Linux philosophy optimization: complete project refactoring

[Copilot]

Applies the recommendations from docs/linux_philosophy_optimization.md — a systematic review of xAgent through the lens of Unix/Linux design principles. Changes span security hardening, dependency hygiene, observability decoupling, and filling in placeholder functionality.

🔴 High Priority

• H1 — Logging pollution: Removed logging.basicConfig() from agent.py. Library code must never configure the root logger.
• H2 — CORS security: AgentHTTPServer now emits SecurityWarning on allow_origins=["*"]; origins are configurable via server.cors.allow_origins in config.
• H3 — Typo fix: Renamed upstach_vector_store.py → upstash_vector_store.py; old file kept as a one-line backward-compat re-export shim.
• H4 — Structured error logging: Background memory task permanent failures now include structured extra fields (task_description, error, attempts); max_iter exhaustion returns a diagnostic message instead of a generic one.
• H5 — DoS prevention: AgentInput HTTP fields now use pydantic.Field(ge=..., le=...) bounds (e.g. max_iter capped at 50) sourced from defaults.py.

🟡 Medium Priority

• M1+M2 — Optional dependencies: langfuse and chromadb moved from required to optional in pyproject.toml. New extras: local, cloud, observability, dev, all.

• M1 — Observability module: New xagent/observability/__init__.py provides @observe decorator and get_openai_client() — both transparent no-ops when Langfuse is absent.

  # agent.py — no longer imports langfuse directly
  from ..observability import observe, get_openai_client
  
  @observe
  async def chat(self, ...): ...
  self.client = get_openai_client(client)  # wraps with Langfuse if installed

• M5 — Preview length: TOOL_RESULT_PREVIEW_LENGTH increased from 20 → 200 characters.

• M8 — Magic numbers: New xagent/defaults.py is the single source of truth for all constants (timeouts, limits, TTLs); AgentConfig references these instead of inline literals.

🟢 Low Priority

• L1 — Swarm implementation: Replaced the empty Swarm placeholder with a working implementation using Agent.as_tool(). Specialist agents register as tools on a coordinator; Swarm is now exported from the top-level xagent package.

  swarm = Swarm(agents=[researcher, writer])
  result = await swarm.run(task="Research and summarize quantum computing.")

• L2 — YAML DSL: Added parse_dependencies_yaml() to workflow_dsl.py. GraphWorkflow now auto-detects YAML-format vs arrow-DSL dependency strings.

  dependencies:
    analyzer:
      - researcher
    writer:
      - analyzer

Tests

Added tests/test_refactoring.py with 20 tests covering all the above. All 37 tests pass; CodeQL reports 0 alerts.

Original prompt

基于 linux_philosophy_optimization.md
完成项目的完整重构

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.