AI Features

AlgoQuill is AI-native — meaning AI is not a bolt-on feature but the foundation the whole product is built on. Here is what each AI feature does and how to use it.

AI Chat Widget

Every AlgoQuill documentation site includes a chat widget in the bottom-right corner. Visitors click it, type a question in plain English, and get a direct answer with a citation to the relevant documentation page. The AI only answers from your published documentation — it never makes up information.

What it looks like to your visitors

A floating chat bubble appears in the bottom-right corner of every page. When clicked, a chat panel slides up. The visitor types their question, hits Enter, and the AI responds within a few seconds. Each answer includes a "Source: [page name]" link so the visitor can read more.

How to use it well

1. Publish enough content first

The widget only knows what's in your published documentation. If you have 1 published page, it will struggle to answer most questions. Aim for at least 5-10 well-structured pages covering your most common topics before promoting it.

2. Use clear headings (H2 and H3)

The AI uses your headings as semantic anchors. A page with ## Authentication followed by ### API Keys will get cited correctly when someone asks "how do I authenticate?". Headings without context — like just "Setup" — make citation harder.

3. Test with the questions visitors actually ask

Before launch, open your hosted docs in incognito and ask the widget the 5 most common questions you expect visitors to have. If any answer is wrong or missing, fix the source page. The widget is only as good as the docs behind it.

4. Monitor what visitors actually ask

Go to Dashboard → AI Insights to see every question asked. Questions with no good answer are gaps in your docs — fix them. Questions asked frequently mean those topics deserve a dedicated page.

Customising the widget

In Dashboard → AI Chat Config you can configure:

Greeting message — What the visitor sees when they open the chat. Default: "Hi! Ask me anything about the documentation."
Widget colour — Defaults to your project's brand colour. Override in Settings → Branding.
Suggested questions — A list of pre-filled prompts visitors see when they open the widget. Pick 3-5 questions that showcase what the docs cover.
Disable on specific pages — Turn the widget off for individual pages where you don't want it (e.g., legal pages).

Embed on your own site

Add the chat widget to any external site with one script tag. Get your embed code from Dashboard → Embed Widget.

<script src="https://algoquill.ai/embed.js" data-project="YOUR_PROJECT_ID"></script>

Drop this before </body>. The widget loads as a floating bubble. It works on any site — your marketing page, your app, your blog. Same AI brain, same documentation source.

Pro tip: The widget is most valuable on landing pages and pricing pages — where prospects have specific questions before signing up. Adding it there can reduce sales-conversation friction significantly.

AI Documentation Generation

AlgoQuill generates complete documentation pages from your codebase. It reads your actual code — not just the README — and produces structured, accurate documentation.

RAG Docs (recommended)

RAG (Retrieval Augmented Generation) works by first indexing your entire repository, then generating docs using the most relevant code chunks as context. This produces more accurate and specific output than sending the full codebase.

1
Sync your repo
AlgoQuill indexes your codebase into searchable chunks. Run once, re-run when the code changes significantly.
2
Choose your audience
Developers, Customers, Enterprise/Internal, or Portfolio/Recruiters. Each produces completely different documentation from the same code.
3
Specify a topic
"Authentication flow", "Payment API", "Database models" — be specific. The more specific the topic, the better the output.
4
Choose a type
API Reference, Quickstart, Tutorial, or Reference. This determines the structure of the generated page.
5
Generate and review
AlgoQuill retrieves the most relevant code context, generates the page, and saves it as a draft for your review.

AI Generate (direct)

The AI Generate page lets you generate from a prompt, from GitHub directly, from pasted code, or from an OpenAPI spec. Use this for documentation that doesn't require deep code context — getting started guides, conceptual explanations, migration guides.

Generate Full Suite

The "Generate All" button in RAG Docs generates 5 pages at once: Quickstart Guide, API Reference, Components Overview, Configuration Guide, and Troubleshooting. All saved as drafts for your review.

Important: Generated pages are always saved as drafts. You must manually publish each one. Only published pages are visible on your public documentation site.

Drift Detection

Drift detection compares your published documentation against your current codebase and identifies mismatches. It catches four types of issues:

Outdated
Documentation describes something that has changed — a renamed function, removed parameter, different behaviour.
Missing
Code has functions, endpoints, or components that have no documentation page at all.
Deprecated
Documentation describes something that no longer exists in the codebase.
Inconsistent
Naming or usage in docs doesn't match the code — docs say onClick but code uses onPress.

Go to Drift Detection in the dashboard and click Run Analysis. Results show the specific page affected, the specific code file that caused the issue, and a suggested fix.

AI Insights

AI Insights tracks what visitors ask the chat widget and uses that data to identify gaps in your documentation. The most valuable metrics:

Unanswered questions
Questions the AI couldn't answer confidently — these are your documentation gaps.
Top search queries
What users search for most. If they're searching for it, it needs a dedicated page.
Zero-result searches
Searches that returned nothing — strong signal that content is missing entirely.