1.3.1. Other Potential Components - SamuraiBarbi/jttw-ai GitHub Wiki

AI Memory

• LongMem
• MemGPT
• AutoGen

---

AI Assistants

• LocalAGI
• Chestshire-cat AI

---

AI Application Builders

• Flowise
• Anchoring AI
• OpenCopilot

---

AI Single Agent

• BabyAGI
• AutoGPT
• GPTEngineer
• TaskWeaver

---

AI Multiple Agents

• Crew AI
• Langroid
• Langchain
• MetaGPT
• Superagent
• AutoGen

---

AI Machine Vision

• VideoLLaVA
• ShareGPT4V
• ML-Farret

---

AI Coders

• Tabby
• Deepseek Coder
• Aider
• Sweep AI

---

AI Servers

• Llama.cpp
• GPT-Fast
• vLLM
• LM Studio
• Ollama
• Text Generation WebUI
• KobaldCpp
• LocalAI
• JanAI

---

AI Benchmarking/Evaluation

• PromptFoo
• ChainForge
• Fiddler-Auditor

---

AI Proxies

• LiteLLM
• LLMClient

---

AI API

• Gorilla OpenFunctions could be used to make function calling abilities available to models
  • Demo 1

---

AI Text to Speech

• StyleTTS2
• Bark
• Coqui
• TortioseTTS
• EdgeTTS
• RVC-TTS
• Apollo RVC-TTS
• Piper
• RVC-TTS Webui

---

AI Distributed Inference

• Ray
• Petals

---

AI Generative Image/Video Creation

• Comprehensive Resource

  • https://github.com/showlab/Awesome-Video-Diffusion

• Additional Resources:

  • Stable Diffusion Subreddit
  • Unstable Diffusion Subreddit
  • ComfyUI Subreddit
  • AI Artist Workflows Subreddit
  • CivitAI

• Stability Matrix

• Stable Diffusion WebUI

• ComfyUI

AI RAG - Chat With Documents

• Embed Chain

• MemGPT

• PrivateGPT could be used to ingest documents and query the language model about information from those documents

  • Documentation
  • Demo 1
  • Thoughts: I’ve gotten this working on Linux. Installation is somewhat fussy. Feeding it my 1080 ti 11GB it seemed to take a reasonable amount of time to ingest, and embed documents. I feed it a text document containing erotic fiction to test the models censor sensitivity and if it’s able to summarize the document, tell me specific details about characters, what it considers the last sentence of the document to be, and a number of other inquiries. Its answers were half right. I then fed it a short story of roughly 2930 words and questioned it about characters, and aspects of the story. It would describe characters in broad terms, and exhibited issues correctly identifying how the story ended despite being able to identify the last sentence of the story. Ultimately tests using a short story in .txt format were disappointing.

• LocalGPT could be used to ingest documents and query the language model about information from those documents

  • Documentation
  • Demo 1, Demo 2
  • Thoughts: I’ve gotten this working on Linux. Installation is somewhat fussy. Feeding it my 1080 ti 11GB it seemed to take a reasonable amount of time to ingest, and embed documents. I feed it a text document containing erotic fiction to test the models censor sensitivity and if it’s able to summarize the document, tell me specific details about characters, what it considers the last sentence of the document to be, and a number of other inquiries. Its answers were more correct than PrivateGPT. I then fed it a short story of roughly 2930 words and questioned it about characters, and aspects of the story. It would describe characters in broad terms, and exhibited issues correctly identifying how the story ended despite being able to identify the last sentence of the story. Ultimately tests using a short story in .txt format were disappointing.