Description

Resolves #21232

Checklist

• I fully tested it as developer / designer and created unit / integration tests
• I documented it (or no need to document or I will create a separate documentation issue)
• I worked with the design team to get their idea for UI/UX design (or no need to design for this PR)
• I've assigned reviewer, related labels and set a milestone for this PR

How to test it?

Setup

• Run ollama in container:

docker run -d --name ollama --gpus all -p 11434:11434 ollama/ollama

--gpus all parameter is crucial for LLM performance. It allows container to use your GPU.

• Make sure llama3.2 model is downloaded:

docker exec ollama ollama pull llama3.2

• Pull the embedder:

docker exec ollama ollama pull nomic-embed-text

• Deploy PgVector container:

docker run -d --name pgvector -e POSTGRES_PASSWORD=myPassw0rd -p 5432:5432 ankane/pgvector

• Run the Volo.AIManagement.DemoApp with --migrate-database flag once

dotnet run -- --migrate-database

---

Testing UI

• Configure your workspace with your credentials:

Workspace Name: Test
Provider: Ollama
Model Name: llama3.2
Provider Base URL: http://localhost:11434
Embedder Provider: Ollama
Embedder Model Name: nomic-embed-text
Embedder Base URL: http://localhost:11434
Vector Store Provider: Pgvector
Vector Store Settings: Host=localhost;Port=5432;Database=postgres;Username=postgres;Password=myPassw0rd

System Prompt (optional):

You are a helpful assistant that can answer questions and help with tasks. 
Make tool calls to search required information from uploaded documents, knowledge bases, or specific domain knowledge.
Return the most relevant document chunks with their sources.
Multiple alternative queries can be provided to the function to increase the chances of finding relevant documents.

• Upload documents to the workspace:
  robot-mop.pdf

• Ask a question to the workspace:

Dokümana göre, yazılım güncellemesi yapmak için şarjımın en az kaç olması gerekiyor?

10.2.0-rc.1 24 February, 2026 abp-module-ai-management feature