LLM Observability Platform

Spanlens solves the problem of limited visibility and tracking in LLM (Large Language Models) usage. It enables developers to log every request made to services like OpenAI, Anthropic, and Gemini, allowing them to monitor costs, latency, and token usage easily. By implementing Spanlens, users can catch anomalies, track multi-step workflows, and even receive recommendations for cheaper models to optimize spending. Key features include comprehensive logging of requests, cost tracking with alerts, anomaly detection, PII scanning, and detailed analytics for user sessions. Notable alternatives include Langfuse, Helicone, and LangSmith, but Spanlens stands out due to its open-source nature and ease of integration without modifying existing SDKs. Compared to other platforms, Spanlens simplifies the setup with just a single line of code, providing rapid insights without overhead. Spanlens offers various pricing tiers. The Free plan allows up to 50K requests per month, while the Pro plan is priced at $29 per month for up to 100K requests. The Team plan at $149 per month supports 1M requests, and Enterprise options offer customized solutions based on needs. FAQs: 1. How does Spanlens instrument requests? - Simply swap your existing provider SDK with our drop-in solution. 2. What is the impact on latency? - There is minimal overhead, typically under 3ms, with asynchronous ingestion ensuring requests complete without delay. 3. How does Spanlens handle PII? - It flags potential PII during logging and masks sensitive information before storage. 4. Can data be exported from Spanlens? - Yes, users can export data in JSON, CSV, or Parquet formats at any time. 5. Is Spanlens suitable for team usage? - Absolutely! It features project isolation, role management, and audit logs for collaborative teams.