Evaluate quality
Vectara’s observability tools reveal detailed insights into query behavior, making it easier to troubleshoot and optimize performance. Paired with evaluation features that detect inaccuracies, these capabilities help developers fine-tune their systems, ensure factual consistency, and build confidence in AI-driven outputs.
Choose the task
- For aggregate agent activity, latency, token usage, and custom metrics, start with Monitor agents.
- For an individual agent conversation or trace, use Inspect agent runs.
- For corpus query history, use Query observability.
- For navigation and developer handoff in the simplified interface, see View metrics in Business Suite.
- For factual consistency and quality evaluation, use the resources below. Operational counts and latency do not establish answer quality.
The Console label Observability and this documentation category, Evaluate quality, are not the same navigation surface. The existing quality and query-history URLs remain unchanged.
Diagnose, debug, and optimize query performance with detailed query histories and execution flow analysis to improve your generative AI applications.
Gain insights into every stage of the query lifecycle for better understanding and optimization.
Use the Vectara Factual Consistency Score to evaluate and detect hallucinations in AI-generated outputs.
Achieve accurate summaries with calibrated scores across multiple languages.
Detect hallucinations in AI-generated summaries and view corrected versions updated with factual source material.
Ensures that summaries passed through HCM retain their intended meaning.
Use the open-source Vectara Open RAG Eval toolkit to evaluate and optimize your RAG pipelines.
Compare configurations systematically and validate improvements with measurable metrics.