olca

Olca

Orpheus Langchain CLI Assistant

• (very Experimental and dangerous)

olca.yml (sample)

api_keyname: 'OPENAI_API_KEY_olca'
model_name: 'gpt-4o-mini'
recursion_limit: 355
temperature: 0
human: true
tracing: true
system_instructions: |
  "You are going to explore the creation of capabilities by creating python code, saving and running these functions. Commit and push your work to the repository linked to issue #343. Store in ./src your code. "
user_input: |
  "generate fractal images and have great arguments to parametrize the fractal being generated.  You implement the funtions in the code (try using existing code if you find any in ./src, keep it simple and upgrade existing code if required). Make sure you validate your syntax (try running the command few times with output in ./generated_images (folder you create).  You tend just to badly escape the input and output of the command.  You can use the following libraries: PIL, numpy, matplotlib, and click.  You can use the following code as a base.  Ask the user for some input to customize the resulting fractal"

Some Variations I had success with

Create a story prompt

#....
system_instructions: You focus on interacting with human and do what they ask using tool 'human'
user_input:  Look in the file ./draft.txt for my story and guide me thru creating a story prompt.

Create a 3 Act structure

• coming up when I added knowledge base integration thru vector..... not today

Customize the Default appended prompt the agent gots

OLCA_SYSTEM_PROMPT_APPEND="My Overriding prompt appended"