Skip to content

SETUP.md

Latest commit

 

History

History
114 lines (86 loc) · 3.39 KB

SETUP.md

File metadata and controls

114 lines (86 loc) · 3.39 KB
Raw

ChatPilot Setup Guide

Prerequisites

  • Python 3.9 or higher
  • pip (Python package manager)
  • Docker (required for secure code execution sandbox)
  • LLM API Key - One of the following is required:
    • OpenAI (GPT-4o)
    • Anthropic (Claude 3.5 Sonnet)
    • DeepSeek (DeepSeek-V3/R1)
    • Google (Gemini-1.5 Pro/Flash)
  • RAG Configuration (Embedding & Reranking):
    • Local (BAAI): Fast, offline, no API key needed
    • Cloud (OpenAI / Cohere): High-performance embeddings and reranking
  • Web Search API Key (Optional):
    • Tavily API Key (Optimized for research/agents)
    • Google Custom Search API key + CSE ID

Installation

1. Clone the repository

git clone https://github.com/Sreehari05055/ChatPilot.git
cd chatpilot

2. Install dependencies

pip install -r requirements.txt

3. Get API Keys

ChatPilot supports multiple providers for both LLM and RAG features.

Required (LLM Provider): Choose your primary LLM and get its API key:

RAG Support (Embeddings & Reranking): You can mix and match any combination of providers:

  • Embeddings: openai, cohere, or local (BAAI BGE)
  • Reranker: cohere or local (BAAI BGE)

Note

If you select cohere or openai as a provider but do NOT provide the corresponding API key, ChatPilot will automatically fall back to local models to ensure the app continues to function.

Optional (Search):

  • Tavily: Recommended for web research.
  • OpenAlex: For scholar research tools.

4. Set up environment variables

Create a .env file in the project root with the following structure:

# --- LLM CONFIGURATION ---
LLM_PROVIDER=openai  # openai, anthropic, deepseek, or google
MODEL_NAME=gpt-4o     # The specific model version to use
LLM_API_KEY=your_api_key_here

# --- RAG CONFIGURATION (MIX & MATCH) ---
# Choose your embedding and reranker providers independently
EMBEDDING_PROVIDER=openai  # Options: openai, cohere, local
RERANKER_PROVIDER=local    # Options: cohere, local

# --- SEARCH CONFIGURATION (OPTIONAL) ---
TAVILY_API_KEY=your_tavily_api_key
OPENALEX_API_KEY=your_openalex_api_key

# OR Google Custom Search
WEB_SEARCH_API_KEY=your_google_api_key
CSE_ID=your_cse_id

5. Build the Docker sandbox image

For secure code execution / data analysis:

docker build -t python-sandbox:latest -f Dockerfile .

6. Run ChatPilot

python main.py

Advanced Configuration

Resource Limits (Docker Sandbox)

Adjust container resource limits in .env to match your hardware:

  • SANDBOX_NANO_CPUS: CPU limit (e.g., 500000000 is 0.5 CPU)
  • SANDBOX_MEM_LIMIT: Memory limit (e.g., 2g)

RAG Performance Tuning

Edit app/core/config.py to adjust:

  • chunk_size: Size of document segments (default 512)
  • top_k: Number of chunks picked by initial search (default 20)
  • top_n: Number of chunks kept after reranking (default 10)

Usage

Document Ingestion

Upload files via the UI (PDF, TXT, DOCX, MD). ChatPilot supports OCR automatically. Documents are indexed instantly and available for the RAG pipeline.

Web & Scholar Research

  • WebResearch: Uses Tavily/Google to synthesize answers from the web.
  • Scholar Research: Uses OpenAlex for multi-step academic research.