Local AI Agent (Django + Ollama + pgvector)

This project implements your workflow:

• Intent lookup from CustomerIntent
• Ask for contract number and resolve balance from CustomerBalance
• RAG fallback over PDF chunks stored in Postgres pgvector

1) Create virtual environment and install dependencies

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2) Ensure Ollama models are present

ollama run qwen2.5:7b
ollama pull nomic-embed-text

3) Run Postgres with pgvector (Docker example)

docker run --name pgvector-local \
  -e POSTGRES_PASSWORD=postgres \
  -e POSTGRES_USER=postgres \
  -e POSTGRES_DB=postgres \
  -p 5432:5432 \
  -d pgvector/pgvector:pg16

Enable extension:

docker exec -it pgvector-local psql -U postgres -d postgres -c "CREATE EXTENSION IF NOT EXISTS vector;"

4) Set environment variables

export POSTGRES_DB=postgres
export POSTGRES_USER=postgres
export POSTGRES_PASSWORD=postgres
export POSTGRES_HOST=localhost
export POSTGRES_PORT=5432

export OLLAMA_TEXT_MODEL=qwen
export OLLAMA_EMBED_MODEL=nomic-embed-text
export PGVECTOR_CONNECTION="postgresql+psycopg://postgres:postgres@localhost:5432/postgres"
export PGVECTOR_COLLECTION="pdf_knowledge_base"

5) Migrate and run server

python manage.py makemigrations
python manage.py migrate
python manage.py createsuperuser
python manage.py runserver

Open:

• Chat UI: http://127.0.0.1:8000/
• Admin: http://127.0.0.1:8000/admin/

6) Seed data

Use admin to create:

• CustomerIntent rows (intent + answer)
• CustomerBalance rows (contract number + contract balance)

7) Ingest your PDF into pgvector

python ingest_pdf.py "/absolute/path/to/your_qa.pdf"

Notes

• If intent lookup misses, the API falls back to RAG.
• session_id from frontend local storage is used to track awaiting_contract.
• Main endpoint: POST /api/chat with body:

{
  "session_id": "abc123",
  "message": "What is my renewal policy?"
}

newTestAudio