
2 · configure
3 · deploy
NewAvailable product — files, demo URL and secure download get mapped in EDF 0.3.
Ask-your-docs chat widget with cited answers
Add an "Ask the docs" chat to your documentation site. Visitors ask in plain language, DocsChat finds the right passages in your Markdown docs and answers with clickable citations. One script tag on the front end, one Vercel function (or a tiny Node server) on the back end, written in TypeScript with zero runtime dependencies. Uses Claude by default, any OpenAI-compatible model (OpenAI, Ollama, LM Studio, Groq) as an alternative, or no AI at all. The endpoint protects your wallet with an origin allow-list, rate limits and an off-topic gate that skips the model.
Origin: EWDS Store (epicwebdesignstudio.com) · Roadmap window: EDF 0.2


Placed in the private vault layout (deploy/vault/products). Downloads stay disabled until checkout is connected.
Derived by the mapping engine (family · layer · tags · dependency rules). Stored in the catalog as compatibility, dependencies, relations[].
| VERSION | STATE | NOTES |
|---|---|---|
| v1.0.0 2026-10-02 | AVAILABLE | Initial release. This package contains: Markdown to search index in one command: splits docs by headings, keeps code blocks intact, deep-links to the exact section, supports front matter, drafts, synonyms and English, Polish and German; no embeddings API or vector database; Widget in about 11 KB gzipped (Shadow DOM, no dependencies): streaming answers, safe Markdown with code blocks, numbered citations with source chips, suggested questions, copy, 👍/👎 events, retry, Stop, dark/light/auto themes, full screen on phones, EN/PL/DE, JS API and events; Serverless endpoint for Vercel plus a zero-dependency standalone Node server for any host; Claude by default or any OpenAI-compatible API (OpenAI, Ollama, LM Studio, Groq, OpenRouter); Answers only from your docs: strict citation rules, "not in the docs" replies, and an off-topic gate that skips the model call entirely; Protects your wallet: origin allow-list, per-IP minute and day limits, global daily cap, input and history limits, token cap, optional shared limits via Upstash Redis REST, no keys in the browser, no questions stored; Works without any AI: local mode answers in the browser from the index, and the server falls back to extractive cited answers when no model key is set. |
| → EDF 0.2 | Roadmap | Target window for the next stage. |
Verified for real with a local Ollama model (llama3.2:3b) in a real browser: that small model cited well and answered the signature question correctly but missed facts on two of three further questions, so use a stronger model (Claude Haiku/Sonnet or similar) for production. The Anthropic and OpenAI calls follow their public API formats and are tested against fake servers, not run with real keys; the Vercel deployment and the Upstash limiter are implemented and tested against fakes, not on real accounts. Retrieval is keyword-based (BM25 + synonyms), not semantic, so add synonyms for words your docs never use. AI answers can be wrong: the widget shows sources and a disclaimer. The sample Nimbus docs and the demo are fictional; the demo here runs in local mode (no AI, answers extracted in your browser). Requires Node 20.19+ for the tooling.
Standard (personal or commercial use in one end project or your own business) · Extended ×3 (SaaS, unlimited client projects, finished products only). Agency licence comes later.
Personal: use on websites and docs you own. Commercial: use for one client or company. Extended: unlimited sites, clients and products. No resale, redistribution or hosting as a service for third parties.
Activated by 1.0 (download) and 2.0 (one-click deploy). Download will be a 10-minute signed URL from the R2 vault — never a public /files/product.zip.



