Packages

codex_sdk

0.2.2
0.17.0 0.16.1 0.16.0 0.14.0 0.13.0 0.12.0 0.11.0 0.10.1 0.10.0 0.9.0 0.8.0 0.7.2 0.7.1 0.7.0 0.6.0 0.5.0 0.4.5 0.4.4 0.4.3 0.4.2 0.4.1 0.4.0 0.3.0 0.2.5 0.2.4 0.2.3 0.2.2 0.2.1 0.2.0 0.1.0

Idiomatic Elixir SDK for OpenAI's Codex agent. Provides a complete, production-ready interface with streaming support, comprehensive event handling, and robust testing utilities.

HexDocs

Current section

Files

Jump to
Readme
30 Versions
6 Dependencies
Dependants
Files
Activity
codex_sdk docs 20251202 gap_analysis 003-image-input-support.md
Raw

docs/20251202/gap_analysis/003-image-input-support.md 

# ADR 003: Add image input support to thread turns



- Status: Proposed

- Date: 2025-12-02



## Context

- Upstream SDKs accept mixed text/image inputs and forward local images via `--image` (codex/sdk/typescript/src/thread.ts:27-153, codex/sdk/typescript/src/exec.ts:99-103).

- The Elixir API only accepts plain string prompts (`@spec run(t(), String.t(), ...)` in lib/codex/thread.ex:66-69) and the exec arg builder has no image flag handling (lib/codex/exec.ex:187-255).



## Problem

- Elixir users cannot send local images with their prompts, so image-grounded flows available in Python/TypeScript are unavailable here. This also blocks parity testing for multimodal turns.



## Decision

- Extend the public thread API to accept a structured input list (text + local image entries) and normalize it into prompt + image paths.

- Update `Codex.Exec.build_args/1` to emit `--image` arguments for each path while retaining existing attachment handling.



## Actions

- Add unit tests covering mixed inputs and ensure streamed/buffered paths both forward images.

- Document the new input shape and add an example mirroring the upstream SDK.