> ## Documentation Index > Fetch the complete documentation index at: https://www.doc-reviewer.site/llms.txt > Use this file to discover all available pages before exploring further. # Doc Reviewer: evaluate instruction quality with AI > Doc Reviewer is a desktop app for technical writers that uses LLMs to automatically detect and evaluate the quality of instructions in your documentation. Doc Reviewer is a desktop application that scans your technical documentation — PDF, Word documents, Markdown files, and web pages — finds instructions and evaluates them against a configurable set of criteria using an LLM. Each instruction receives a rating with visual color indication and recommendations for improvement. Everything runs locally on your computer — no cloud account, no data transfer, no subscription. Download the `.exe`, install Chromium for web page support, and get up and running in minutes. Connect an LLM, upload your first document, and run your first evaluation step by step. Group documents by product, generate product context, and improve evaluation accuracy. Understand how Doc Reviewer scores instructions and what each color result means. Connect OpenAI, Anthropic, Ollama, any OpenAI‑compatible API, or add your own custom model. ## What Doc Reviewer does When you upload a document, Doc Reviewer parses it into sections and classifies each one using morphological analysis. Sections that contain instructions — procedural steps, task descriptions, action sequences — are sent to the LLM for evaluation. The LLM checks each instruction against criteria such as heading style, step structure, completeness, and result descriptions, then returns a score and recommendations. Results are color‑coded for fast triage: * **Green** — the instruction meets all criteria * **Yellow** — minor issues that are easy to fix * **Orange** — problems that may affect usability * **Red** — critical issues that need immediate attention You can review results in the document tree, mark false positives, export to XLS or other formats, and save evaluation snapshots to compare quality across document versions. ## How it works Download `doc-reviewer.exe` and run it. The app starts a local server on `localhost:8000` and automatically opens the interface in the browser. Go to **Settings → Models** and add your LLM provider — OpenAI, Anthropic, Ollama, any OpenAI‑compatible API, or a custom model you have configured. Your API key is stored locally in the database. Create a project for a product, then upload PDF, DOCX, Markdown, or TXT files — or paste a URL to load a web page directly using the built‑in headless browser. Click **Evaluate** to analyze all detected instruction sections. Review color‑coded results with recommendations, mark any false positives, and export to XLS or other formats. ## Key features Parse PDF, DOCX, Markdown, and TXT files. Load web pages by URL using Playwright and a headless Chromium browser — including JavaScript‑rendered pages and SPAs. Doc Reviewer identifies instruction sections automatically using morphological analysis. No manual tagging or preprocessing required. Use the built‑in criteria set or create your own. Edit and manage criteria directly in **Settings → Criteria** using Markdown format. Generate a product context summary for each project. The LLM uses it to understand your product's terminology and audience when evaluating instructions. Automatically extract a glossary of product‑specific terms from each document. The glossary is added to LLM prompts to reduce false positives and improve relevance. When a document is replaced, Doc Reviewer matches sections with the previous version. Unchanged sections reuse prior evaluations, partially changed sections receive diff hints, and only new content is re‑evaluated. Save evaluation results as named snapshots and compare quality across document versions to track improvements over time. All data is stored in a local SQLite database next to the `.exe`. Nothing leaves your machine except LLM API calls to the provider you configure.