DotNetSearch

Reduces code indentation from 4-space (or configurable) to 1-space per indent level and removes blank lines. Saves 10-20% tokens with zero semantic change for brace-scoped languages. SAFE for: C#, Java, JavaScript, TypeScript, C, C++, Kotlin, Swift, Rust, Scala, PHP, Ruby, CSS, SCSS, JSON, XML. DO NOT USE on indentation-sensitive languages: Python, YAML, CoffeeScript, Pug/Jade, HAML, Makefile — compression WILL corrupt these. Accepts either an absolute file path or a raw code string. Returns compressed code plus metrics: originalLength, compressedLength, savingsPercent, line counts. Use when reading source files to minimize context window usage.

Decompiles a .NET type (class, interface, enum, struct, delegate) from the NuGet package assemblies referenced by a .NET project. Pass the short class name (e.g. 'Session') or the fully qualified name (e.g. 'Opc.Ua.Client.Session'). The tool scans the project's obj output folder — which contains all referenced NuGet DLLs after a build — so no assembly path is needed. Use this when the user or LLM needs to understand the API surface of a third-party or NuGet type without access to its source code. IMPORTANT: The target project must have been built at least once so the bin and obj folders are populated. If typeName is ambiguous (found in multiple assemblies), all matches are returned up to maxResults.

Extract embedded file attachments from a PDF file and save them to an output directory. PDFs can carry arbitrary embedded files (documents, spreadsheets, etc.) as attachments, distinct from the PDF's own embedded images. Returns the list of saved file paths and metadata.

Extract embedded raster images from a PDF file (e.g. photos, diagrams) and save them as PNG files to an output directory. Optionally limited to a page range. Returns the list of saved file paths with their source page and size. Note: this extracts embedded image objects, not vector graphics/text — use RenderPdfPageAsImage to rasterize an entire page as a picture instead.

Read the bookmark/outline tree (table of contents) of a PDF file. Returns each bookmark's title, target page number (1-based, when resolvable), and nested child bookmarks. Use this to understand a PDF's structure before reading specific pages.

Read all interactive AcroForm fields from a PDF file (text boxes, checkboxes, radio buttons, combo/list boxes, signature fields). Returns each field's name, type, and current value. Use this to extract data entered into fillable PDF forms.

Read hyperlinks and internal links from a PDF file, optionally limited to a page range. Returns each link's source page, on-page location, and target — an external URI, an internal destination page number, or a linked external file name.

Read the metadata of a PDF file: page count, per-page size, and document info (title, author, subject, keywords, creator, producer, creation/modification date). Use this first to plan which pages to read text/images from.

Read a text file from disk and return its contents. Useful when you need the full contents of a specific file after a search. Supports both absolute and relative paths, and can optionally read a specific line range.

Extract the text content of a PDF file, optionally limited to a page range. Returns text per page along with the overall extracted text. Note: scanned/image-only PDFs will return little or no text — use ExtractPdfImages or RenderPdfPageAsImage instead for those.

Render a single PDF page as a full picture (PNG) — a screenshot of the whole page including text, vector graphics and images. Useful for scanned documents or when embedded text/images aren't sufficient. Returns the saved image file path and pixel dimensions.

Search a local directory for files whose content matches a query keyword or phrase. Supports both exact keyword search and natural-language / semantic-style queries by tokenising the query, expanding CamelCase and snake_case identifiers, and ranking results with a BM25 relevance score. Returns the top matching lines with file path, line number, relevance score and content snippet. Pass 'repo' as an absolute or relative local directory path.

Find code chunks semantically similar to a specific location in a file. Use after `search` to explore related implementations, callers, or patterns. Pass file_path and line from a prior search result. Extracts a smart chunk around the given line and finds the most structurally similar code blocks across the entire repo using BM25 token scoring.