Replaced synchronous requests with aiohttp to enable non-blocking, async HTTP requests. Updated `smd_detail_article` and `smd_research` to utilize asyncio for parallel execution, significantly improving efficiency for handling multiple API calls. Added aiohttp and asyncio dependencies to requirements.txt.
Removed the unused `smd_detail_article` decorator and expanded the response structure to include detailed article data in nested JSON. Adjusted the pagination default `pageSize` from 10 to 25 to handle larger datasets effectively.
Revised the API endpoint URL from smd.ch to swissdox.ch for accuracy and updated the function docstring to clarify the purpose of the article_id parameter. These changes improve clarity and ensure API requests are routed correctly.
Clarify the usage instructions and context for the smd_detail_article function. This ensures better understanding of its relationship with the smd_research tool.
The `onlyResults` parameter was changed from `True` to `False` in the API request configuration. This ensures more comprehensive data retrieval by including additional metadata in the API response.
Remove unnecessary SMD utilities and shift query creation logic directly into `smd_research` for simplicity. This eliminates duplicate functionality and reduces reliance on external utility files, consolidating responsibilities within server implementation.
Introduce an MCP server for GPT Researcher using FastMCP, enabling AI assistants to perform web research and generate reports via SSE. Includes async processing steps with real-time updates and a runnable server script.
Updated `write_report` and `get_research_sources` to stream results via server-sent events (SSE) using `AsyncGenerator`. This allows incremental data delivery and better error reporting, improving client-server communication.
Updated `quick_search` to return an `AsyncGenerator` for streaming server-sent events (SSE) during quick research execution. Enhanced error handling to provide detailed feedback and upgraded `gpt-researcher` dependency to version 0.13.3 for compatibility.
Replaced various custom SSE event types like "tool_update", "tool_result", and "tool_error" with the unified "message" event type. This simplifies event handling logic and ensures consistency across all server-sent event flows.
Updated the `deep_research` function to use asynchronous generators and send Server-Sent Events (SSE) for real-time streaming updates. Added a new utility for formatting SSE events and improved research lifecycle visibility with intermediate progress steps and error handling.
Updated the FastMCP initialization to explicitly bind to 0.0.0.0, set the port to 8000, and specify a keep-alive timeout of 720 seconds. This ensures the server is properly configured for broader accessibility and stability.
Add a `research://{topic}` resource endpoint for direct access to research context, reducing redundant API calls. Introduced `research_store` for caching research results and modularized helper methods like `store_research_results` and `format_context_with_sources` for better reusability and clarity. Refactored existing researcher initialization for simplicity and improved comments to clarify intended usage.
Added the `timeout_keep_alive` parameter to FastMCP initialization in both GPT Researcher and SMD Researcher servers. This ensures connections remain active for longer, improving server reliability for long-lived client interactions.
The `host` and `port` parameters were added to FastMCP initialization for both GPT Researcher and SMD servers. This ensures the servers are accessible on all network interfaces at port 8000.
Changed the APP_ENTRYPOINT environment variable to reflect the updated directory structure. This ensures the application points to the correct script path during execution.
Switched CMD to use shell form instead of exec form for running the application. This simplifies the command syntax and ensures compatibility in environments where exec form parsing may cause issues.
Implemented the SMD MCP Server with functionality for AI-driven research and article retrieval via the MCP protocol. Added utilities for prompt creation and enhanced dependency configuration. This serves as a foundation for handling SMD-related queries and data processing.
Eliminated the `research://{topic}` resource API, associated utilities, and the `research_store`. These components were redundant due to existing alternatives using the `conduct_research` tool. This cleanup reduces complexity and improves maintainability.
Reorganized module directories under `phoenix_technologies` for better namespace clarity and maintainability. Updated the Dockerfile to use an environment variable for the application entry point, enhancing flexibility in deployment. Additionally, revamped the README to reflect the new structure and provide clearer project documentation.
Simplified the descriptions in `deep_research` and `quick_search` docstrings to improve readability and eliminate redundant details. This ensures concise and focused explanations of each function's purpose.
Introduced a CustomLogsHandler class to handle and log JSON data during research or search operations. Updated GPTResearcher initialization to include the CustomLogsHandler for improved logging and debugging.
Updated the `mcp.run` call to include the "sse" mode, ensuring proper handling of server events. This change improves the server's event loop management and prepares it for SSE-specific operations.
Introduced the quick_search function for faster, snippet-based searches and adjusted mcp.run to remove the "sse" argument. Additionally, modified the deep_research docstring to remove the citation requirement note for simpler usage documentation.
Added `RESEARCH_TYPE` environment variable to configure the default research type for GPTResearcher. Updated the initialization of GPTResearcher to use this dynamic configuration, improving flexibility and adaptability.
This update enables specifying the report type during researcher initialization. It introduces "deep" as the default report type, enhancing flexibility for future use cases.
Replaced FastAPI app with an MCP server implementation, enhancing flexibility and modularity for research operations. Deprecated `phoenix_technologies` package, updated server logic, added utility functions, and revised dependencies in `requirements.txt`. Updated Dockerfile and README to align with the new architecture.
Replaced custom log handler and async report generation logic with a simplified fake data streamer for the StreamingResponse. Added uvicorn server startup code for direct script execution.
Simplified the logic for log streaming by consolidating into a single async generator (`log_stream`). Removed redundant tasks and streamlined report generation to improve code readability and maintainability.
Enhanced report generation by integrating `CustomLogsHandler` and separating tasks for generating reports and streaming logs. Replaced the previous implementation with a unified async generator to handle concurrent execution and improve error handling. Updated module exports to include the new `CustomLogsHandler` component.
Replaced `generator.researcher.complete` with the correct `generator.complete`. This resolves a potential logic error and ensures the completion condition is properly evaluated.
Replaced `.is_complete()` method calls with direct `.complete` attribute access to streamline the code. Removed the redundant `is_complete()` method from `deepresearch.py` to reduce unnecessary indirection. This simplifies the logic and improves readability.
The is_complete method was incorrectly referenced without parentheses, causing a potential runtime error. This fix ensures proper method invocation and prevents the loop from behaving incorrectly.
Modified the report generation process to stream logs incrementally during execution. Introduced new methods `init` and `is_complete` to enhance modularity and track completion status. Adjusted generator logic to better manage and yield logs in real-time.
This change introduces status updates to indicate the start and successful completion of the report generation. These progress messages improve user feedback during the asynchronous operation.
Removed the unnecessary `logs` initialization and streamlined the `CustomLogsHandler` setup. This simplifies the code and avoids redundant log storage.
Changed `send_json` to return `None` instead of yielding and updated log retrieval to use `yield` in place of `return`. These modifications align function behavior with intended asynchronous use cases and improve consistency.
Removed unused log yielding in `main.py` to simplify the flow. Updated `send_json` in `deepresearch.py` to use an async generator for more flexible streaming of log data.
The `generate_report` method is now explicitly called within the main process, ensuring the researcher's report generation is initiated properly. Updated its return type annotation for clarity in `deepresearch.py`.
Refactored `CustomLogsHandler` to accept an optional logs list and modified its method to append data instead of yielding. Enhanced `ReportGenerator` with proper asynchronous iterator implementation, enabling smooth log handling and retrieval. Simplified variable naming for better readability.
Introduced `CustomLogsHandler` to manage log handling and WebSocket integration in `ReportGenerator`. Simplified and restructured report generation logic for improved maintainability. Removed obsolete methods and enhanced overall readability with cleaner code structure.
Enhanced docstrings for better clarity and consistency across methods. Added error handling for report generation and ensured default values for research details to avoid null errors. Improved comments and logic flow for readability and maintainability.
Added asynchronous iteration methods (`__aiter__` and `__anext__`) to the `DeepResearch` class to enable streaming of research report chunks. Introduced functionality to generate and split reports into smaller chunks for improved report handling.
Included new environment variable placeholders for better configuration guidance. These updates clarify setup requirements and improve documentation completeness.
Implement a modular FastAPI-based service for generating research reports using `GPTResearcher`. Includes secure API key authentication, a streaming response endpoint, and a Dockerized deployment setup. Also adds documentation, core dependencies, and project structure.
