- Added custom scrollbar styles for improved aesthetics and usability across the interface. - Implemented responsive adjustments for chat sidebar and input components to enhance mobile experience. - Integrated tooltip components in select elements for better user guidance and information display. - Updated chat input and message components for improved styling and functionality. - Enhanced settings page with semantic query expansion options and improved model selection UI. - Refactored sidebar list to include better date formatting and tooltip support for session titles. - Introduced GPU stats monitoring in the system panel for better hardware insights.
128 lines
5.2 KiB
TypeScript
128 lines
5.2 KiB
TypeScript
export const DEEP_RESEARCH_PROMPTS = {
|
|
DECOMPOSITION: `Role: Research Lead specializing in web search optimization.
|
|
Task: Decompose the user's query into 3-5 atomic, search-friendly questions optimized for web search engines.
|
|
|
|
User Query: "{{userQuery}}"
|
|
|
|
## CRITICAL INSTRUCTIONS
|
|
|
|
1. **PRESERVE THE SUBJECT**: If the user asks about "Garuva", ALL queries must mention "Garuva".
|
|
2. **USE SPECIFIC KEYWORDS**: Include relevant technical terms, proper names, locations, dates, or context that will help search engines find precise results.
|
|
3. **AVOID GENERIC QUERIES**: Never create queries like "what is this" or "tell me about" without specific context. Instead, use descriptive terms.
|
|
4. **OPTIMIZE FOR SEARCH ENGINES**:
|
|
- Use terms that are likely to appear in search results
|
|
- Include synonyms or related terms when appropriate
|
|
- For academic/research queries, include terms like "estudo", "pesquisa", "resultado", "premiado"
|
|
- For location queries, include geographic context (state, country, region)
|
|
5. **COVER DIFFERENT ANGLES**: Each query should explore a different aspect or perspective of the original question.
|
|
6. **DO NOT TRANSLATE**: Preserve proper names, technical terms, and specific terminology in their original form.
|
|
7. **OUTPUT FORMAT**: Return ONLY a JSON array of strings. No markdown, no explanations, no additional text.
|
|
|
|
## EXAMPLES
|
|
|
|
Input: "Onde fica Garuva?"
|
|
Output: ["localização garuva sc", "mapa garuva santa catarina", "história cidade garuva"]
|
|
|
|
Input: "quais os ultimos resultados academicos mais premiados?"
|
|
Output: ["resultados academicos premiados 2024", "pesquisas cientificas premiadas recentes", "estudos academicos reconhecidos premios", "descobertas academicas premiadas"]
|
|
|
|
Input: "como funciona inteligencia artificial?"
|
|
Output: ["inteligencia artificial como funciona", "IA machine learning algoritmos", "sistemas inteligencia artificial aplicacoes"]
|
|
|
|
## YOUR TASK
|
|
|
|
Decompose this query into search-optimized sub-queries:
|
|
"{{userQuery}}"
|
|
|
|
Return ONLY a JSON array of strings:`,
|
|
|
|
CONTEXT_CHECK: `Role: Fact Checker.
|
|
Task: Analyze the Knowledge Base below and determine if it contains sufficient information to answer the user's query.
|
|
|
|
## KNOWLEDGE BASE
|
|
{{knowledgeBase}}
|
|
|
|
## USER QUERY
|
|
"{{userQuery}}"
|
|
|
|
## OUTPUT
|
|
1. **Sufficiency**: Does the Knowledge Base contain enough information to answer the query? (YES/NO/PARTIAL)
|
|
2. **Key Facts Found**: List the main facts relevant to the query (bullet points).
|
|
3. **Missing Information**: What information is NOT in the Knowledge Base but would be needed?
|
|
4. **Reliability Assessment**: Rate the overall reliability of sources (HIGH/MEDIUM/LOW).
|
|
|
|
Be concise and factual.`,
|
|
|
|
STRICT_GENERATION: `Role: Rigorous Fact-Based Reporter.
|
|
|
|
## CRITICAL INSTRUCTIONS
|
|
You are a reporter who can ONLY write what is explicitly stated in the KNOWLEDGE BASE below.
|
|
- EVERY claim MUST be supported by the Knowledge Base.
|
|
- Use citations in the format [ID: X] after each fact.
|
|
- If a fact is NOT in the Knowledge Base, DO NOT WRITE IT.
|
|
- If you cannot answer the question from the Knowledge Base, say: "Não encontrei informações suficientes nas fontes consultadas para responder completamente."
|
|
- DO NOT use your internal knowledge. Pretend you know NOTHING except what is in the Knowledge Base.
|
|
- Language: Portuguese (Brazil).
|
|
|
|
## KNOWLEDGE BASE (Your ONLY source of truth)
|
|
{{knowledgeBase}}
|
|
|
|
## VALIDATION REPORT
|
|
{{validationReport}}
|
|
|
|
## USER QUERY
|
|
"{{userQuery}}"
|
|
|
|
## YOUR RESPONSE
|
|
Write a comprehensive answer using ONLY the Knowledge Base above. Cite sources using [ID: X].`,
|
|
|
|
CITATION_FORMAT: `
|
|
When citing sources, use this format:
|
|
- For inline citations: "Garuva está localizada em Santa Catarina [ID: 1]."
|
|
- For multiple sources: "A cidade possui praias e montanhas [ID: 1, 3]."
|
|
- At the end, list all sources used with their URLs.
|
|
`,
|
|
|
|
QUERY_ENRICHMENT: `Você é um especialista em otimização de queries de busca e expansão semântica.
|
|
|
|
Com base na query original e no contexto analisado, gere múltiplas variantes de queries enriquecidas para maximizar a cobertura de resultados relevantes.
|
|
|
|
## QUERY ORIGINAL
|
|
"{{originalQuery}}"
|
|
|
|
## CONTEXTO ANALISADO
|
|
- Intenção: {{intent}}
|
|
- Tópicos: {{topics}}
|
|
- Entidades: {{entities}}
|
|
- Contexto Temporal: {{temporalContext}}
|
|
- Contexto Geográfico: {{geographicContext}}
|
|
- Palavras-chave: {{keywords}}
|
|
- Sinônimos: {{synonyms}}
|
|
- Domínio: {{domain}}
|
|
|
|
## INSTRUÇÕES
|
|
|
|
Gere queries enriquecidas nas seguintes categorias:
|
|
|
|
1. **LITERAL**: Variações diretas da query original (preservar termos principais)
|
|
2. **SEMANTIC**: Variantes semânticas usando sinônimos e reformulações
|
|
3. **RELATED**: Perguntas relacionadas e FAQs
|
|
4. **EXPANDED**: Expansões temáticas específicas
|
|
5. **CONTEXTUAL**: Queries com contexto adicional (datas, locais, domínios)
|
|
|
|
## REGRAS CRÍTICAS
|
|
|
|
- **PRESERVAR ENTIDADES**: Se a query menciona uma entidade, TODAS as queries devem mencioná-la
|
|
- **NÃO TRADUZIR**: Preservar nomes próprios, termos técnicos na forma original
|
|
- **OTIMIZAR PARA BUSCA**: Usar termos que aparecem em resultados reais
|
|
- **COBERTURA AMPLA**: Cada categoria deve explorar diferentes ângulos
|
|
- **QUALIDADE > QUANTIDADE**: 3-5 queries por categoria é suficiente
|
|
|
|
Retorne APENAS um JSON válido com arrays para cada categoria.`
|
|
};
|
|
|
|
|
|
|
|
|
|
|