mirror of
https://github.com/agentscope-ai/ReMe.git
synced 2026-09-23 00:43:18 +00:00
d0c9d89092
* feat(memory): add ContextChecker component for context size management * refactor(memory): restructure file-based memory tools and update imports * docs(readme): update documentation with detailed architecture and components * docs(readme): update Chinese documentation with enhanced memory management diagrams * refactor(cookbook): move cookbook files to test directory and clean up docs * docs(readme): update link path for old version documentation * docs(readme): update documentation with improved architecture diagrams and component details * docs(readme): update documentation with improved clarity and structure * refactor(docs): update in-memory memory documentation * docs(readme): add experiment reproduction link to quickstart guide
1 line
No EOL
6.4 KiB
JSON
1 line
No EOL
6.4 KiB
JSON
[{"workspace_id": "test_workspace", "memory_id": "5a8480e36cf84d5d8b8d4acfad78f156", "memory_type": "task", "when_to_use": "When analyzing a large, multi-faceted company with global operations and diverse product lines, especially when the goal is to generate a comprehensive, structured overview.", "content": "The agent successfully decomposed the analysis of Xiaomi Corporation into four distinct subtasks\u2014company overview, financial performance, product portfolio/innovation, and market position\u2014each addressed with targeted web searches. This systematic breakdown ensured coverage of all critical dimensions without redundancy. By using multiple search queries with varying focus (e.g., 'financial performance 2024' vs. 'market position in smartphones'), the agent obtained granular, up-to-date insights from different angles, including forward-looking data (Q1 2025) and historical benchmarks (2024). The synthesis of results into a well-organized, categorized report demonstrated strong integration of external information into a coherent narrative.", "score": 0.92, "time_created": "2025-09-07 00:02:29", "time_modified": "2025-09-07 00:02:29", "author": "qwen3-30b-a3b-instruct-2507", "metadata": {"when_to_use": "When analyzing a large, multi-faceted company with global operations and diverse product lines, especially when the goal is to generate a comprehensive, structured overview.", "experience": "The agent successfully decomposed the analysis of Xiaomi Corporation into four distinct subtasks\u2014company overview, financial performance, product portfolio/innovation, and market position\u2014each addressed with targeted web searches. This systematic breakdown ensured coverage of all critical dimensions without redundancy. By using multiple search queries with varying focus (e.g., 'financial performance 2024' vs. 'market position in smartphones'), the agent obtained granular, up-to-date insights from different angles, including forward-looking data (Q1 2025) and historical benchmarks (2024). The synthesis of results into a well-organized, categorized report demonstrated strong integration of external information into a coherent narrative.", "tags": ["company analysis", "multi-dimensional research", "structured decomposition", "financial performance", "market position", "product ecosystem"], "confidence": 0.9, "step_type": "reasoning", "tools_used": ["web_search"]}}, {"workspace_id": "test_workspace", "memory_id": "bee8011f546e449899f85a0213fce340", "memory_type": "task", "when_to_use": "When initial search results are incomplete or inconsistent, particularly when key metrics span across different time periods or reports (e.g., 2024 vs. 2025 data).", "content": "The agent proactively used the same tool (web_search) multiple times with different parameters to gather information from various perspectives\u2014such as financials, market share, product innovation, and company history. This iterative approach allowed the agent to cross-validate facts (e.g., confirming 2024 smartphone revenue and Q1 2025 growth), identify temporal inconsistencies (e.g., Q1 2025 data being reported in 2025 but relevant for 2024 context), and fill knowledge gaps. The ability to distinguish between forward-looking and retrospective data enhanced accuracy and contextual relevance.", "score": 0.85, "time_created": "2025-09-07 00:02:29", "time_modified": "2025-09-07 00:02:29", "author": "qwen3-30b-a3b-instruct-2507", "metadata": {"when_to_use": "When initial search results are incomplete or inconsistent, particularly when key metrics span across different time periods or reports (e.g., 2024 vs. 2025 data).", "experience": "The agent proactively used the same tool (web_search) multiple times with different parameters to gather information from various perspectives\u2014such as financials, market share, product innovation, and company history. This iterative approach allowed the agent to cross-validate facts (e.g., confirming 2024 smartphone revenue and Q1 2025 growth), identify temporal inconsistencies (e.g., Q1 2025 data being reported in 2025 but relevant for 2024 context), and fill knowledge gaps. The ability to distinguish between forward-looking and retrospective data enhanced accuracy and contextual relevance.", "tags": ["iterative search", "data validation", "temporal consistency", "cross-referencing", "information triangulation"], "confidence": 0.85, "step_type": "action", "tools_used": ["web_search"]}}, {"workspace_id": "test_workspace", "memory_id": "10d95a6544a14a979dc63d7d71f5cf1a", "memory_type": "task", "when_to_use": "When synthesizing information from fragmented sources into a unified, high-quality output that meets professional or strategic decision-making standards.", "content": "After collecting raw data from multiple search results, the agent applied a clear categorization framework\u2014Company Overview, Market Position, Financials, Product Portfolio, Recognition\u2014to structure the final response. This not only improved readability but also ensured no critical insight was overlooked. The agent correctly inferred missing values (e.g., completing the 'nearly X% of revenue from IoT' statement based on context), prioritized recent and verified data, and highlighted trends such as premium segment growth and ecosystem expansion. This demonstrates effective transformation of raw data into actionable intelligence.", "score": 0.85, "time_created": "2025-09-07 00:02:29", "time_modified": "2025-09-07 00:02:29", "author": "qwen3-30b-a3b-instruct-2507", "metadata": {"when_to_use": "When synthesizing information from fragmented sources into a unified, high-quality output that meets professional or strategic decision-making standards.", "experience": "After collecting raw data from multiple search results, the agent applied a clear categorization framework\u2014Company Overview, Market Position, Financials, Product Portfolio, Recognition\u2014to structure the final response. This not only improved readability but also ensured no critical insight was overlooked. The agent correctly inferred missing values (e.g., completing the 'nearly X% of revenue from IoT' statement based on context), prioritized recent and verified data, and highlighted trends such as premium segment growth and ecosystem expansion. This demonstrates effective transformation of raw data into actionable intelligence.", "tags": ["synthesis", "information structuring", "narrative coherence", "trend identification", "knowledge integration"], "confidence": 0.9, "step_type": "observation", "tools_used": ["web_search"]}}] |