I have written myself a web fronetend using langchain that uses a json based agent to do advanced tool use (websearch, calculations, document summarisation/information extraction, etc) and chain of thought reasoning. At the moment im using openai gpt-4-turbo for the main agent and gpt-3.5-turbo for some of the summation tools but openai models are limited in capabilities (wont talk about certain topic to the point of disregarding required output json formatting). What service can use that will run a preferably foss model as powerfull as gpt-4 that is pay per use with a langchain integration? here’s the project https://github.com/muntedcrocodile/Sydney [https://github.com/muntedcrocodile/Sydney] EDIT: Added link to project
  • MrPoopyButthole@lemmy.worldEnglish
    1·
    6 months ago

    The only really capable models are from OpenAI, Anthropic, and Meta. They all have guard rails of some kind and you need to use a supervisor agent to decide if the output is as you wanted if you can’t use something more basic like a JSON parser. Google Bard/Gemini output is not that high quality IMO.

    • 🇦🇺𝕄𝕦𝕟𝕥𝕖𝕕𝕔𝕣𝕠𝕔𝕕𝕚𝕝𝕖@lemm.eeOPEnglish
      1·
      6 months ago

      I have a guardrails ensuring proper json output that works 99% of the time except chatgpt has been so lobotomised that it ignores that for the I can’t answer that default responces. What about mistral models I’ve played with then but haven’t found any services that let me just point at an api and fire.