Facebook Rag Pipeline

Initial Setup

• Install Docker
• Use Docker to pull and run Qdrant
• Export your Facebook Data and unzip to a local folder
• Install Ollama
• Pull qwen3-embedding:8b and qwen3:14b using ollama. If you have limited RAM and/or GPU memory you can try qwen3:8b instead of qwen3:14b.

ollama pull qwen3-embedding:8b
ollama pull qwen3:14b

Local repo set up

• Clone this git repo
• Copy sample_config.json to config.json. Replace settings as necessary.
  • collection_name the name of the qdrant collection to create
  • document_path the json file in which to store the cleaned and processed Facebook data
  • log_folder the folder to write the log files to
  • raw_data_path The path to the raw facebook post data
  • model The LLM model to use to generate answers.
  • qdrant_host Where qdrant is hosted
  • qdrant_port The port where qrdant is hosted
• Install the required packages: pip install -r requirements.txt

Database set up

• Run facebook_processor.py. This will process the Facebook posts stored at raw_data_path and store them at document_path. Note I don't use Facebook a lot so I have less than 300 total posts. If you have more posts, modifications may be required.
• Make sure Qdrant is running and run store_docs.py.

Ask Questions

• Run query_docs.py and start asking questions!