pan3793 commented on code in PR #7431: URL: https://github.com/apache/kyuubi/pull/7431#discussion_r3298383922
########## docs/quick_start/quick_start_with_data_agent.md: ########## @@ -0,0 +1,211 @@ +<!-- +- Licensed to the Apache Software Foundation (ASF) under one or more +- contributor license agreements. See the NOTICE file distributed with +- this work for additional information regarding copyright ownership. +- The ASF licenses this file to You under the Apache License, Version 2.0 +- (the "License"); you may not use this file except in compliance with +- the License. You may obtain a copy of the License at +- +- http://www.apache.org/licenses/LICENSE-2.0 +- +- Unless required by applicable law or agreed to in writing, software +- distributed under the License is distributed on an "AS IS" BASIS, +- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +- See the License for the specific language governing permissions and +- limitations under the License. +--> + +# Data Agent Engine + +The Data Agent engine lets users explore and query data with natural language. It is a Kyuubi engine type (`DATA_AGENT`) that runs an LLM-driven [ReAct](https://arxiv.org/abs/2210.03629) agent. The agent reasons about the user's request, calls a small set of SQL tools against a JDBC-accessible backend (Spark, Trino, MySQL, SQLite, ...), and streams its responses to the bundled web UI. + +This page covers how to configure the engine, how the web UI behaves at runtime, the approval / risk model that gates SQL tool execution, and a few deployment notes worth knowing. + +## At a glance + +- **Engine type**: `DATA_AGENT`, share level follows `kyuubi.engine.share.level` like every other engine +- **LLM provider**: any OpenAI-compatible chat-completion endpoint (OpenAI, Azure OpenAI, vLLM, llama.cpp, or any other provider that exposes the OpenAI chat-completion API) +- **Datasource**: any JDBC URL the engine can reach. When unset, defaults to ZooKeeper service discovery for ZK-HA deployments, or `jdbc:hive2://localhost:<thrift-port>/default` otherwise +- **Tools** (built in): `run_select_query` (low-risk, read-only) and `run_mutation_query` (high-risk, mutating). Per-dialect prompt templates ship for Spark / Trino / MySQL / SQLite / Generic +- **Interface**: the bundled Kyuubi web UI. Pick **Data Agent** from the left navigation after logging in + +## Quick start + +### 1. Configure the LLM provider + +Add the following to `$KYUUBI_HOME/conf/kyuubi-defaults.conf`: + +``` +kyuubi.engine.data.agent.provider OPENAI_COMPATIBLE + +# OpenAI-compatible endpoint and credentials +kyuubi.engine.data.agent.openai.endpoint https://api.openai.com/v1 +kyuubi.engine.data.agent.openai.api.key sk-xxxxxxxxxxxxxxxxxxxxxxxx +kyuubi.engine.data.agent.model gpt-4o-mini +``` + +Substitute the endpoint, API key, and model ID for whichever OpenAI-compatible provider you use. + +### 2. Restart Kyuubi and open the UI + +``` +$KYUUBI_HOME/bin/kyuubi restart +``` + +Open `http://<kyuubi-host>:10099/ui` in a browser (the port comes from `kyuubi.frontend.rest.bind.port`, default `10099`). Pick **Data Agent** in the left navigation. + +### 3. Ask a question Review Comment: is there a known dataset and question set that can be used for a demo? for example, kyuubi provides Spark TPC-H connector, which could provide the TPC-H dataset on the fly, is it suitable for the demo purpose? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
