Echoes

Set of tools to allow easy implementation of Ollama LLM on Python, with context and memories.
The project is still in development, here are the aimed features;

✅ Model Init: Introduces the Model type that allows interaction with Ollama LLM.
✅ Context: Able to follow the flow of the conversation and use previous elements used to evaluate the context and store it in a file in case of reboot.
✅️ Memory/Memories: Will be able to remember key elements of a conversation to re-use later (Experimental feature, still in testing, have bugs regarding the assignment of existing user_id, create double memories).

All those features needs work/rework/upgrades, so far it's a very minimalistic version, a proof of concept.

Tested Implementations

import echoes

user = {
    'id': "",                               # Define user id in program.
    'name': "",                             # The user's name.
    'favorite': "",                         # What the user's love.
    'dislike': "",                          # What the user's hate.
    'relationship_lvl': "",                 # The relationship with the AI.
}

model_data = {
    'name': "LLama",                        # Attributing a name to the LLM (anything can be assigned).
    'model': "llama3.2:latest",             # Specify the LLM model.
    'persona': "",                          # Adds the persona of the model.
    'emotions': "",                         # The current emotion the AI is feeling (WORK IN PROGRESS NEED TO IMPLEMENT IN MODEL.PY).

    'user': user,                           # The user's information that are relevant to the generation.

    'context': [],                          # Store the conversation.
    'context_file': "context.json",         # Specify the file that should be created for remembering context, if the feature "is_remembering", None or '' values can be assigned.
    'context_enable': True,                 # Define if a file should be created containing the context for the LLM to remember when rebooted.

    'memories': {},                         # Where the memories will be stored.
    'memories_file': "memories.json",       # The filename for the memory storage on drive.
    'memories_enable': True,                # Specify if the LLM will create memories to remember.

    'instructions': "",                     # Define the instructions the AI should follow when generating a prompt.
}

model = echoes.create_model(model_data)
model.talk()

Will allow the user to input in the console the prompt for the AI to reply, creating a back and forth conversation.
You can also replace the last line with the tested function to generate the output of an input.

model = echoes.create_model(model_data)
output = asyncio.run(model.generate_response("Hello!"))

Will display in the console the LLM answer to "Hello!" and store it to the "output" variable as an 'str' for later uses.