Bayard Documentation

Bayard is a Flask application that provides a natural language processing API endpoint. It leverages the power of Elasticsearch for document search and OpenAI's language models for generating contextual responses based on user input and relevant documents. The application is designed to be scalable and robust, with a focus on efficient data storage and retrieval.

‍

At its core, the Bayard App consists of a Flask server that exposes a single API endpoint: /api/bayard. This endpoint accepts POST requests with a JSON payload containing the input_text field, which represents the user's natural language input.

‍

python
@app.route("/api/bayard", methods=["POST"])
def handle_bayard_request():
    input_text = request.json.get("input_text")
    # ... process input_text ...

‍

Upon receiving a request, the application first searches Elasticsearch for relevant documents based on the input_text. Elasticsearch is a powerful search engine that allows for efficient and flexible document retrieval, ensuring that the most relevant context is provided to the language model.

‍

python
filtered_docs = search_elasticsearch(input_text)

‍

Next, the application leverages OpenAI's language model to generate a contextual response, taking into account both the input_text and the relevant documents retrieved from Elasticsearch. OpenAI's state-of-the-art natural language processing models are capable of understanding and generating human-like text, providing accurate and insightful responses.

‍

python
model_output_json = generate_model_output(input_text, filtered_docs)
model_output_data = json.loads(model_output_json)

‍

To ensure data persistence and enable potential future analysis, the application stores each model run in a PostgreSQL database hosted on Render. The database schema includes fields for the run_id, timestamp, input_text, and model_output, allowing for efficient storage and retrieval of model run data.

‍

python
with db_connection.cursor() as cursor:
    cursor.execute("""
        INSERT INTO model_runs (run_id, timestamp, input_text, model_output)
        VALUES (%s, %s, %s, %s)
        ON CONFLICT (run_id) DO UPDATE SET
            timestamp = EXCLUDED.timestamp,
            input_text = EXCLUDED.input_text,
            model_output = EXCLUDED.model_output;
    """, (run_id, timestamp, input_text, model_output_data['modelOutput']))
db_connection.commit()

‍

Finally, the endpoint returns a JSON response containing the runId, timestamp, modelOutput, and relevantDocuments, providing the client with the necessary information for further processing or display.

The Bayard App is designed with scalability and robustness in mind, leveraging industry-standard technologies and best practices for natural language processing, data storage, and API development. Its modular architecture and integration with powerful services like Elasticsearch and OpenAI make it a versatile solution for various natural language processing tasks.

Elasticsearch Utility OpenAI Utility WL1.0GP License Terms Ethical Use Policy for the Bayard Corpus