AI Journalist for Government Meeting Analysis
This challenge focuses on building an advanced AI-powered journalistic assistant. Participants will create a graph-based agent workflow using LangGraph to analyze raw video transcripts of local government meetings. The system, powered by Claude Sonnet 4, will identify key speakers, extract discussion topics, summarize decisions, and generate actionable journalistic insights. It will feature RAG over meeting data and external public records, with MCP-enabled tool integration for accessing transcription services and structured databases. Extended thinking will be applied to synthesize complex information and identify potential stories.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop a graph-based AI journalist using LangGraph and Claude Sonnet 4 for analyzing government meeting transcripts with MCP tool integration and RAG.
How work is evaluated
The evaluation will focus on the accuracy of information extraction, the quality of summaries and journalistic insights, and the efficiency of the LangGraph workflow. It will assess the system's ability to identify key events, speakers, and decisions from meeting transcripts and provide relevant contextual information.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Design and implement a graph-based agent workflow using LangGraph for multi-stage analysis
Utilize Claude Sonnet 4 for advanced NLP tasks like summarization, entity extraction, and insight generation
Develop tools for transcribing videos and querying external data sources
Integrate an MCP-enabled RAG system with meeting transcripts and public records
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The evaluation will focus on the accuracy of information extraction, the quality of summaries and journalistic insights, and the efficiency of the LangGraph workflow. It will assess the system's ability to identify ke...
Challenge input
{'raw_transcript': 'string', 'public_records_rag_query': 'string'}
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{'extracted_topics': ['string'], 'key_speakers': ['string'], 'decisions_summarized': ['string'], 'journalistic_insights': [{'title': 'string', 'sum...
- Successfully transcribes the video and integrates the transcript into the workflow.
- At least 3 journalistic insights are generated, and they are relevant and coherent.
- InformationExtractionAccuracy target: 0.8
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Successfully transcribes the video and integrates the transcript into the workflow.
- At least 3 journalistic insights are generated, and they are relevant and coherent.
Metrics
- InformationExtractionAccuracy · Target 0.8
- WorkflowExecutionTime · Target 120
Workflow tasks
- Task 1
Input: {'raw_transcript': 'string', 'public_records_rag_query': 'string'}
Output: {'extracted_topics': ['string'], 'key_speakers': ['string'], 'decisions_summarized': ['string'], 'journalistic_insights': [{'title': 'string', 'summary': 'string', 'relevance_score': 'float'}]}
Execution harness: Python
- Task 2
Input: {'video_url': 'string', 'query_type': 'string'}
Output: {'transcription_status': 'string', 'rag_tool_calls_count': 'integer', 'total_workflow_time_s': 'float'}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.