You are viewing a single comment's thread from:

RE: LeoThread 2025-11-05 15-48

in LeoFinance21 days ago

Part 6/12:

  • Signing up on Pine Cone and creating an index optimized for higher throughput (P1 tier), with 1536-dimensional vectors corresponding to OpenAI’s ADA 2 embeddings.

  • Writing Python scripts employing the pinecone client library to:

  • Generate embeddings with OpenAI's API

  • Upsert messages, complete with metadata such as timestamps, speaker identity, and message content

  • Retrieve relevant memories based on similarity scores for subsequent use

The system is designed to index entire chat messages, which include user input, system replies, timestamps, and other metadata, stored as JSON files for external backup or manual inspection.

Sample Workflow

The script:

  • Initializes the Pine Cone index environment and sets necessary API keys.