Sazabi Backend

The Sazabi backend is Sazabi's native, managed storage layer. It is powered by ClickHouse, a high-performance columnar database optimized for analytical queries. When you send logs to Sazabi, they are stored here by default.

About the Sazabi backend

The Sazabi backend provides a unified storage layer for all your observability data. It is designed to handle high-volume log ingestion while maintaining fast query performance for Chat.

Key characteristics:

• Fully managed: Sazabi handles provisioning, scaling, and maintenance
• High performance: ClickHouse provides sub-second queries even on large datasets
• AI-optimized: Indexing is designed to support the queries the AI assistant runs
• Default for all projects: Every Sazabi project uses this backend automatically

How it works

When logs flow into Sazabi from your data sources:

1. Intake: Logs arrive at Sazabi's regional intake endpoints
2. Processing: Logs are validated, enriched, and formatted
3. Storage: Logs are written to ClickHouse with appropriate indexes
4. Indexing: Full-text search indexes are built automatically
5. Available: Logs are immediately queryable by Chat

This entire pipeline typically completes in seconds, meaning logs are available for analysis almost immediately after ingestion.

Capabilities

Full-text search

The Sazabi backend indexes all log content for full-text search. When you ask Chat to "find errors containing timeout," it can quickly locate matching logs across your entire retention window.

Time-series queries

ClickHouse excels at time-series analysis. Chat can efficiently run queries like:

• "How many errors occurred in the last hour?"
• "What is the 95th percentile latency trend over the past week?"
• "Compare error rates between today and yesterday"

Aggregations

The backend supports complex aggregations that help Chat identify patterns:

• Count logs by service, status code, or any field
• Calculate averages, percentiles, and distributions
• Group and filter by multiple dimensions

Schema auto-detection

When logs arrive, Sazabi automatically detects and indexes common fields. JSON logs are parsed, and nested fields become queryable without any configuration. Chat can query fields like error.message, user.id, or response.status_code without manual schema setup.

Query experience

With the Sazabi backend, you interact with your logs through natural language. Instead of writing ClickHouse SQL or learning a query language, you simply ask questions:

• "Why are checkout errors elevated right now?"
• "Show me the slowest API requests from the payments service"
• "What changed before the spike in 5xx responses?"

Chat translates your questions into optimized ClickHouse queries, retrieves the results, and provides plain-language answers with supporting evidence from your logs.

Data retention

Default data retention is 30 days for all projects. Logs older than your retention window are automatically deleted.

Enterprise plans support custom retention periods. Contact us to discuss your requirements.

Storage limits

The Sazabi backend has the following limits:

For detailed information about rate limits and quotas, see Limits.

When to use the Sazabi backend

The Sazabi backend is the best choice for most users:

• Unified storage: All your logs in one place, regardless of where they originated
• Optimized for AI: Indexing and query patterns designed for the AI assistant
• Simple setup: Send logs to Sazabi, and they are automatically stored and indexed
• Fast queries: ClickHouse provides excellent performance for analytical workloads

Limitations

The Sazabi backend is designed for log storage and analysis. It is not a replacement for:

• APM tools: While Sazabi stores traces, dedicated APM tools provide more specialized trace analysis
• Real-time monitoring: Sazabi ingestion has a small delay; for sub-second alerting, use dedicated monitoring tools
• Compliance archival: For long-term compliance storage, consider dedicated archival solutions

Next steps