This talk was presented at the AI The Docs online conference on April 4, 2024. We are thrilled to share the recording and the summary with you.
Visit the talk summary page to see all of the presentations from the conference.
Neelesh's presentation
In this presentation, Neelesh Pateriya from DevNet, Cisco's developer platform, explores the integration of advanced AI techniques to enhance developer experience through better API documentation and search functionalities.
Key Takeaways
Key Concepts: Using a coffee shop analogy, Neelesh explains how embedding and summarization work. LLMs learn quickly from extensive data, like a barista mastering coffee recipes faster than through years of experience. They improve developer portals by providing more intuitive and effective search and interaction features.
Semantic Search: Traditional keyword-based search often falls short for new developers who prefer natural language queries. By creating embeddings from content and building a vector database, semantic search translates user queries into vectors, enhancing search accuracy and relevance.
Incremental Development: The team adopted an incremental approach, starting with summarization, then moving to Q&A functionalities, and finally integrating chatbots. This method ensures that each feature delivers accurate and useful responses before proceeding.
API Documentation & Excellence: Cisco’s API Excellence program helps maintain consistency in API documentation. OpenAPI specs are transformed into self-sufficient API endpoints, vectorized, and embedded. Prompts are used to refine summaries, generate examples, and ensure compliance.
Practical Use Cases: Experiments include generating SEO metadata and social media content. The setup allows for easy expansion and reuse of components. Data chunking, particularly with large documents, and managing costs for training and inference remain significant hurdles.
Future Directions: Neelesh highlights ongoing exploration of cost-effective models like AWS Bedrock and evaluations of training and retrieval-augmented generation (RAG) approaches. The aim is to enhance semantic search, summarization, Q&A, chatbots, and authoring tools, making the developer portal a more valuable resource for both users and documentation authors.
Sign up to our Developer Portal Newsletter so that you never miss out on the latest API The Docs recaps and our devportal, API documentation and Developer Experience research publications.
