Node CLI for indexing local sources and composing task-specific context with Contextrie.

The CLI writes its working files under .contextrie/ in the current project:

• config.json: saved model/provider configuration
• sources.json: indexed source manifest
• context.md: composed context for the current task

Install globally with npm:

npm install -g @contextrie/cli

Then run:

contextrie --index --all --openai-api-key "$OPENAI_API_KEY" --openai-base-url "$OPENAI_BASE_URL" --openai-model "gpt-5.4"
contextrie --task "Summarize the files most relevant to parser source path handling."

For local development with Bun:

bun install
bun run ./index.ts --index --all --openai-api-key "$OPENAI_API_KEY" --openai-model "gpt-5.4"
bun run ./index.ts --task "Summarize the files most relevant to parser source path handling."

Use --openai-base-url as well if you are targeting an OpenAI-compatible provider.

For the standalone macOS arm64 binary, use the separate compiled artifact and Homebrew flow. That is distinct from the npm bin entrypoint.

contextrie --help

Usage
  $ contextrie --index [paths...]
  $ contextrie --task "describe the task"

Index explicit files, directories, or globs:

contextrie --index README.md core parsers

Index all supported source files in the current repository:

contextrie --index --all

Supported files are filtered through @contextrie/parsers, and indexed metadata is written to .contextrie/sources.json.

After indexing, compose a task-specific context bundle:

contextrie --task "Explain which files matter most for the CLI indexing flow."

This writes the final output to .contextrie/context.md.

The CLI resolves configuration from:

1. command flags
2. .contextrie/config.json
3. environment variables

Required:

• OPENAI_API_KEY
• OPENAI_MODEL

Optional:

• OPENAI_BASE_URL

Early development; expect breaking changes.