prompt2json v0.3.0: Multi-Provider Support Arrives

Released: February 4, 2026

We’re excited to announce prompt2json v0.3.0, a major release that transforms our CLI tool from a single-provider utility into a universal LLM-to-JSON interface. This release brings OpenAI provider support, enabling you to use prompt2json with OpenAI models, Ollama local deployments, Google Cloud’s OpenAI-compatible endpoint, or any server implementing the OpenAI Chat Completions API.

What’s New

Multi-Provider Support

The headline feature of v0.3.0 is OpenAI provider support. You can now choose between two providers:

  • gemini - Google Cloud’s Vertex AI (the previous single provider)
  • openai - OpenAI Chat Completions API and compatible endpoints

This flexibility means you can switch between providers based on your budget, infrastructure preferences, or model requirements—all with the same prompt2json interface.

New Command-Line Flags

v0.3.0 introduces several new flags to support multi-provider functionality:

  • --provider (required) - Choose your backend: gemini or openai
  • --api-key - Supply bearer token authentication for the OpenAI provider
  • --url - Override default API endpoints (works with both providers)
  • --strict-schema - Enable OpenAI’s strict mode for structured outputs

Local LLM Support

With the --url flag and OpenAI provider, you can now point prompt2json at local LLM deployments:

echo "extract entities" | prompt2json \
    --provider openai \
    --url "http://localhost:11434/v1/chat/completions" \
    --model "llama3.2" \
    --schema '{"type": "object", "properties": {...}}'

This opens up privacy-sensitive use cases and eliminates cloud dependency for development and testing.

OpenAI Structured Outputs

The new --strict-schema flag leverages OpenAI’s structured outputs feature, which rejects responses that don’t conform exactly to your JSON schema. This provides stronger guarantees for automation workflows.

Breaking Changes

Important: If you’re upgrading from v0.2.0, your scripts need a quick update. The --provider flag is now required:

Before (v0.2.0):

prompt2json --system-instruction "..." --schema '...' \
    --project "my-project" --location "us-central1" \
    --model "gemini-2.5-flash"

After (v0.3.0):

prompt2json --provider gemini --system-instruction "..." --schema '...' \
    --project "my-project" --location "us-central1" \
    --model "gemini-2.5-flash"

Note that --project and --location are now Gemini-only flags. Using them with the OpenAI provider will result in an error.

Why It Matters

prompt2json solves a fundamental problem: turning free-form LLM prompts into reliable, schema-validated JSON for automation and batch processing. Before v0.3.0, you were locked into Google’s Vertex AI. Now you have choices:

  • Cost optimization: Run local models with Ollama for zero per-call costs
  • Model experimentation: Test the same prompt across Gemini, GPT-4o, or open-source alternatives
  • Infrastructure flexibility: Use cloud APIs during development and switch to local deployment in production
  • Vendor independence: No more lock-in; your prompts work anywhere

Provider Comparison

Feature Gemini OpenAI
Attachments (images/files) ✅ Yes ❌ Text-only
Requires API key ADC or --url Required (or custom --url)
Strict schema mode ❌ ✅ Via --strict-schema
Local deployment ❌ ✅ Via --url

Upgrade & Installation

Install v0.3.0

Using Go install:

go install github.com/UnitVectorY-Labs/prompt2json@v0.3.0

From source:

git clone https://github.com/UnitVectorY-Labs/prompt2json.git
cd prompt2json
git checkout v0.3.0
go build -o prompt2json

Pre-built binaries: Available for macOS (Intel/Apple Silicon), Linux (x86_64/ARM/32-bit), and Windows from the GitHub release page.

Getting Started with OpenAI Provider

  1. Set your API key: 
    export OPENAI_API_KEY="sk-..."
  2. Run your first OpenAI prompt: 
    echo "summarize this" | prompt2json \
    --provider openai \
    --model "gpt-4o" \
    --schema '{"type": "object", "properties": {"summary": {"type": "string"}}}'

Documentation

Full usage guides, examples, and installation instructions are available in the docs.

Transparency Note

This release announcement was AI-generated using the unsloth/Qwen3.5-122B-A10B-GGUF:Q4_K_M model on February 4, 2026, based on the v0.3.0 release of the prompt2json repository. The article was authored by release-storyteller.