These CKAN Versions are exactely matched:

• 2.10.0
• 2.10.1
• 2.10.2
• 2.10.3
• 2.10.4
• 2.10.5
• 2.10.6
• 2.10.7
• 2.10.8
• 2.10.9
• 2.11.0
• 2.11.1
• 2.11.2
• 2.11.3
• 2.11.4
• 2.12.0
• 2.9.0
• 2.9.0a
• 2.9.1
• 2.9.10
• 2.9.11
• 2.9.13
• 2.9.2
• 2.9.3
• 2.9.4
• 2.9.5
• 2.9.6
• 2.9.7
• 2.9.8
• 2.9.9

ckanext-chat

A plugin integrating a chat interface in ckan with a pydanticai agent that can use all available ckan actions and url_patterns. All actions are done with user aware context. The chat interface uses marked and highightjs to display responses. Chat histories are saved in the local storage of the user. The agent is chat history aware. LLMs to use are configured in the bot/agent.py a section Model & Agent Setup. An Azure Openai and a local impementation using openai compartible api of ollama is implemeneted.

Option Rag Search

It has a rag_search tool that can facilitate a Milvus vector store if it is set up. Currently it relies on the Azure OpenAI embeddings api and will not work with local deployments. It uses the embedding model text-embedding-3-small to form the search vector. To make use of the documents it returns, the metadata of te vectores should include the dataset and resource ids at least. See the class VectorMeta for expected fields.

LLM Compartibility

Openai Models starting from gpt-35 on work very well. Local LLMs tested with ollama server are listed below.

in general reasoning models dont perform well

Requirements

A completion endpoint of the LLM model to use with the agent is needed. Currently uses Azure Cognitive Service Integrati