Skip to content

Reference

Technical reference documentation.

Performance

Benchmarks

Comprehensive performance benchmarks:

Performance Benchmarks →

  • Query latency (<100ms p95)
  • Throughput (1000+ req/s)
  • Memory usage profiles
  • Scaling characteristics

Configuration

Server Configuration

All server settings are in config.toml:

[server]
host = "0.0.0.0"
port = 8080
api_key = "sk-sekha-your-secure-key-min-32-chars-long"

[database]
url = "sqlite:///data/sekha.db"
max_connections = 10

[vector_store]
host = "localhost"
port = 8000
collection_name = "sekha_embeddings"

[llm]
embedding_model = "nomic-embed-text"
summarization_model = "llama3.1:8b"

[rate_limiting]
requests_per_second = 100
burst_size = 200

See Configuration Guide for all options.

Environment Variables

Alternatively, use environment variables:

SEKHA_HOST=0.0.0.0
SEKHA_PORT=8080
SEKHA_API_KEY=your-api-key
DATABASE_URL=sqlite:///data/sekha.db
CHROMA_HOST=localhost
CHROMA_PORT=8000
OLLAMA_HOST=localhost
OLLAMA_PORT=11434

Database Schema

SQLite Tables

conversations - id - Primary key - label - Display name - folder - Organization folder - importance - 1-10 score - created_at - Timestamp - updated_at - Timestamp

messages - id - Primary key - conversation_id - Foreign key - role - user/assistant/system - content - Message text - timestamp - Creation time

embeddings - id - Primary key - message_id - Foreign key - embedding_id - ChromaDB ID - model - Embedding model used

Vector Store

ChromaDB Collections

sekha_embeddings - Embedding dimension: 768 (nomic-embed-text) - Distance metric: Cosine similarity - Metadata: conversation_id, timestamp, importance

Metrics & Monitoring

Available Metrics

Access via /stats endpoint:

{
  "total_conversations": 150,
  "total_messages": 1250,
  "storage_size_mb": 45.2,
  "embeddings_count": 1250,
  "uptime_seconds": 86400,
  "requests_total": 5420,
  "errors_total": 12
}

Health Checks

# Overall health
curl http://localhost:8080/health

# Component status
curl http://localhost:8080/health/detailed

API Versioning

Current version: v1

Base path: /api/v1

Version strategy: - Major versions for breaking changes - Minor versions for new features - Backward compatibility within major version

Limits & Quotas

Default Limits

  • Max message length: 100,000 characters
  • Max messages per conversation: 10,000
  • Max conversations: Unlimited
  • Max query results: 1,000
  • Rate limit: 100 req/s (configurable)

Storage

  • SQLite: ~1KB per message average
  • ChromaDB: ~3KB per embedding
  • Recommended: 1GB per 100,000 messages

Next Steps

Built with MkDocs and Material for MkDocs