Open-source project lets DeepSeek text-only models use Codex's vision tool
teortaxesTex · x · 2026-08-02
Developer Anionex released open-source project codex-deepseek-vision, enabling text-only models like DeepSeek to use Codex's built-in viewimage tool without extra MCP, skills, or CLI. It includes proxy scripts and a vision toolkit supporting image Q&A, screenshot analysis, OCR, and visual grounding, verified in real Codex + DeepSeek sessions.
More from coding & agent
- To Maximize AI Agents, Developers Must Let Go of the Code — ericelliott_ · 2026-08-03
- AI Agents Cut Threat Investigation Time from Days to Seconds in SOC — brucemacv · 2026-08-03
- Turn Customer Feedback into Roadmaps Automatically with Codex — gdb · 2026-08-03
- Rewriting Bioinformatics Tool pydREG with Claude & Codex — anshulkundaje · 2026-08-03
- Opus 5 Generates AAA-Quality Game Scene via Agent Loops — mattshumer_ · 2026-08-03
- Building a Ghibli-Style Interactive SF Map with Coding Agents — keerthanpg · 2026-08-03