Run Locally

Three Ways to Parse

AILANG Parse offers three ways to parse documents. Choose what fits your needs:

Quick Start

Parse a document in two commands:

# Install AILANG
curl -fsSL https://ailang.sunholo.com/install.sh | bash

# Install AILANG Parse from the package registry
ailang install sunholo/ailang_parse

Now parse a document:

# Parse a DOCX file
ailang run --entry main --caps IO,FS,Env \
  ~/.ailang/cache/registry/sunholo/ailang_parse/*/docparse/main.ail your-file.docx

git clone https://github.com/sunholo-data/ailang-parse.git
cd ailang-parse
./bin/docparse data/test_files/sample.docx

Convert between formats:

# DOCX to HTML
./bin/docparse input.docx --convert output.html

# CSV to DOCX report
./bin/docparse data.csv --convert report.docx

# Markdown to PowerPoint slides
./bin/docparse notes.md --convert slides.pptx

# Any format to Quarto Markdown (for Quarto rendering)
./bin/docparse report.docx --convert report.qmd

Generate documents with AI:

# Generate a DOCX report from a prompt
ailang run --entry main --caps IO,FS,Env,AI --ai gemini-2.5-flash \
  docparse/main.ail --generate report.docx --prompt "Q1 sales report with revenue table"

Install AILANG

Step 1: Install the AILANG CLI

AILANG is a single binary with no dependencies. Install it for your platform:

# macOS / Linux
curl -fsSL https://ailang.sunholo.com/install.sh | bash

# Windows (PowerShell)
irm https://ailang.sunholo.com/install.ps1 | iex

Or download directly from GitHub releases for any platform.

Verify the installation:

ailang --version
# ailang v0.9.2 (or later)

Step 2: Install AILANG Parse

Install the parsing package from the AILANG registry:

ailang install sunholo/ailang_parse

This downloads all parsing modules and their dependencies. No build step needed.

Optional: Clone the Source

For development, testing, and benchmarks, clone the repository:

# macOS / Linux
git clone https://github.com/sunholo-data/ailang-parse.git
cd ailang-parse
./bin/docparse data/test_files/sample.docx

# Type-check all modules
./bin/docparse --check

# Windows
git clone https://github.com/sunholo-data/ailang-parse.git
cd ailang-parse
bin\docparse.cmd data\test_files\sample.docx

# Type-check all modules
bin\docparse.cmd --check

Batch Mode & Folder Parsing

Parse multiple files in one command. The CLI compiles once and runs all inputs — 5–15x faster than parsing sequentially:

# Parse all emails in a directory
docparse ~/inbox/

# Parse specific files (auto-detects batch mode)
docparse report.docx slides.pptx data.xlsx

# Glob patterns work too
docparse *.eml

What is AILANG?

AILANG is a general-purpose programming language designed for AI-powered applications. It treats AI calls as a first-class effect — just like file I/O or network access — with hard capability budgets that cap how many AI calls, file reads, or network requests a program can make. These limits are enforced at compile time, not just by convention.

AILANG also supports formal contracts verified by Z3, deterministic execution for non-AI code paths, and a package registry for sharing modules. The result: programs that are auditable, reproducible, and cost-predictable.

AILANG Parse is one application built on AILANG. Once you have the CLI installed, you can build your own AI-powered tools — document processors, data pipelines, code generators, agents — using the same language and toolchain.

AI Models

AILANG Parse uses AI for PDFs, images, audio, and video. Office formats (DOCX, PPTX, XLSX, etc.) are parsed deterministically with zero AI calls.

Google (Vertex AI / AI Studio)

# Via Application Default Credentials (Cloud Run, GKE)
GOOGLE_API_KEY="" ailang run --entry main --caps IO,FS,Env,AI \
  --ai gemini-2.5-flash docparse/main.ail document.pdf

# Via API key (local development)
GOOGLE_API_KEY="your-key" ailang run --entry main --caps IO,FS,Env,AI \
  --ai gemini-2.5-flash docparse/main.ail document.pdf

Anthropic (Claude)

ANTHROPIC_API_KEY="sk-ant-..." ailang run --entry main --caps IO,FS,Env,AI \
  --ai claude-haiku-4-5 docparse/main.ail document.pdf

Ollama (Local, Free)

# Start Ollama and pull a model
ollama pull granite3.2-vision

# Run AILANG Parse with local model (no API key needed)
ailang run --entry main --caps IO,FS,Env,AI \
  --ai granite-docling docparse/main.ail document.pdf

Audio & Video

# Transcribe audio — returns transcription, speaker count, language
./bin/docparse recording.mp3 --ai gemini-2.5-flash

# Extract video content — visual scenes, tables, transcription
./bin/docparse presentation.mp4 --ai gemini-2.5-flash

# Convert audio transcription to DOCX
./bin/docparse interview.wav --ai gemini-2.5-flash --convert transcript.docx

# Convert video content to HTML report
./bin/docparse tutorial.mp4 --ai gemini-2.5-flash --convert report.html

Audio formats: WAV, MP3, AIFF, AAC, OGG, FLAC. Video formats: MP4, MOV, AVI, WebM, WMV, MPEG, MPG, FLV, 3GPP.

SDKs (API, not local)

The Python, JavaScript, and Go SDKs connect to the hosted API — they don't run parsing locally. Use them when you want zero-maintenance scaling without installing AILANG.

# Python
pip install ailang-parse

# JavaScript / TypeScript
npm install @ailang/parse

# Go
go get github.com/sunholo-data/ailang-parse-go

Example (Python):

from ailang_parse import AilangParse

client = AilangParse()
result = client.parse("report.docx")
print(result.markdown)

SDKs include a free tier (1,000 requests/month + 50 AI parses). See the API documentation for full details, authentication, and pricing.

Configuration

Local parsing is configured via environment variables and AILANG capability budgets.

Environment Variables

Office formats (DOCX, PPTX, XLSX, ODT, ODP, ODS) need no API keys — parsing is fully deterministic and offline.

Capability Budgets

AILANG enforces hard limits on what each program can do. These are compile-time guarantees — exceed them and the runtime halts deterministically.

Supported Formats

Docker

Prefer containers? A CLI-only Dockerfile is included in the repository:

# Build the image
git clone https://github.com/sunholo-data/ailang-parse.git
cd ailang-parse
docker build -t docparse .

# Parse a file (mount your documents into /data)
docker run -v $(pwd):/data docparse /data/report.docx

# Parse with AI (pass your API key)
docker run -e GOOGLE_API_KEY="your-key" \
  -v $(pwd):/data docparse --ai gemini-2.5-flash /data/document.pdf

The image builds AILANG from source, installs all dependencies, and uses an ENTRYPOINT for direct CLI parsing. No API server included — this is for local batch processing in containers.