search

fiatrete / OpenDAN-Personal-AI-OS

OpenDAN is an open source Personal AI OS , which consolidates various AI modules in one place for your personal use.
https://opendan.ai
MIT License
1.65k stars 134 forks source link

Jarvis prompt hangs for me (What is the latest working status on main branch?) #127

Open smarath3 opened 1 month ago

smarath3 commented 1 month ago

I tried to build from latest 'main' this weekend after several months (Version: 0.5.2) (have been watching this exciting space and the updates!). After some build struggles (my suggestion->stick with Python 3.11 and don't update to 3.12); was able to build and run from source.

But the prompt response hangs! (log snippet below with some errors). Appreciate any help.
Since gpt-3.5-turbo-0613 is deprecated, I updated JARVIS_LLM_MODEL to gpt-3.5-turbo. But nothing changed. Not sure if its picking up what is set at JARVIS_LLM_MODEL or GPT4o based on @waterflier's recent changes to default llm_model (I can look more next week). But Jarvis right now is unusable for me!

Bigger question: Whats is the latest status 'working' status on main branch? Can anyone else who have tried it, chime in? Should we stick to 0.5.1 for now?

@waterflier looks like you are main (sole?) developer for past few months. Appreciate any recent updates you can share? How can the community help?

There is interest in this project. Personally for me, this is super exciting! Love the concept of Personal AI OS, feels like the future of how AI is going to be used with privacy and personal-use (models, assistants) in mind. Real value-prop, and would love to see this project grow, and succeed.

So any inputs, and anything else on how community can help would be appreciated.

Here is my aios_shell log snippet btw:

[2024-08-11 12:27:02,743]aios.agent.llm_context[ERROR]: load_function_set_from_config failed! enable function id aigc.text_2_image not found!
[2024-08-11 12:27:02,743]aios.agent.llm_context[ERROR]: load_from_config failed! load_function_set_from_config failed!
[2024-08-11 12:27:02,743]aios.agent.llm_context[ERROR]: load_function_set_from_config failed! enable function id system.shell.run_code not found!
[2024-08-11 12:27:02,743]aios.agent.llm_context[ERROR]: load_from_config failed! load_function_set_from_config failed!
[2024-08-11 12:27:02,744]aios.agent.agent_memory[ERROR]: load memory meta failed: [Errno 2] No such file or directory: '/Users/sanmar/myai/agent_data/Jarvis/memory/meta.json'
[2024-08-11 12:27:02,744]aios.agent.llm_process[INFO]: workspace is not none,enable workspace functions
[2024-08-11 12:27:02,837]aios.frame.compute_kernel[INFO]: compute_kernel get task: ComputeTask: 80b791e1f2d3481a8e62886323eedd81 ComputeTaskType.LLM_COMPLETION ComputeTaskState.INIT
[2024-08-11 12:27:02,837]aios.frame.compute_kernel[WARNING]: task ComputeTask: 80b791e1f2d3481a8e62886323eedd81 ComputeTaskType.LLM_COMPLETION ComputeTaskState.INIT is not support by any compute node
[2024-08-11 12:27:07,746]aios.agent.agent[INFO]: agent Jarvis self thinking start!
[2024-08-11 12:27:07,749]aios.agent.llm_process[INFO]: load_chat_history: no history messages,return NONE
[2024-08-11 12:27:07,749]aios.agent.llm_process[INFO]: prepare_prompt: no history messages,return NONE
[2024-08-11 12:27:07,749]aios.agent.llm_process[WARNING]: prepare_prompt return None, break llm_process
[2024-08-11 12:27:07,749]aios.agent.agent[ERROR]: llm process self thinking error:None
waterflier commented 1 month ago

Due to huge architectural changes, the main branch is now mainly used for daily development ~ but I will always try my best to keep main branch availability. I'll go back and take a look at your question later. The branch that is currently working stably is MVP 0.5.1~

You can take a look at some issues I wrote recently to learn about the difficult journey of 0.5.2. I adjusted some goals. Right now I am making a relatively large change in my local working directory to fully use the knowledge graph. I have entered the testing phase and will submit it in the next 1 or 2 weeks.

waterflier commented 1 month ago

Bug fixed, try again?

smarath3 commented 4 weeks ago

It works now! thanks for the fix, and for heads-down working on ambitious goal of 0.5.2!

On a side note, I was considering integrating/playing with Llama 3.1 and tinkering with code a bit. Wondering, do you think its better to wait for you large architectural changes to land?

Appreciate it.