Best Practices

from moorcheh_sdk import (
    MoorchehClient,
    MoorchehError,
    AuthenticationError,
    InvalidInputError,
    NamespaceNotFound,
    ConflictError,
    APIError
)

try:
    client = MoorchehClient(api_key="your-api-key")
    
    # Perform operations
    client.delete_namespace("non-existent-namespace")
    
except NamespaceNotFound as e:
    print(f"Namespace not found: {e}")
except ConflictError as e:
    print(f"Conflict error: {e}")
except AuthenticationError as e:
    print(f"Authentication failed: {e}")
except InvalidInputError as e:
    print(f"Invalid input: {e}")
except APIError as e:
    print(f"API error: {e}")
except MoorchehError as e:
    print(f"General Moorcheh error: {e}")

def chunk_document(text, chunk_size=1000, overlap=200):
    """Split document into overlapping chunks"""
    chunks = []
    start = 0
    
    while start < len(text):
        end = start + chunk_size
        chunk = text[start:end]
        chunks.append(chunk)
        start = end - overlap
    
    return chunks

# Upload chunked documents
document_text = "Very long document content..."
chunks = chunk_document(document_text)

documents = [
    {
        "id": f"doc-1-chunk-{i}",
        "text": chunk,
        "metadata": {
            "document_id": "doc-1",
            "chunk_index": i,
            "total_chunks": len(chunks)
        }
    }
    for i, chunk in enumerate(chunks)
]

client.upload_text(namespace_name="my-documents", documents=documents)

# Upload in batches
batch_size = 100
for i in range(0, len(all_documents), batch_size):
    batch = all_documents[i:i + batch_size]
    client.upload_text(namespace_name="my-documents", documents=batch)
    print(f"Uploaded batch {i//batch_size + 1}")

import time

# Upload documents
client.upload_text(namespace_name="my-documents", documents=documents)

# Wait for embedding generation
time.sleep(2)

# Now search
results = client.search(namespaces=["my-documents"], query="my query")

# Development
client.create_namespace("dev-documents", "text")

# Staging
client.create_namespace("staging-documents", "text")

# Production
client.create_namespace("prod-documents", "text")

# Different namespaces for different content
client.create_namespace("blog-posts", "text")
client.create_namespace("documentation", "text")
client.create_namespace("support-tickets", "text")

# Keep versions separate
client.create_namespace("embeddings-v1", "vector", 1536)
client.create_namespace("embeddings-v2", "vector", 768)

documents = [{
    "id": "doc-1",
    "text": "Document content...",
    "metadata": {
        "title": "Product Documentation",
        "category": "api",
        "author": "engineering",
        "date": "2024-01-15",
        "tags": ["api", "reference", "v2"],
        "priority": "high",
        "department": "engineering"
    }
}]

# Good: Consistent schema
documents = [
    {"id": "1", "text": "...", "metadata": {"category": "tech", "date": "2024-01"}},
    {"id": "2", "text": "...", "metadata": {"category": "sales", "date": "2024-02"}}
]

# Bad: Inconsistent keys
documents = [
    {"id": "1", "text": "...", "metadata": {"cat": "tech", "created": "2024-01"}},
    {"id": "2", "text": "...", "metadata": {"category": "sales", "date": "2024-02"}}
]

# Focused search: 3-5 results
results = client.search(namespaces=["docs"], query="API endpoint", top_k=5)

# Comprehensive search: 10-20 results
results = client.search(namespaces=["docs"], query="pricing", top_k=15)

# Filter by category and date
results = client.search(
    namespaces=["docs"],
    query="authentication #category:api #date:2024",
    top_k=10
)

# Only return high-confidence results
results = client.search(
    namespaces=["docs"],
    query="security",
    kiosk_mode=True,
    threshold=0.7,
    top_k=10
)

# General queries: Claude Sonnet 4
answer = client.get_answer(
    namespace="docs",
    query="What features do you offer?",
    ai_model="anthropic.claude-sonnet-4-20250514-v1:0"
)

# Code-related: DeepSeek R1
answer = client.get_answer(
    namespace="code-docs",
    query="How to implement authentication?",
    ai_model="deepseek.r1-v1:0"
)

# Long context: Llama 4 Maverick
answer = client.get_answer(
    namespace="legal-docs",
    query="Summarize the contract",
    ai_model="meta.llama4-maverick-17b-instruct-v1:0"
)

# Factual answers: Low temperature
answer = client.get_answer(
    namespace="docs",
    query="List the API endpoints",
    ai_model="anthropic.claude-sonnet-4-20250514-v1:0",
    temperature=0.1
)

# Creative content: Higher temperature
answer = client.get_answer(
    namespace="blog",
    query="Write an introduction",
    ai_model="anthropic.claude-sonnet-4-20250514-v1:0",
    temperature=0.9
)

answer = client.get_answer(
    namespace="docs",
    query="Explain the pricing",
    ai_model="anthropic.claude-sonnet-4-20250514-v1:0",
    headerPrompt="You are a helpful sales assistant. Be concise and friendly.",
    footerPrompt="End with a call to action to contact sales."
)