undefined

The Setup Skill will:
1. Ask you to select providers (OpenAI, Anthropic, Cohere, etc.)
2. Configure API keys
3. Install dependencies
4. Generate configuration files
5. Launch the dashboard

undefined

undefined

from src.core.config import get_config

config = get_config()

undefined

undefined

undefined

class LLMConfig:
    provider: str = "openai"  # or "anthropic", "cohere"
    model: str = "gpt-4"
    temperature: float = 0.7
    max_tokens: int = 2048

class EmbeddingConfig:
    provider: str = "openai"  # or "cohere", "huggingface"
    model: str = "text-embedding-3-small"
    dimension: int = 1536

class RerankerConfig:
    enabled: bool = True
    provider: str = "cohere"  # or "jina", "cross-encoder"
    model: str = "rerank-english-v3.0"
    top_k: int = 5

from src.ingestion.pipeline import IngestionPipeline
from src.core.config import get_config

config = get_config()
pipeline = IngestionPipeline(config)

undefined

from src.retrieval.hybrid_search import HybridSearchRetriever
from src.core.config import get_config

config = get_config()
retriever = HybridSearchRetriever(config)

undefined

undefined

Configure in Claude Desktop (`claude_desktop_config.json`):

```json
{
  "mcpServers": {
    "rag-knowledge-hub": {
      "command": "python",
      "args": ["/path/to/project/src/mcp/server.py"],
      "env": {
        "PYTHONPATH": "/path/to/project"
      }
    }
  }
}

streamlit run src/dashboard/app.py

from src.evaluation.evaluator import RAGEvaluator
from src.core.config import get_config

config = get_config()
evaluator = RAGEvaluator(config)

undefined

undefined

Or programmatically:

```python
from src.core.config import get_config

config = get_config()
config.embedding.provider = "cohere"
config.embedding.model = "embed-english-v3.0"
config.embedding.dimension = 1024

from src.ingestion.splitters.base import BaseSplitter
from typing import List

class CustomSplitter(BaseSplitter):
    def __init__(self, chunk_size: int = 500, overlap: int = 50):
        self.chunk_size = chunk_size
        self.overlap = overlap
    
    def split(self, text: str, metadata: dict = None) -> List[dict]:
        chunks = []
        start = 0
        while start < len(text):
            end = start + self.chunk_size
            chunk_text = text[start:end]
            chunks.append({
                "text": chunk_text,
                "metadata": {
                    **(metadata or {}),
                    "chunk_index": len(chunks),
                    "start_char": start
                }
            })
            start += self.chunk_size - self.overlap
        return chunks

undefined

from src.retrieval.rerankers.base import BaseReranker
from typing import List

class CustomReranker(BaseReranker):
    def rerank(self, query: str, documents: List[dict], top_k: int = 5) -> List[dict]:
        # Custom reranking logic
        scored_docs = []
        for doc in documents:
            # Example: simple keyword matching score
            score = sum(1 for word in query.lower().split() 
                       if word in doc['text'].lower())
            scored_docs.append({**doc, 'rerank_score': score})
        
        # Sort by score and return top_k
        scored_docs.sort(key=lambda x: x['rerank_score'], reverse=True)
        return scored_docs[:top_k]

undefined

from src.ingestion.pipeline import IngestionPipeline

pipeline = IngestionPipeline(config)

undefined

import os
from pathlib import Path

pipeline = IngestionPipeline(config)
docs_dir = Path("./documents")

results = []
for pdf_file in docs_dir.glob("*.pdf"):
    try:
        result = pipeline.ingest_document(
            file_path=str(pdf_file),
            collection_name="batch_collection",
            metadata={"filename": pdf_file.name}
        )
        results.append(result)
        print(f"✓ Ingested {pdf_file.name}")
    except Exception as e:
        print(f"✗ Failed {pdf_file.name}: {e}")

print(f"Total successful: {len(results)}")

from src.core.config import get_config
from src.retrieval.hybrid_search import HybridSearchRetriever
from src.generation.generator import Generator
from src.evaluation.evaluator import RAGEvaluator

config = get_config()

# Generate
response = generator.generate(
    query=query,
    contexts=[c.text for c in contexts],
    metadata=[c.metadata for c in contexts]
)

# Evaluate (optional)
evaluator = RAGEvaluator(config)
metrics = evaluator.evaluate_single(
    query=query,
    response=response,
    contexts=[c.text for c in contexts]
)

return {
    "response": response,
    "contexts": contexts,
    "metrics": metrics
}

undefined

from src.rag_system import RAGSystem
from src.core.config import get_config

undefined