Developer expertise modifications quickly. Many builders and the Cisco DevNet group make the most of Generative AI instruments and language fashions for code era and troubleshooting.
Higher information – higher mannequin completion
The primary problem for GenAI customers is discovering legitimate information for his or her prompts or Vector Databases. Builders and engineers must care in regards to the information they plan to make use of for LLMs/GenAI interplay.
OpenAPI documentations is now obtainable to obtain
The OpenAPI documentation is a specification that defines a normal strategy to describe RESTful APIs, together with endpoints, parameters, request/response codecs, and authentication strategies, selling interoperability and ease of integration.
We at Cisco DevNet care about builders’ expertise and wish to make your expertise working with Cisco APIs environment friendly and with minimal growth/testing prices.
Yow will discover hyperlinks to OpenAPI documentation in JSON/YAML format right here: Open API Documentation web page and Search associated product API – Navigate to API Reference -> Overview part in left-side menu
Observe: Some API documentation can include a number of OpenAPI Paperwork
For which objective you need to use associated OpenAPI documentation as part of immediate/RAG:
- Assemble code or script that makes use of associated Cisco API
- Discover associated API operations or ask to repair present code utilizing the knowledge within the API documentation
- Create integrations with Cisco merchandise via API
- Create and check AI brokers
- Make the most of associated Cisco OpenAPI documentation domestically or utilizing authorized AI instruments in your group.
Structured vs Unstructured information
I’ve in contrast two LLM mannequin completions with a immediate that accommodates two components.
The primary a part of the immediate was the identical and contained the next data:
Primarily based on the next API documentation, please write step-by-step directions that may assist mechanically tag roaming computer systems utilizing Umbrella API.
Excessive-level workflow description:
- Add API Key
- Generate OAuth 2.0 entry token
- Create tag
- Get the checklist of roaming computer systems and determine associated ‘originId’
- Add tag to units.
API documentation:
Second half:
- In a single case, it accommodates copy and paste information instantly from the doc,
- The opposite one accommodates LLM-friendly structured information like OpenAPI paperwork pasted one after the other
A part of CDO OpenAPI documentation
Claude 3 Sonnet mannequin completion. Immediate with OpenAPI paperwork
Claude 3 Sonnet mannequin completion. Immediate with copy and paste information
Advantages of utilizing LLM-friendly documentation as part of the immediate
I’ve discovered that mannequin output was extra correct once we used OpenAPI paperwork as part of a immediate. API endpoints supplied in every step have been extra correct. Suggestions in sections like “Get Record of Roaming Computer systems” include higher and extra optimum directions and API operations.
I’ve examined this with different foundational fashions, and mannequin completion was extra correct after I used the OpenAPI doc as part of the immediate.
Some hyperlinks on the Cisco APIs OpenAPI Paperwork
- https://developer.cisco.com/docs/cisco-defense-orchestrator/overview/
- https://developer.cisco.com/docs/cloud-security/token-overview/
- https://developer.cisco.com/docs/cisco-xdr/automation-api-guide/#sample-code
Share:
