feat(python/langgraph): upgrade langgraph agent to a2a-sdk 1.1+ (protocol v1.0)

[1Wizzy]

Description

Thank you for opening a Pull Request!
Before submitting your PR, there are a few things you can do to make sure it goes smoothly:

• Follow the CONTRIBUTING Guide.

Fixes # 🦕

---

Summary of Changes

This PR upgrades the LangGraph Currency Agent sample (samples/python/agents/langgraph) from legacy a2a-sdk 0.3.x to the official stable a2a-sdk >= 1.1.0 (A2A Protocol v1.0 specification) and langgraph >= 1.0.0.

1. A2A Protocol v1.0 & LangGraph 1.0 Upgrade

• Dependencies (pyproject.toml, uv.lock):
  • Upgraded to a2a-sdk>=1.1.0 and langgraph>=1.0.0.
• Server Architecture (app/__main__.py):
  • Replaced deprecated A2AStarletteApplication with modular route factories: create_agent_card_routes and create_jsonrpc_routes on Starlette.
  • Declared AgentCard with v1.0 AgentInterface (protocol_binding='JSONRPC', url='http://{host}:{port}/', protocol_version='1.0').
  • Passed agent_card into DefaultRequestHandler.
• Execution & Lifecycle (app/agent_executor.py):
  • Migrated from legacy a2a.utils to a2a.helpers (new_task_from_user_message, new_text_message, new_text_part).
  • Implemented event and status management using a2a.server.tasks.TaskUpdater (start_work(), requires_input(), complete(), add_artifact()).
  • Standardized error handling using InvalidParamsError, InternalError, and UnsupportedOperationError.
• Client & Verification (app/test_client.py):
  • Migrated to await create_client(agent=..., client_config=...).
  • Upgraded messaging requests to SendMessageRequest using protobuf enums (Role.ROLE_USER).
• Documentation (README.md):
  • Modernized CLI run instructions (uv run python -m app).
  • Updated JSON-RPC payload examples to protocol v1.0 standards (SendMessage, SendStreamingMessage, A2A-Version: 1.0, protobuf JSON structures).
  • Updated Mermaid sequence diagram with TASK_STATE_INPUT_REQUIRED.
  • Wrapped disclaimer text to comply with Markdownlint (MD013).

2. Robustness Improvements

• Frankfurter API 301 Redirect: Added follow_redirects=True to httpx.get in app/agent.py to handle upstream 301 redirect to api.frankfurter.dev.
• OpenAI-Compatible LLMs: Added _CompatibleChatOpenAI defaulting with_structured_output to function_calling instead of strict json_schema, ensuring compatibility with non-OpenAI gateways (Ollama, DeepSeek, vLLM).
• Client Timeout: Provided custom httpx.AsyncClient with a 60s timeout to ClientConfig in test_client.py to avoid premature client timeouts during multi-step tool calls.

---

Verification & Testing

• Linting & Formatting:
  • uv run ruff check --config .ruff.toml samples/python/agents/langgraph (Passed, 0 errors).
  • uv run ruff format --check --config .ruff.toml samples/python/agents/langgraph (Passed, clean).
• Type Checking:
  • uv run pyright samples/python/agents/langgraph (Passed, 0 errors, 0 warnings).
• End-to-End Runtime Test:
  • Single-turn execution with artifact generation (conversion_result).
  • Multi-turn interaction with TASK_STATE_INPUT_REQUIRED.
  • Real-time streaming updates (SendStreamingMessage via SSE).

Note on CI Checks: Lint Code Base (super-linter, ruff, markdownlint) is passing. The failing Lint Go check is an upstream issue in samples/go/agents/deepresearch (being resolved in #704) and is unrelated to this Python PR.

[samirbajaj]

@1Wizzy - are we just waiting for the upstream fix in #704 ?
I'd love to see this merged.