feat(agent): add long-term memory extraction and visualization

[newcat22]

Summary

• Add LLM-powered memory extraction that compresses agent execution steps into human-readable Markdown documents with problem/conclusion pairs
• Support manual and periodic triggers via factory pattern, with context compression when messages exceed 20
• Add trae-cli memory subcommand for offline extraction from trajectory JSON files, and --extract-memory flag on run command
• Add console visualization for memory documents in SimpleCLIConsole

New files

• trae_agent/utils/long_term_memory.py — Core module: LLM extraction, Markdown rendering (MemoryDocument with to_markdown/from_markdown round-trip), memory message builder
• trae_agent/utils/memory_trigger.py — Trigger factory: ManualMemoryTrigger, PeriodicMemoryTrigger, create_memory_trigger() factory
• tests/utils/test_long_term_memory.py — 22 test cases covering config, triggers, document round-trip, memory injection, and end-to-end extraction

Modified files

• trae_agent/utils/config.py — LongTermMemoryConfig dataclass + YAML parsing
• trae_agent/agent/base_agent.py — Memory trigger integration in execute_task(), context compression in _run_llm_step(), inject_memory_into_messages(), extract_memory_now()
• trae_agent/agent/agent.py — Wire up LongTermMemory creation and set_task()
• trae_agent/cli.py — memory subcommand + --extract-memory flag
• trae_agent/utils/cli/cli_console.py — set_memory_doc() base method
• trae_agent/utils/cli/simple_console.py — Memory summary visualization with Rich Panels
• trae_config.yaml.example — long_term_memory config examples

Test plan

• All 84 existing + new tests pass
• Ruff lint check passes
• YAML config parsing for long_term_memory section
• MemoryDocument to_markdown/from_markdown round-trip
• ManualMemoryTrigger never fires, PeriodicMemoryTrigger fires at interval
• create_memory_trigger() factory returns correct types
• inject_memory_into_messages preserves system + memory + recent messages
• extract_memory() with mock LLM produces correct sections
• extract_and_save() creates Markdown file with correct content

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}