Skip to content

README.md

Latest commit

 

History

History
71 lines (47 loc) · 1.56 KB

README.md

File metadata and controls

71 lines (47 loc) · 1.56 KB
Raw

Answer Architect

Skeleton project for RAG/LLM orchestration inspired by the anomaly-detector implementation.

Quick start

  • poetry run python tools/pas_estimate.py

Local development (quick)

  1. Install dependencies (Poetry):
poetry install
  1. Run tests:
poetry run pytest -q
  1. Start the FastAPI app locally (set environment variables as needed):
export VECTOR_STORE_TYPE=opensearch
export OPENSEARCH_ENDPOINT=https://your-opensearch-endpoint
export AWS_REGION=us-west-2
export S3_DATA_BUCKET=your-dev-bucket
poetry run uvicorn src.rag_api.app:app --reload
  1. Use the checklist updater CLI to list/mark items:
# List tasks
python scripts/update_checklist.py --file checklist.md --list

# Mark a task by substring
python scripts/update_checklist.py --file checklist.md --mark "Initialize Poetry project"

# Mark by index (1-based)
python scripts/update_checklist.py --file checklist.md --index 3

Project structure

Docker

Build image:

docker build -t answer-architect:local .

Run container (example):

docker run --rm -p 8000:8000 -e AWS_REGION=us-west-2 -e VECTOR_STORE_TYPE=opensearch answer-architect:local

Testing strategy

Unit tests in tests/ are designed to run quickly and use lightweight fakes for heavy external dependencies (OpenSearch, boto3, SentenceTransformers). Integration tests that exercise real services should be placed under tests/integration/ and run in a CI job or developer environment that installs the real dependencies.

See docs/TESTING.md for more details.