wehub-resource-sync
77 lines
3.0 KiB
YAML
77 lines
3.0 KiB
YAML
rag_search_task:
|
|
description: >
|
|
Search for information relevant to: {query}. Use query='{query}'. Search through documents that are already loaded in the vector database.
|
|
expected_output: >
|
|
JSON formatted response with status, answer, citations, and confidence score
|
|
agent: rag_agent
|
|
|
|
memory_retrieval_task:
|
|
description: >
|
|
Retrieve relevant conversation history and context for: {query}
|
|
expected_output: >
|
|
JSON formatted response with memory context and relevance assessment
|
|
agent: memory_agent
|
|
|
|
web_search_task:
|
|
description: >
|
|
Search the web for recent information and developments related to: {query}
|
|
expected_output: >
|
|
JSON formatted response with web search results and relevance assessment
|
|
agent: web_search_agent
|
|
|
|
arxiv_search_task:
|
|
description: >
|
|
Search ArXiv for academic papers related to: {query}. Find relevant research papers, authors, and recent developments in the field.
|
|
expected_output: >
|
|
JSON formatted response with ArXiv search results including paper titles, authors, abstracts, and publication details
|
|
agent: arxiv_agent
|
|
|
|
context_evaluation_task:
|
|
description: >
|
|
Evaluate the relevance of the following context sources for the query: "{query}"
|
|
|
|
Context Sources:
|
|
1. RAG Result: {rag_result}
|
|
2. Memory Result: {memory_result}
|
|
3. Web Search Result: {web_result}
|
|
4. Tool Result (ArXiv): {tool_result}
|
|
|
|
For each source, determine:
|
|
- Is it relevant to answering the query? (yes/no)
|
|
- Confidence score of relevance (0-1)
|
|
- Key information that should be included in the final response
|
|
- What information should be filtered out as irrelevant
|
|
|
|
Note: If a source has ERROR status, it should not be included in relevant_sources, but mention it in the reasoning.
|
|
|
|
Return only the relevant context that should be used for generating the final response.
|
|
|
|
IMPORTANT: Your response must strictly follow the provided JSON schema structure.
|
|
expected_output: >
|
|
JSON response with:
|
|
- relevant_sources: list of source names that are relevant (e.g., ['RAG', 'Web', 'Memory', 'ArXiv'])
|
|
- filtered_context: dictionary with only relevant information from each source
|
|
- relevance_scores: confidence scores (0-1) for each source's relevance
|
|
- reasoning: brief explanation of filtering decisions
|
|
|
|
The response must be valid JSON that conforms to the ContextEvaluationResult schema.
|
|
agent: evaluator_agent
|
|
|
|
synthesis_task:
|
|
description: >
|
|
Create a comprehensive, coherent response to the query: "{query}"
|
|
|
|
Use the following filtered and relevant context:
|
|
{filtered_context}
|
|
|
|
Guidelines:
|
|
- Synthesize information from multiple sources into a coherent narrative
|
|
- Cite sources appropriately
|
|
- Maintain accuracy and don't add information not present in the context
|
|
- Structure the response clearly and logically
|
|
expected_output: >
|
|
Well-structured response with:
|
|
- Clear answer to the user's query
|
|
- Proper citations from all relevant sources
|
|
agent: synthesizer_agent
|