followed Conventional Commits for commit messages

Additional Notes

PR #322 addressed get_mcp_tools_lob only. This PR also fixes get_agent_cards_lob with the same pattern, and adds a timing log line to make the improvement observable in production logs.

[NicoleMGomes]

Missing user-guide and version update

[NicoleMGomes]

Great work!!