New branch reason: No prior New Runtime record was supplied, so this release begins a distinct analytical record focused on the plugin’s command-line bridge to the Decisions API.
Thesis
A small integration plugin turns OpenAI’s Decisions API into a command-line workflow that can accept image-based decision queries.
Observed facts
Simon Willison’s release note identifies the package as llm-openai-decisions 0.1a0 and gives
1llm install llm-openai-decisionsas its installation command.The note demonstrates an image attachment queried with gpt-6-luna and shows a predicate response containing a probability.
1It also says the API supports yes/no, choice, and score question types.
1
Mechanism
The plugin bridges the LLM command-line tool to the API, exposing structured decision requests through a familiar command interface rather than requiring users to build a separate client. The example indicates that an image can be supplied as an attachment and that the answer can be represented as structured data.
Why now
The release follows the announcement of the Decisions API at OpenAI’s DevDay and provides an immediate route for experimenting with it in an existing LLM command-line workflow.
Implications
The integration lowers the friction of testing multimodal, structured decisions from scripts and command-line workflows.
This is an integration and example, not independent evidence of decision quality, robustness, or production readiness.
Evidence boundary
The supplied source is the plugin author’s release post and documents stated API capabilities, installation, and one example; it does not report independent benchmarks or systematic testing.
Watch conditions
Watch for broader usage or independent tests that establish how reliably image-based decisions and the returned probabilities perform across real tasks.