docs: link upstream projects in README and add PDF.js roadmap item

Author: sousekd

README.md

@@ -54,9 +54,9 @@ flowchart LR
 
 ## What Ships Today
 
-- HTTP adapters for Open WebUI external web loader requests and limited Jina Reader-style `GET /r` requests.
-- A default `full` pipeline with source loading, Readability, mdream HTML-to-markdown conversion, optional LLM clean/summarize passes, URL verification, truncation, and optional XML diagnostics.
-- Source providers for native HTTP fetch, Firecrawl, and Docling.
+- HTTP adapters for [Open WebUI](https://github.com/open-webui/open-webui) external web loader requests and limited [Jina Reader](https://github.com/jina-ai/reader)-style `GET /r` requests.
+- A default `full` pipeline with source loading, [Readability](https://github.com/mozilla/readability), [mdream](https://github.com/harlan-zw/mdream) HTML-to-markdown conversion, optional LLM clean/summarize passes, URL verification, truncation, and optional XML diagnostics.
+- Source providers for native HTTP fetch, [Firecrawl](https://github.com/firecrawl/firecrawl), and [Docling](https://github.com/docling-project/docling).
 - Docker and GHCR image support.
 
 The `smoke` pipeline is also shipped for testing and debugging. It uses only native HTTP fetch, aggressive mdream conversion, and truncation.

docs/ROADMAP.md

@@ -24,15 +24,16 @@ Current provider implementations are intentionally few. The interfaces exist so
 ## Coming soon
 
 1. **Playwright source provider** to remove the hard dependency on a running Firecrawl instance for HTML pages.
-2. **LLM pass auto-repair.** Repair URLs and code blocks, instead of mere detection.
+2. **Support PDF.js** to extract text from PDF documents without an external service.
 3. **Diagnostics footer cleanup.** Consolidate the `<loader_info ... />` payload for consistency and readability.
 4. **Better defaults and prompts.** Test and tune what ships.
 
 ## Short term
 
 Near-term additions once the work above settles.
 
-- **Docling content transformer** to convert PDFs, Office documents, and other document-type files to markdown.
+- **LLM pass auto-repair.** Repair URLs and code blocks, instead of mere detection.
+- **Jina Reader source provider** to get markdown from a self-hosted Jina Reader instance.
 - **Crawl4AI source provider** as a potentially better alternative to Firecrawl.
 
 ## Mid term