Show HN: Oodle.ai – $10 per million agent traces
3 hours ago
- #LLMOps
- #Performance Monitoring
- #Agent Observability
- Agent traces have large metadata, causing slow search results.
- Agents fail in non-deterministic ways, requiring 100% trace retention for debugging while managing costs.
- Oodle offers a solution with <1s P99 queries, columnar storage, and serverless compute for fast trace retrieval.
- Pricing is based on flat ingestion ($0.30/GB) and retention ($0.001/GB/mo), with no per-query charges.
- Every trace is analyzed to detect issues like user frustration, tool call inefficiencies, and anomalies without manual evaluation.
- Features include monitoring silent failures, performance bottlenecks, sentiment analysis, and cost optimizations.
- A complete LLMOps platform provides tools for debugging, eval pipelines, prompt iteration, and cost estimation across models.
- Oodle supports quick setup with a free tier and aims to surface agent issues before users notice.