{"repo": "billy-liu-dut/g-frame", "files": {"deep_research/config.yaml": {"2ede9623fe34f1e0311659690b23b28329404e31": {"t": 1785745218.5289984, "n": 3932, "h": "6069d00985c4b880", "c": "# ------------------ API Keys and Endpoints ------------------\n# IMPORTANT: Fill in your API keys here.\napi_keys:\n  clarivate_wos: \"YOUR_CLARIVATE_WOS_API_KEY\"\n  deepseek_api_key: \"YOUR_DEEPSEEK_API_KEY\"\n  qwen_api_key: \"YOUR_QWEN_API_KEY\"\n\nllm_endpoints:\n  deepseek_endpoint: \"https://ark.cn-beijing.volces.com/api/v3\"\n  qwen_endpoint: \"https://dashscope.aliyuncs.com/compatible-mode/v1\"\n\n# ------------------ Model Configuration ------------------\n# Select which LLM to use for the agent's tasks.\n# Options: \"Qwen7B\", \"DeepSeekR1\"\nactive_model: \"Qwen7B\"\n\nmodels:\n  Qwen7B:\n    nickname: \"Qwen7B\"\n    model_name: \"qwen2.5-7b-instruct\"\n  DeepSeekR1:\n    nickname: \"DeepSeek\"\n    model_name: \"ep-20250218211825-q28hr\"\n\n# ------------------ File Paths ------------------\n# All output files will be saved relative to the project root.\npaths:\n  results_dir: \"./results\"\n\n# ------------------ Web of Science Search Parameters ------------------\nwos_search_params:\n  database: \"WOS\" # Web of Science Core Collection\n  # Number of articles to retrieve per search query.\n  limit_per_query: 10\n  # How to sort the articles. TC+D = Times Cited (Descending)\n  sort_field: \"TC+D\"\n\n# ------------------ Prompt Templates ------------------\n# You can customize the prompts sent to the LLM for each step.\nprompts:\n  article_summary: |\n    You are a seasoned researcher skilled in summarizing literature.\n    Analyze the following article content and extract key elements including:\n    - Study significance\n    - Background information\n    - Research objectives\n    - Methodology\n    - Mechanistic explanations\n    - Methodological innovations\n    Article content:\n    {content}\n\n  outline_generation: |\n    You are a senior professor drafting a literature review. Based on the user's request and summarized content,\n    create a structured research report with standard academic sections including:\n    - Research Background & Significance\n    - Common Methodologies\n    - Key Mechanisms and Findings\n    - Performance Optimization Strategies\n    - Future Directions and Challenges\n    User's requirement: {research_topic}\n    Summarized content:\n    {summary_content}\n\n  reflection_and_expansion: |\n    You are a research advisor identifying knowledge gaps. Analyze the outline to find the most important subtopic that requires deeper investigation.\n    Requirements:\n    - Provide a specific English research term.\n    - It must be directly related to the core research problem.\n    - It should be suitable for a Web of Science API query.\n    - It must be more focused than the initial keyword.\n    Example: \"narrowband deep-blue TADF emitters\"\n    Core research problem: {research_topic}\n    Current outline: {current_outline}\n    Return ONLY the most important keyword in this format: \"keyword\"\n\n  expand_section: |\n    Your task is to expand a specific section of a research report by synthesizing the provided reference materials.\n    Follow these requirements:\n    1. Write in a formal, academic tone using technical terminology.\n    2. Synthesize the information, do not just list summaries.\n    3. Include mathematical formulas using LaTeX notation ($equation$ or $$equation$$) where appropriate.\n    Core research problem: {research_topic}\n    Section to develop: {section_title}\n    Reference materials:\n    {summary_content}\n\n  report_polishing: |\n    You are a professional academic editor. Your task is to rewrite the following draft report into a comprehensive and polished review article.\n    Requirements:\n    1. Preserve all original DOI citations.\n    2. Write entirely in academic English.\n    3. Follow a standard review structure (Introduction, Core Concepts, Methodologies, Applications, Challenges, Conclusion).\n    4. Ensure logical flow and coherence between sections.\n    5. Format equations with LaTeX notation.\n    Report to polish:\n    {report_content}\n    Output ONLY the polished academic review in Markdown format."}, "fd4483938587cef7ab313ea4096f9bb343df7bf0": {"t": 1785748834.9406688, "n": 3932, "h": "6069d00985c4b880", "c": "# ------------------ API Keys and Endpoints ------------------\n# IMPORTANT: Fill in your API keys here.\napi_keys:\n  clarivate_wos: \"YOUR_CLARIVATE_WOS_API_KEY\"\n  deepseek_api_key: \"YOUR_DEEPSEEK_API_KEY\"\n  qwen_api_key: \"YOUR_QWEN_API_KEY\"\n\nllm_endpoints:\n  deepseek_endpoint: \"https://ark.cn-beijing.volces.com/api/v3\"\n  qwen_endpoint: \"https://dashscope.aliyuncs.com/compatible-mode/v1\"\n\n# ------------------ Model Configuration ------------------\n# Select which LLM to use for the agent's tasks.\n# Options: \"Qwen7B\", \"DeepSeekR1\"\nactive_model: \"Qwen7B\"\n\nmodels:\n  Qwen7B:\n    nickname: \"Qwen7B\"\n    model_name: \"qwen2.5-7b-instruct\"\n  DeepSeekR1:\n    nickname: \"DeepSeek\"\n    model_name: \"ep-20250218211825-q28hr\"\n\n# ------------------ File Paths ------------------\n# All output files will be saved relative to the project root.\npaths:\n  results_dir: \"./results\"\n\n# ------------------ Web of Science Search Parameters ------------------\nwos_search_params:\n  database: \"WOS\" # Web of Science Core Collection\n  # Number of articles to retrieve per search query.\n  limit_per_query: 10\n  # How to sort the articles. TC+D = Times Cited (Descending)\n  sort_field: \"TC+D\"\n\n# ------------------ Prompt Templates ------------------\n# You can customize the prompts sent to the LLM for each step.\nprompts:\n  article_summary: |\n    You are a seasoned researcher skilled in summarizing literature.\n    Analyze the following article content and extract key elements including:\n    - Study significance\n    - Background information\n    - Research objectives\n    - Methodology\n    - Mechanistic explanations\n    - Methodological innovations\n    Article content:\n    {content}\n\n  outline_generation: |\n    You are a senior professor drafting a literature review. Based on the user's request and summarized content,\n    create a structured research report with standard academic sections including:\n    - Research Background & Significance\n    - Common Methodologies\n    - Key Mechanisms and Findings\n    - Performance Optimization Strategies\n    - Future Directions and Challenges\n    User's requirement: {research_topic}\n    Summarized content:\n    {summary_content}\n\n  reflection_and_expansion: |\n    You are a research advisor identifying knowledge gaps. Analyze the outline to find the most important subtopic that requires deeper investigation.\n    Requirements:\n    - Provide a specific English research term.\n    - It must be directly related to the core research problem.\n    - It should be suitable for a Web of Science API query.\n    - It must be more focused than the initial keyword.\n    Example: \"narrowband deep-blue TADF emitters\"\n    Core research problem: {research_topic}\n    Current outline: {current_outline}\n    Return ONLY the most important keyword in this format: \"keyword\"\n\n  expand_section: |\n    Your task is to expand a specific section of a research report by synthesizing the provided reference materials.\n    Follow these requirements:\n    1. Write in a formal, academic tone using technical terminology.\n    2. Synthesize the information, do not just list summaries.\n    3. Include mathematical formulas using LaTeX notation ($equation$ or $$equation$$) where appropriate.\n    Core research problem: {research_topic}\n    Section to develop: {section_title}\n    Reference materials:\n    {summary_content}\n\n  report_polishing: |\n    You are a professional academic editor. Your task is to rewrite the following draft report into a comprehensive and polished review article.\n    Requirements:\n    1. Preserve all original DOI citations.\n    2. Write entirely in academic English.\n    3. Follow a standard review structure (Introduction, Core Concepts, Methodologies, Applications, Challenges, Conclusion).\n    4. Ensure logical flow and coherence between sections.\n    5. Format equations with LaTeX notation.\n    Report to polish:\n    {report_content}\n    Output ONLY the polished academic review in Markdown format."}}}}