LlamaIndex: Implement suggestions by CodeRabbit

Author: amotl

docs/integrate/llamaindex/index.md

@@ -16,14 +16,13 @@
 :::{rubric} About
 :::
 
-[LlamaIndex] is a data framework for Large Language Models (LLMs). It comes with
-pre-trained models on massive public datasets such as GPT-4 or Llama 2, and
-provides an interface to external data sources allowing for natural language
-querying on your private data.
+[LlamaIndex] is a data framework for Large Language Models (LLMs). It integrates
+with providers and models such as GPT‑4 or Llama 3/4, and provides interfaces to
+external data sources for natural‑language querying of your private data.
 
-Azure Open AI Service is a fully managed service that runs on the Azure global
+Azure OpenAI Service is a fully managed service that runs on the Azure global
 infrastructure and allows developers to integrate OpenAI models into their
-applications. Through Azure Open AI API one can easily access a wide range of
+applications. Through the Azure OpenAI API, you can access a wide range of
 AI models in a scalable and reliable way.
 
 :::{rubric} Use case examples
@@ -55,8 +54,8 @@ contemporary large language models, optionally offline.
 :::{grid-item-card} Demo: Using LlamaIndex with OpenAI and CrateDB
 :link: llamaindex-tutorial
 :link-type: ref
-- Connect your CrateDB data to an LLM using OpenAI or Azure OpenAI.
-- Text-to-SQL / Talk to your data: Query the database in human language; query CrateDB in plain English.
+- Connect CrateDB to an LLM via OpenAI or Azure OpenAI.
+- Text‑to‑SQL: Query CrateDB in natural language.
 
 {tags-primary}`Cloud LLM`
 {tags-secondary}`LLM`

docs/integrate/llamaindex/synopsis.md

@@ -9,11 +9,12 @@ contemporary large language models, optionally offline.
 ## Install
 Project dependencies. For example, use them in a `requirements.txt` file.
 ```shell
-langchain-openai<0.3
-llama-index-embeddings-langchain<0.4
-llama-index-embeddings-openai<0.4
-llama-index-llms-azure-openai<0.4
-llama-index-llms-openai<0.4
+langchain-openai<0.4
+llama-index-embeddings-langchain<0.5
+llama-index-embeddings-openai<0.6
+llama-index-llms-azure-openai<0.5
+llama-index-llms-ollama<0.8
+llama-index-llms-openai<0.6
 sqlalchemy-cratedb
 ```
 
@@ -33,14 +34,14 @@ import os
 from llama_index.llms.ollama import Ollama
 from llama_index.llms.openai import OpenAI
 from llama_index.core import SQLDatabase
-from llama_index.core.indices.struct_store import NLSQLTableQueryEngine
+from llama_index.core.query_engine import NLSQLTableQueryEngine
 import sqlalchemy as sa
 ```
 :::
 
 :::{grid-item}
 :columns: 4
-Provision LLM using OpenAI model.
+Provision an LLM using an OpenAI model.
 :::
 :::{grid-item}
 :columns: 8
@@ -55,14 +56,14 @@ llm = OpenAI(
 
 :::{grid-item}
 :columns: 4
-Alternatively, provision LLM using self-hosted model.
+Alternatively, provision an LLM using a self-hosted model.
 :::
 :::{grid-item}
 :columns: 8
 ```python
 llm = Ollama(
     base_url=os.getenv("OLLAMA_BASE_URL", "http://localhost:11434"),
-    model="gemma3:1b",
+    model=os.getenv("OLLAMA_MODEL", "gemma3:1b"),
     temperature=0.0,
     request_timeout=120.0,
     keep_alive=-1,
@@ -78,7 +79,6 @@ Connect to CrateDB.
 :columns: 8
 ```python
 database = sa.create_engine(os.getenv("CRATEDB_SQLALCHEMY_URL", "crate://crate@localhost:4200"))
-database.connect()
 ```
 :::
 
@@ -117,7 +117,7 @@ answer = nlsql.query("Quelle est la valeur moyenne pour le capteur 1 ?")
 import os
 import sqlalchemy as sa
 
-from llama_index.core.utilities.sql_wrapper import SQLDatabase
+from llama_index.core import SQLDatabase
 from llama_index.core.query_engine import NLSQLTableQueryEngine
 from llama_index.core import Settings
 

docs/integrate/llamaindex/tutorial.md

@@ -3,20 +3,23 @@
 
 ## Introduction
 
-[LlamaIndex](https://www.llamaindex.ai/) is a data framework for Large Language Models (LLMs). It comes with pre-trained models on massive public datasets such as [GPT-4](https://openai.com/index/gpt-4/) or [Llama 4](https://www.llama.com/models/llama-4/) and provides an interface to external data sources allowing for natural language querying on your private data.
+[LlamaIndex](https://www.llamaindex.ai/) is a data framework for Large Language Models (LLMs).
+It integrates with models such as [GPT‑4](https://openai.com/index/gpt-4/) or [Llama 4](https://www.llama.com/models/llama-4/) and provides interfaces to external data sources for natural‑language querying of your private data.
 
-[Azure Open AI Service](https://azure.microsoft.com/en-us/products/ai-services/openai-service) is a fully managed service that runs on the Azure global infrastructure and allows developers to integrate OpenAI models into their applications. Through Azure Open AI API one can easily access a wide range of AI models in a scalable and reliable way.
+[Azure OpenAI Service](https://azure.microsoft.com/en-us/products/ai-services/openai-service) is a fully managed service on the Azure global infrastructure that lets developers integrate OpenAI models into applications. Through the Azure OpenAI API, you can access a wide range of AI models in a scalable and reliable way.
 
-In this tutorial, we will illustrate how to augment existing LLMs with data stored in CrateDB through the LlamaIndex framework and Azure Open AI Service. By doing this, you will be able to use the power of generative AI models with your own data in just a few lines of code.
+This tutorial shows how to augment LLMs with data stored in CrateDB using LlamaIndex and Azure OpenAI, enabling natural‑language queries over your data.
 
 If you want to run this in your own environment, we've provided all of the code and supporting resources that you'll need in the [`cratedb-examples`](https://github.com/crate/cratedb-examples/tree/main/topic/machine-learning/llama-index) GitHub repository.
 
 ## Prerequisites
 
 * Python 3.10 or higher
 * Recent version of LlamaIndex, please follow the [installation instructions](https://gpt-index.readthedocs.io/en/latest/getting_started/installation.html)
+* `sqlalchemy-cratedb`
+* `SQLAlchemy` (if not pulled transitively)
 * Running instance of [CrateDB](https://console.cratedb.cloud/)
-* [An Azure subscription](https://azure.microsoft.com/en-gb/free/cognitive-services/) and [Azure OpenAI resource](https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/create-resource?pivots=web-portal) in your desired subscription
+* [Azure subscription](https://azure.microsoft.com/en-gb/free/cognitive-services/) and [Azure OpenAI resource](https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/create-resource?pivots=web-portal)
 
 ## Deploy models in Azure OpenAI
 
@@ -28,9 +31,9 @@ To deploy the models required for this tutorial, follow these steps:
 
 ![92c692f4-2cc8-4661-9fc3-7c037cccbd28|690x263](https://us1.discourse-cdn.com/flex020/uploads/crate/original/2X/e/e0a411357b62eb67995b818b4e235d8c64aebc99.png)
 
-2. This will open Azure AI Studio. Azure AI Studio enables developers to build, run, and deploy AI applications. Click on the *Create new deployment* button to deploy the following models:
-  1. **GPT-35-turbo** for text generation tasks
-  2. **text-embedding-ada-002** for generating embeddings
+2. This opens Azure AI Studio. Click *Create new deployment* and deploy:
+  1. A chat/completions model (e.g., **gpt-4o-mini**)
+  2. An embeddings model (e.g., **text-embedding-3-large**)
 
 The basic deployment of each model is straightforward in Azure OpenAI Studio: You need to select the model you want to deploy and specify the unique name:
 
@@ -136,7 +139,6 @@ We use sqlalchemy, a popular SQL database toolkit, to connect to CrateDB and SQL
 
 ```python
 engine_crate = sa.create_engine(os.getenv("CRATEDB_SQLALCHEMY_URL"))
-engine_crate.connect()
 ```
 The value of `CRATEDB_SQLALCHEMY_URL` should be a URL format connection string containing the hostname, username and password for your CrateDB instance:
 
@@ -153,8 +155,7 @@ sql_database = SQLDatabase(
 )
 ```
 
-Then use that to create am instance of `NLSQLTableQueryEngine`:
-
+Then create an instance of `NLSQLTableQueryEngine`:
 ```python
 query_engine = NLSQLTableQueryEngine(
     sql_database=sql_database,
@@ -183,15 +184,15 @@ print("Answer was:", answer)
 Often, we are also interested in the query that produces the output. This is included in the answer's metadata:
 
 ```python
-print(answer.metadata))
+print(answer.metadata)
 # {'result': [(17.033333333333335,)], 'sql_query': 'SELECT AVG(value) FROM time_series_data WHERE sensor_id = 1'}
 ```
 
 ## Takeaway
 
 In this tutorial, we've embarked on the journey of using a natural language interface to query CrateDB data. We've explored how to seamlessly connect your data to the power of LLM using LlamaIndex and the capabilities of Azure OpenAI.
 
-This tutorial is just the beginning. You can expect further resources, documentation, and tutorials related to CrateDB and generative AI from us. Also, stay tuned for the CrateDB 5.5 release: we will soon announce the support for the vector store and search, allowing you to implement similarity-based data retrieval efficiently.
+This tutorial is a starting point. Explore additional resources on CrateDB and generative AI as they become available.
 
 If you want to try this out yourself, you can find the full example code and supporting resources in the [`cratedb-examples` GitHub repository](https://github.com/crate/cratedb-examples/tree/main/topic/machine-learning/llama-index).
 

docs/start/query/ai-integration.md

@@ -212,8 +212,8 @@ frameworks and libraries, see CrateDB's support for MLflow and PyCaret at
 
 
 ## Text-to-SQL
-Text-to-SQL: Talk to your data using human language and contemporary large
-language models, optionally offline.
+Text‑to‑SQL lets you query data in natural language with contemporary
+large language models, optionally offline.
 :::{seealso}
 {ref}`llamaindex-synopsis`
 :::