aidoc

AI for Documents

AI assisted document generation

aidoc is a complete suite to build complex targetted documents from an overarching pool of both curated and AI generated document stubs.

The core philosophy of aidoc is to treat documentation as a structured, version-controlled engineering artifact. By utilizing the Sphinx ecosystem as the underlying document engine, we move away from prescribed editors (ie MSWord) and supported output formats, to edit and build with any tool; and a multitude of publishing formats.

The kind of problem aidoc solves is any document that must contain both static and dynamic components. Bid preparation say, where we might include our corporate overview and use-cases; but product/pricing as per a RFP. Legals and licensing agreements, where elements have been signed-off/reviewed (tagged in your git). Even job hunting - there are many such cases.

Sphinx is incredibly flexible, and you can build/define custom tags natively for your Restructured Text. Despite this, it can be frustrating; so aidoc incorporates a Jinja2 preprocessor: simply embed Jinja into your .rst and rename the file to .j2.

Sphinx as a dedicated documentation platform has considerable advantages. There are many validators available to ensure you produce a professional product: spelling and link checkers ensure mistakes are not sent to clients and customers. The Sphinx is treated like code and is very ameniable to CI/CD approaches.

aidoc includes a sophisticated prompt engine where you can create and edit prompts to produce your tags, and create custom documents. It is simply a matter of configuration settings against LiteLLM and/or LangFlow. Telemetry can be sent to MLFlow. Using these tools, you can prompt-engineer your way to quite stunning results.

aidoc is a command line tool by design. It works great on the desktop for an individual; but it also works well running out of Jenkins or AirFlow. It has an excellent web-crawler - where it can ingest url’s and use these as the inputs for prompts to prepare custom documents. Using AirFlow providers - such as IMAP to read a bids mailbox, or periodically hitting API’s in your funnel, you can easily create a documentation pipeline.