Learn Seq.

01

Install

seqc targets Linux x86_64. The reference environment is Ubuntu 24.04 LTS; on Windows, use WSL2.

curl -fsSL https://raw.githubusercontent.com/tomjnet/seq-lang/HEAD/install.sh | bash

The installer needs no root access. By default it downloads the pre-built package of the latest release and verifies its SHA-256 checksum; when no package fits, it builds from source. It never downloads a model and never installs system packages. Run it with --help for the options, including--from-source, --prefix, and --uninstall.

On Ubuntu, the build tools and Google Test come from one command:

sudo apt install build-essential cmake ninja-build libgtest-dev curl python3

llama-server comes from llama.cpp. Put it on your PATH, or name it with the model.llama_server setting.

02

Your first workflow

Five commands take you from nothing to published results.

seqc new top3Companycd top3Companyseqc doctorseqc model pullseqc src/main.seq

• seqc new creates the project. It needs no model and no compiler.
• seqc doctor says, one line per prerequisite, whether this host can compile and run workflows.
• seqc model pull downloads the model, about 1.1 GB, verifies its SHA-256, and writesseq.lock. It is the only command that downloads anything. A build never does.
• seqc src/main.seq validates the workflow, builds it or reuses the last accepted build, runs it, and lists the results.

Results:
  output/company.txt  (212 bytes)
  output/top3Company.png  (4.1 KiB)  final

A project keeps your source, your inputs, and everything the compiler made, side by side:

top3Company/
├── seq.lock                 pinned model (written by seqc model pull)
├── src/main.seq             the workflow
├── input/                   read-only data for the workflow
└── output/
    ├── company.txt          published results
    ├── top3Company.png
    └── temp/                compiler-owned: generated C, assembly,
                             executable, tests, and run records

03

A tour of the language

A Seq source file declares a model, a backend, a name, and an ordered list of steps. This is a complete program:

model("https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct")
backend.C()
name = "hello"

step step1():
    ask("print Hello from Seq")

Header declarations

Each of the three is required exactly once. They may come in any order, and all of them come before the first step.

Steps and requests

step <name>():
    ask("<request>")
    ask("<another request>")

• A workflow has at least one step. Steps run in source order.
• A step name is a label, not something that can be called. Names are unique.
• A step holds one or more ask("...") statements, one per line, indented exactly four spaces.
• Strings are double-quoted and single-line. A comment runs from # to the end of the line.
• Tabs are rejected everywhere. Indentation is zero or four spaces.

ask() is a compile-time directive. The model receives the whole workflow at once, so a later step can use what an earlier step creates. The generated executable never contacts the model.

Limits

Exceeding a limit is a validation error. Nothing is truncated.

Diagnostics

seqc reports every error it can find in one run, each with its line and column, and all of them before any model access. The codes are stable.

src/main.seq:6:9: error[E0204]: ask requires parentheses
        ask "print Hello from Seq"
            ^
    hint: write ask("...")

04

Commands

Options for seqc <file.seq>: --build-only, --rebuild,--force, --quiet. Every command accepts --set section.key=value.

05

When something fails

The exit status says which stage stopped. A failed build never replaces the last accepted build and never runs a stale executable.

06

Reference

• LanguageSyntax, limits, every diagnostic code, exit statuses.
• ArchitecturePipeline, model protocol, build cache, manifests, settings.
• RuntimeThe contract between seqc and generated code.
• SecurityThe sandbox, what it denies, and known limits.
• ExamplesThe projects this site shows, ready to build.