Deep Web Research

from kapso.kapso import Kapso

kapso = Kapso()

# Research returns ResearchFindings
# mode: "idea" | "implementation" | "both" (default: "both")
# depth: "light" | "deep" (default: "deep")

findings = kapso.research(
    "QLoRA fine-tuning best practices for LLaMA models",
    mode="both",
    depth="deep",
)

# Access results
for idea in findings.ideas:
    print(idea.query)      # Original query
    print(idea.source)     # URL
    print(idea.content)    # Full content with sections
    print(idea.to_string())  # Formatted for LLM context

for impl in findings.implementations:
    print(impl.query)      # Original query
    print(impl.source)     # URL
    print(impl.content)    # Full content with code snippet
    print(impl.to_string())  # Formatted for LLM context

findings = kapso.research(
    "QLoRA fine-tuning best practices",
    mode="idea",
    depth="deep",
)

for idea in findings.ideas:
    print(idea.to_string())

findings = kapso.research(
    "unsloth FastLanguageModel example",
    mode="implementation",
    depth="deep",
)

for impl in findings.implementations:
    print(impl.to_string())

from kapso.kapso import Kapso

kapso = Kapso()

# Research
findings = kapso.research(
    "unsloth FastLanguageModel example",
    mode="both",
    depth="deep",
)

# Use research as context for evolve
solution = kapso.evolve(
    goal="Fine-tune a model with Unsloth + LoRA",
    context=[findings.to_string()],
    output_path="./models/unsloth_v1",
)

from kapso.kapso import Kapso

kapso = Kapso()

# Research
findings = kapso.research(
    "LoRA rank selection best practices",
    mode="both",
    depth="deep",
)

# Ingest all results into Knowledge Graph
kapso.learn(
    *findings.ideas,
    *findings.implementations,
    wiki_dir="data/wikis",
)

# .env
OPENAI_API_KEY=your-openai-api-key

def research(
    self,
    objective: str,
    *,
    mode: str = "both",  # "idea" | "implementation" | "both"
    depth: str = "deep",  # "light" | "deep"
) -> ResearchFindings:
    """
    Perform deep public web research.
    
    Args:
        objective: What to research (be specific)
        mode: Research mode - "idea", "implementation", or "both"
        depth: How thorough - "light" (faster) or "deep" (more comprehensive)
    
    Returns:
        ResearchFindings containing ideas and implementations
    """

from kapso.knowledge_base.types import Source, ResearchFindings

@dataclass
class Source.Idea:
    query: str    # Original research query
    source: str   # URL where this idea came from
    content: str  # Full content with sections
    
    def to_string(self) -> str:
        """Format for use as context in evolve()"""
    
    def to_dict(self) -> Dict[str, Any]:
        """Convert to dictionary"""

@dataclass
class Source.Implementation:
    query: str    # Original research query
    source: str   # URL where this implementation came from
    content: str  # Full content with code snippet
    
    def to_string(self) -> str:
        """Format for use as context in evolve()"""
    
    def to_dict(self) -> Dict[str, Any]:
        """Convert to dictionary"""

@dataclass
class ResearchFindings:
    query: str                              # Original research query
    ideas: List[Source.Idea]                # Ideas (if idea mode was run)
    implementations: List[Source.Implementation]  # Implementations (if implementation mode was run)
    report: Optional[Source.ResearchReport]  # Report (if study mode was run)
    
    def to_string(self) -> str:
        """Format all findings for use as context in evolve()"""
    
    def to_dict(self) -> Dict[str, Any]:
        """Convert to dictionary"""