Imported from KaushikML/SpiderSearch-qdrant (AGENTS.md). Install upstream with npx skills add KaushikML/SpiderSearch-qdrant. Copyright stays with the author.

AGENTS.md

Project overview

• spider search : multimodal classifier is a Streamlit app that correlates claims and evidence using Qdrant and optional Ollama.

Setup

• Create a virtualenv and install dependencies: python -m venv .venv and pip install -r requirements.txt.
• Copy the env template: cp .env.example .env.

Services

• Qdrant (required): docker run -p 6333:6333 -p 6334:6334 -v $(pwd)/qdrant_storage:/qdrant/storage qdrant/qdrant.
• Ollama (optional): ollama serve then ollama pull llama3.

Run

• streamlit run streamlit_app.py

Data and storage

• Qdrant data is stored in qdrant_storage/.
• SQLite metadata defaults to data/app.db (see SQLITE_PATH in .env).

Tests

• No automated test suite in this repo.

Useful entry points

• UI and workflows: streamlit_app.py.
• Ollama deduction prompt: models/llm_reasoner.py.
• Stance classification: models/stance_classifier.py.