ACs for MVP

[Tamsen1995]

Functional Requirements

• [x] User Input Handling:

  • [x] The API must accept user input in JSON format.
  • [x] The input must include a list of messages, where each message contains a role (system, user) and content.

• [x] Query Extraction and Validation:

  • [x] The system must extract meaningful keywords from the user's input.
  • [x] If the user input is empty or invalid, the system must return a Bad Request error with a descriptive message.

• [x] SerpApi Integration:

  • [x] The system must call SerpApi with the extracted keywords from the user input.
  • [x] If SerpApi returns an error, the system must handle it gracefully and return an Internal Server Error with a descriptive message.
  • [x] The system must parse the SerpApi response and extract relevant information.

• [x] OpenAI Integration:

  • [x] The system must call the OpenAI API with the user input and the data obtained from SerpApi.
  • [x] The OpenAI API call must include a predefined system message to set the context.
  • [x] If the OpenAI API returns an error, the system must handle it gracefully and return an Internal Server Error with a descriptive message.
  • [x] The system must return the OpenAI API response to the user.

Non-Functional Requirements

• [ ] Performance:

  • [ ] The API must respond to requests within 3 seconds under normal load conditions.
  • [ ] The system must be able to handle at least 100 concurrent requests without degradation in performance.

• [ ] Scalability:

  • [ ] The system architecture must support horizontal scaling to handle increased load.
  • [ ] The API design must allow for easy addition of new features or integrations in the future.

• [ ] Security:

  • [x] API keys for SerpApi and OpenAI must be stored securely and not hardcoded in the source code.
  • [ ] The system must validate all incoming requests to ensure they are well-formed and do not contain malicious content.
  • [x] Sensitive data in logs and error messages must be masked or omitted.

• [ ] Maintainability:

  • [ ] The codebase must follow clean code principles and be modular, with distinct responsibilities for each module.
  • [ ] The system must include comprehensive unit and integration tests to ensure reliability.
  • [ ] The project must include clear documentation for developers, including setup instructions and code comments.

Error Handling and Logging

• [ ] Error Responses:

  • [ ] The system must return appropriate HTTP status codes for different types of errors (e.g., 400 for Bad Request, 500 for Internal Server Error).
  • [ ] Error messages must be descriptive and provide enough information for troubleshooting without exposing sensitive information.

• [ ] Logging:

  • [ ] The system must log all API requests and responses for auditing and troubleshooting purposes.
  • [ ] Error logs must include stack traces and contextual information to aid in debugging.

API Endpoint Requirements

• [x] Chat Completions Endpoint:

  • [x] The endpoint must be defined at /chat/completions.
  • [x] The endpoint must accept POST requests with JSON data.

• [ ] Request and Response Structure:

  • [ ] Request:
  • [x] The request must include a model field specifying the OpenAI model to use.
  • [x] The request must include a messages field containing a list of messages with roles and content.
  • [ ] An optional query field may be included for additional search context.
  • [x] Response:
  • [x] The response must include a response field containing the OpenAI-generated content.

Configuration and Environment

• [ ] Environment Variables:
  • [x] The system must read API keys and other configuration settings from environment variables.
  • [x] The system must include a .env file for local development with placeholders for required environment variables.

Documentation

• [x] README:
  • [x] The project must include a comprehensive README file with setup instructions, usage examples, and a description of the project structure.
  • [ ] The README must provide instructions for running the project locally and deploying it to a production environment.

Deployment

• [ ] Deployment Instructions:
  • [ ] The README must include clear instructions for deploying the API to a production environment.
  • [ ] Deployment scripts or configurations must be provided for common environments (e.g., Docker, cloud services).

[Tamsen1995]

Test with the following Python code:

from openai import OpenAI

client = OpenAI(
    base_url=<DEPLOYMENT_ENDPOINT>,
    api_key=<API_KEY_FOR_DEPLOYMENT>
)

completion = client.chat.completions.create(
    model="<ONLINE_MODEL_ID>",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is the top news about Gen AI today?"}
    ]
)

print(completion.choices[0].message)

[Tamsen1995]

• [ ] Make sure to find a better way to extract the query params from the user (Right now we're just accessing the indexes directly)
• [ ] Maybe we can figure out a way to make the web crawled results better. Right now it seems to only be using snippets from the different results returned by serpAPI

[Tamsen1995]

Stretch Goal:

• [ ] Feed the extracted keywords into an llm to help construct a better search query, before passing it into serpAPI