Add a Cmake build

.gitignore

@@ -16,3 +16,4 @@ install.sh
 /test/*/result
 /test/*/*/result
 /.vscode/**
+/build

CLAUDE.md

@@ -0,0 +1,118 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## What is Pheniqs
+
+Pheniqs (PHilology ENcoder wIth Quality Statistics) is a multithreaded C++11 barcode classifier for high-throughput next-generation sequencing. It demultiplexes samples using probabilistic Bayesian classification and supports FASTQ, SAM/BAM/CRAM formats.
+
+## Build Commands
+
+### CMake (preferred, current branch is `cmake_conversion`)
+
+```bash
+mkdir build && cd build
+cmake ..
+make -j4
+```
+
+**Key CMake options:**
+- `-DBUILD_STATIC=ON` — static binary
+- `-DVENDOR_HTSLIB=ON` — build htslib locally (also set for its deps if vendoring any)
+- `-DVENDOR_ZLIB=ON`, `-DVENDOR_BZIP2=ON`, `-DVENDOR_LZMA=ON`, `-DVENDOR_LIBDEFLATE=ON` — vendor compression libs
+- `-DVENDOR_RAPIDJSON=ON` — vendor header-only RapidJSON
+- `-DCMAKE_INSTALL_PREFIX=/path` — install prefix (default: `/usr/local`)
+
+**Constraint:** When using `BUILD_STATIC=ON`, if any of htslib's deps are vendored, `VENDOR_HTSLIB` must also be `ON`.
+
+Vendored dependencies are downloaded from the internet at build time and installed to `build/external/install/`.
+
+### Makefile (legacy)
+
+```bash
+make -j4
+make with-static=1 -j4   # static build
+```
+
+Dependencies must be installed at `$(PREFIX)` (default `/usr/local`).
+
+### Generated files
+
+Two headers are generated at build time into the build directory:
+- `version.h` — from git describe via `cmake/GenerateVersionHeader.cmake`
+- `configuration.h` — from `configuration.json` via `tool/pheniqs-configuration-api.py header`
+- `_pheniqs` — zsh completion, also from `configuration.json`
+
+## Running Tests
+
+### Unit and integration tests (C++)
+
+All tests live in `test/unit/` and run via a single binary `pheniqs_unit_tests`. Must be run from the repository root.
+
+```bash
+# from build dir
+cmake --build build --target test     # build deps + run all tests
+cmake --build build --target test.unit  # same (alias)
+```
+
+Test files:
+- `test_phred.cpp` — `PhredScale` lookup table correctness
+- `test_sequence.cpp` — `Sequence`, `ObservedSequence`, `Observation`
+- `test_demux.cpp` — full demultiplexing pipeline (calls `Pipeline` in-process with BDGGG test data, checks SAM output at `test/BDGGG/result/demux_test_output.sam`)
+
+
+## Architecture
+
+### Entry point and pipeline
+
+`pheniqs.cpp` → `Pipeline` (pipeline.h/cpp) → reads config, manages job queue → `Job` (job.h/cpp) → `Transcode` (transcode.h/cpp) orchestrates reading, classifying, and writing reads.
+
+### Classification algorithms (three decoders)
+
+All decoders inherit from `Decoder<T>` template:
+
+| File | Class | Algorithm |
+|------|-------|-----------|
+| `naive.h/cpp` | `NaiveMolecularDecoder` | Direct extraction, no classification |
+| `mdd.h/cpp` | `MdDecoder<T>` | Minimum Distance (edit distance + quality masking) |
+| `pamld.h/cpp` | `PamlDecoder<T>` | Probabilistic Accurate ML — Bayesian with confidence scores |
+
+`PamlDecoder` is the primary, publication-described algorithm. `MdDecoder` is the fallback edit-distance approach.
+
+### I/O layer
+
+- `fastq.h/cpp` — FASTQ reader
+- `hts.h/cpp` — HTSLib wrapper (BAM/SAM/CRAM)
+- `feed.h/cpp` — abstract feed interface
+- `proxy.h/cpp` — data source proxy
+
+### Core data model
+
+- `read.h/cpp` — read object
+- `sequence.h/cpp` — DNA sequence + quality scores
+- `barcode.h/cpp` — barcode definitions with concentration priors
+- `auxiliary.h/cpp` — SAM auxiliary tags (RG, BC, QT, etc.)
+- `phred.h/cpp` — Phred quality scale math and substitution lookup tables
+
+### Configuration and CLI
+
+- `interface.h/cpp` — CLI parsing
+- `configuration.json` — canonical definition of all CLI options; used to generate `configuration.h` and zsh completion
+- `json.h/cpp` — RapidJSON wrapper
+- `atom.h/cpp` — JSON element abstraction
+
+### Selector/transform pipeline
+
+`selector.h/cpp` and `transform.h/cpp` implement a rule engine for extracting barcode segments from arbitrary positions in reads. `multiplex.h/cpp` handles the multiplexing/demultiplexing logic on top of decoded barcodes.
+
+## Dependencies
+
+| Library | Version | Purpose |
+|---------|---------|---------|
+| htslib | 1.23 | BAM/CRAM support (primary dep) |
+| zlib | 1.3.1 | Compression |
+| bzip2 | 1.0.8 | Compression |
+| xz/liblzma | 5.8.2 | Compression |
+| libdeflate | 1.15 | Optional, faster gzip |
+| RapidJSON | 1.1.0 | Header-only JSON |
+| pthread | system | Multithreading |