wandbot  by wandb

Support bot for Weights & Biases' AI tools, running in Discord, Slack, ChatGPT, and Zendesk

Created 2 years ago
307 stars

Top 87.3% on SourcePulse

GitHubView on GitHub
Project Summary

WandBot is a technical support assistant for Weights & Biases' AI developer tools, designed to answer questions about their documentation and code. It targets developers using W&B products, offering an efficient way to get support directly within platforms like Discord, Slack, or via an API.

How It Works

WandBot employs a modular Retrieval Augmented Generation (RAG) pipeline. It uses Gemini Flash-2.0 for query expansion and GPT-4o for response synthesis, with Cohere's rerank-v3.5 for re-ranking. Key components include a hosted ChromaDB vector store, parent document retrieval, and sub-query answering for complex questions. This architecture prioritizes efficiency and contextuality, leveraging parallel LLM calls and metadata filtering for improved performance.

Quick Start & Requirements

  • Install: Run bash build.sh to build and install dependencies.
  • Prerequisites: Requires Python 3.12, OpenAI API Key, Cohere API Key, W&B API Key, and potentially Slack/Discord tokens.
  • Setup: Set environment variables as detailed in the README. Run bash run.sh to start the API and bots.
  • Docs: run.sh, Evaluation Config

Highlighted Details

  • Achieved 91.2% response correctness in English evaluations with v1.3.0 and v1.3.1.
  • Supports both English and Japanese documentation.
  • Integrates with Discord and Slack for seamless support.
  • Utilizes Weights & Biases Weave for performance monitoring and evaluation.

Maintenance & Community

  • Active development with frequent updates (v1.3.1 released recently).
  • Uses uv for dependency management and dotenv for environment loading.
  • Evaluation pipeline includes robust error handling and retries.

Licensing & Compatibility

  • No explicit license mentioned in the README.
  • Requires API keys from OpenAI, Cohere, and Weights & Biases, which may have their own usage terms.

Limitations & Caveats

  • The README does not specify a license, which could impact commercial use.
  • Setup requires multiple API keys and specific environment variable configurations.
  • Some features like web search are currently disabled.
Health Check
Last Commit

4 weeks ago

Responsiveness

1+ week

Pull Requests (30d)
0
Issues (30d)
0
Star History
2 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems") and Vasek Mlejnsky Vasek Mlejnsky(Cofounder of E2B).

super-rag by superagent-ai

0%
384
RAG pipeline for AI apps
Created 1 year ago
Updated 1 year ago
Starred by Li Jiang Li Jiang(Coauthor of AutoGen; Engineer at Microsoft), Elie Bursztein Elie Bursztein(Cybersecurity Lead at Google DeepMind), and
1 more.

AutoRAG by Marker-Inc-Korea

0.3%
4k
RAG AutoML tool for optimizing RAG pipelines
Created 1 year ago
Updated 1 day ago
Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Nir Gazit Nir Gazit(Cofounder of Traceloop), and
4 more.

llmware by llmware-ai

0.6%
14k
Framework for enterprise RAG pipelines using small, specialized models
Created 2 years ago
Updated 1 month ago
Feedback? Help us improve.