Langbase

1. 1
   Define LLM Workflows with Pipes

   Design and implement a Langbase 'pipe' to orchestrate your LLM application's workflow. Chain together LLM calls, data transformations, and external API integrations to define a clear sequence of operations for tasks like query processing or content generation.
2. 2
   Integrate Conversational Memory

   Integrate Langbase's 'memory' feature to maintain conversational state and context across user interactions. Configure memory types (e.g., conversational) and link it to your application's sessions or pipes to ensure coherent and personalized experiences.

class LLMComponent:
    def __init__(self, name, model, template):
        self.name = name
        self.model = model
        self.template = template
        print(f"  - LLMComponent '{name}' created for model '{model}'")

class DataTransformer:
    def __init__(self, name, function, input_key, output_key):
        self.name = name
        self.function = function
        self.input_key = input_key
        self.output_key = output_key
        print(f"  - DataTransformer '{name}' created for function '{function}'")

class ExternalAPI:
    def __init__(self, name, endpoint, method, payload):
        self.name = name
        self.endpoint = endpoint
        self.method = method
        self.payload = payload
        print(f"  - ExternalAPI '{name}' created for endpoint '{endpoint}'")

class Pipe:
    def __init__(self, name, steps):
        self.name = name
        self.steps = steps
        print(f"\nPipe '{name}' defined with {len(steps)} steps:")
        for step in steps:
            print(f"    Step: {step.name} ({type(step).__name__})")
        print("This structure is ready for deployment to a platform like Langbase.")


# Define a simple LLM prompt component
prompt_component = LLMComponent(
    name="initial_prompt",
    model="gpt-4",
    template="You are a helpful assistant. User query: {query}"
)

# Define a data transformation component (e.g., summarizing)
summarizer_component = DataTransformer(
    name="summarize_output",
    function="summarize_text",
    input_key="llm_output",
    output_key="summary"
)

# Define an external API call component (e.g., saving to a database)
db_saver_component = ExternalAPI(
    name="save_to_database",
    endpoint="https://api.your-db.com/save",
    method="POST",
    payload={"content": "{summary}"}
)

# Chain them into a pipe
my_llm_pipe = Pipe(
    name="customer_query_processor",
    steps=[
        prompt_component,
        summarizer_component,
        db_saver_component
    ]
)

print("\n--- End of Pipe Definition ---")
# To run this example, save it as a .py file and execute: python your_file.py