🧬

AiScientist

UCI RUIC20 lab · single-cell research console

No account? Create one with your UCI email. Forgot your password? An admin can reset it.

← Back to sign in

Enter the 6-digit code we emailed you.

Resend code · Start over

🧬

AiScientist Console

SSH → UCI HPC3 · GPU · vLLM / Qwen3

Disconnected

Connect to HPC3

Cluster GPU waits for a free card. Your own API key is ready in seconds — your HPC3 account still runs the analysis jobs either way.

Provisioning

  • SSH login
  • vLLM detect / image
  • GPU allocation (Slurm)
  • vLLM serve + tunnel
  • Model ready
  • Live
Connection & compute
connected
GPU —

Step 2 · Duo


                
                

                  
                  
                

              


              

                

                  
Chats

                  
                

                

              


              
            


            

              

                
New chat

                
              

              
              

              

                
Connect to a GPU-backed Qwen3 on HPC3, then ask a research question.

              

              

                
                
                
                
                
                

                

                
                
                
                
                

                  
                  

                    
                    
                    
                    

                      
                      
                    

                    

                    
                    

                    
Recent uploads (click to add/remove)

                    

                  

                

                

                
                
                

              

              

                Advanced — force a preset pipeline (the PI chooses by default)
                

                  
                  

                

                
Optional. Tick preset pipelines to require them in this conversation (mandatory) — the PI still auto-adds the best-fit one on top. Leave empty to let the PI choose.

                
Require atomic skills (composable capabilities)

                

                
Optional. Tick atomic skills the run must apply (e.g. E-distance ranking) — they compose, so pick several. Leave empty and the agent fetches skills on demand.

              

              

              

                
                
                
              

            


            

              
Results

              

                
Run a pipeline to get a results bundle (.zip) and browse the report, figures, and tables here.

              

            

          

        


        
        

          

            
My datasets

            

            
Files you've uploaded to the server. Click "Use" to load one into a Research run.

            
Loading…

          

        


        
        

          

            
My runs

            
Past research-lab runs. Download a run's results bundle (.zip) — it contains the report, figures, tables, and the technical event log.

            
Loading…

          

        


        
        

          

            

              
              
            

            

            
User management

            
Create accounts, disable/enable them, reset a forgotten password, or delete an account. Admin only.

            

              
              
              
              
              
            

            

              search
              
              
            

            

              
IDUsernameEmailRoleStatusLast loginActions

              
Loading…

            

            

            

              
Skill review

              
Skills the lab learned on its own from earlier runs. Each is one run's code turned into a template, with that dataset's format and column names inside it, so no run can plan with it or use it until an admin approves it. Hand-written skills are not listed here; they are always available.

              

                
                
                
                
              

              
Loading…

            

          

        


        
        

          

            
My account

            

            
Change my password

            

              
              
              
            

            
          

        


        
        

          

            
System

            
What this AiScientist defines and what this server can actually run — generated live from the code.

            
Workflow graph (live — reflects the current code)

            

              
Loading…

              
Click a node to see its details.

            

            
Loading…

          

        

      

    

  


  

    

      

        
HPC3 storage

        
      

      

      
Total used: —

      

      
Loading…

      
Deletes run rm -rf on HPC3, scoped to your own directory only. This cannot be undone.

    

  


  
  

    

      

        
Cluster models

        
      

      

        Weight sets the cluster GPU can serve. Pick one in the LLM endpoint box. Switching
        during a session keeps your current card when the new model can run on it (only the model
        reloads, a few minutes); otherwise the card is released and a new GPU job is queued. The
        default is what a new session gets unless it picks another.
      

      

      
Loading…

      


      

        
Add a model

        
        
        
        
        
        
        
        

        

          
          
        

      

      

    

  


  
  

    

      

        
API endpoints

        
      

      

        Use your own LLM account instead of the cluster GPU. Your HPC3 login is still what runs
        the analysis. Most providers show a new key only once — paste it here and AiScientist
        keeps it, so editing an endpoint later never asks you to find that key again.
      

      
      

      
Loading…


      

        
Add an endpoint

        
        

        
        
        
        
        
        
        
        

        

          
          
        

      

    

  


  
  

    

      

        
This endpoint is outside UCI

        
      

      

      
Using it sends the following to that provider, on your own account:

      
    
      
    These stay at UCI:
    
      
      
      
      
        
        
      
      
    
      
  
      

  
      
    
      
      
      
        
      Preview
      
        
      
          ⬇ Download