Yes, absolutely. We have Self-hosted, SaaS and hybrid deployments available, more here.

We have a trial setup with synthetic data you can use right away. If you have a database you want to test Waii with (recommended), we can get you set up in as little as 1-2 hours. If you have specific privacy requirements and want to integrate additional data sources or LLMs, we can give you a custom setup in just a few days.

You can achieve the highest possible accuracy with Waii - with real-world workloads (1000s of tables, 20-way joins, semi-structured data, etc). You get the fastest answers with sophisticated caching and optimization techniques. All this while using the fewest tokens possible.The system is also truly enterprise ready, with all required security, governance, and operational capabilities you need.

Waii supports SSO, RBAC and pass-through authorization. You have control over what your users can access and change in Waii, they will see only the data and metadata they have access to. Row level access and column masking is supported as well.

Waii offers fine-grained control over what data it uses. You can make sure no PII data is touched. We also offer PII detection and access logs as additional guardrails.

Depends on the complexity of the environment. For analytical workloads with many tables and complex queries, we recommend models with high reasoning capabilities: Typically OpenAI's GPT-4+ family, Anthropic Claude 3 Opus, or Gemini 1.5 ultra / pro.

‍

For less complex scenarios (simpler aggregates, more lookup-based queries, fewer tables), you can use smaller models: OpenAI GPT-3.5, Anthropic Claude 3 Sonnet, etc. For these cases OSS models can also achieve good results: E.g.: Llama 2/3, Mistral Large

‍

If you are dealing with one/few databases and the rate of change is slow, you can get more complex scenarios working on smaller models with appropriate fine tuning.

That is controllable, minimum is schema information and context to understand the schema. Sharing some categorical data, structure of json / xml data is recommended to get better results.

No. At Waii we believe that the system should automatically make the most of the available information and give an incremental path to improvement. Many of our customers are using Waii on previously unknown databases and rely on the automatic generation of the knowledge graph to get good query results.

Snowflake, Postgres, MongoDB, Trino, Athena, Presto, Redshift, MySQL, SQLite and OSQuery at the moment. We are constantly adding more. If you don't see your database here, check in and see when it will be available.

Waii supports Datahub, DBT, cube.dev, and OpenMetadata.org. We are constantly adding more. If you don't see your database here, check in and see when it will be available.

No. While we are an API first company and make everything we do available to developers, we can give you access to a SQL builder and chat experiences without coding required.

Waii can use a wide variety of vector databases. By default we use Postgres/pg_vector, but can use any vector db you have already set up.

Conversational intelligence or sometimes conversational BI or conversational analytics, is about gaining insights from data via natural language. It combines databases, visualizations and NLP to enable rapid analysis of complex datasets. That's at the heart of Waii, our API and tools are all tailor made to support this use case.

No. We are supporting customers and partners with a variety of different schemas. Star schemas and snowflake schemas are common and supported, but highly normalized schemas and "raw" schemas with highly nested data models are supported as well.

Yes! Waii supports a variety of schema types and you can get high accuracy queries out of all of them.

We have tested the system with databases up to thousands of tables. Tables with thousands of columns are supported. There are no bounds on data size and row numbers.

‍

With respect to query complexity: 10-20 level joins are common, multiple aggregates, window functions, and set functions are common. Semi-structured and geospatial data is supported.

No! You are in control of your and your customers' data.

Yes, Waii has a built in optimizer and optimization knowledge base and can write optimal queries.

Yes! Waii let's you visualize data all thru natural language. Multiple visualization frameworks are supported to best suit your needs. You can also build visualizations and publish them to popular BI tools such as metabase and superset.