Change Log

Copy Markdown View Source

All notable changes to this project will be documented in this file.

v0.0.1 - Initial Release

  • Initial Text Only Generation.

v0.0.2 - Local Model Support

  • Local LLama support added for pulling in gguf models for inference.

v0.0.3 - Vision Support and Internal Structure Update

Warning - This update may break existing code.

  • Added Vision support for OpenAI, Gemini, and Anthropic.
  • Updated internal structure to allow for more advanced use cases such as prompt loops/fitness checks, added support for how to handel local models, etc.

v0.1.0 - Split into core and extension libraries

To use local llama you must replace 

config :genai_local, :local_llama,
       enabled: true,
       otp_app: :my_app

with 

config :genai_local, :local_llama,
       otp_app: :my_app

and add {:genai_local, "~> 0.1"} to your dep list. ## v0.2.0 Update to use revamped core libs. ## v0.2.3 XAI, and DeepSeek support added. ## v0.3.9 OpenRouter chat provider (GenAI.Provider.OpenRouter): OpenAI-compatiblehttps://openrouter.ai/api/v1`, Bearer OPENROUTER_API_KEY, live GET /models, optional HTTP-Referer / X-OpenRouter-Title attribution headers. ## v0.3.8 Qwen/DashScope media providers: GenAI.Provider.Qwen.Image (qwen-image-3.0), Qwen.Speech (qwen3-tts-flash), Qwen.Video (wan2.7-t2v async poll). Keys: DASHSCOPE_API_KEY / QWEN_API_KEY / QWEN_TOKEN_KEY. Token-plan host via settings[:plan] => :token_plan. ## v0.3.7 Qwen token-plan mode: token_plan: true (or mode: :token_plan) switches the OpenAI-compatible host to https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1 and Bearer auth to token_api_key / QWEN_TOKEN_KEY. On-demand QWEN_API_KEY + dashscope-intl remains the default. ## v0.3.6 Qwen / Alibaba Cloud Model Studio (DashScope) chat provider added. OpenAI-compatible endpoint defaults to https://dashscope-intl.aliyuncs.com/compatible-mode/v1, Bearer auth from config :genai, :qwen, api_key: (QWEN_API_KEY, with DASHSCOPE_API_KEY fallback). Catalog helpers include qwen3.8-max; models/0 lists the live compatible-mode catalog. Thinking (reasoning_content, enable_thinking, reasoning_effort) and vision image parts are supported. ## v0.3.5 ElevenLabs media provider added (ADR-016): sync text->speech (/v1/text-to-speech/{voice_id}), text->sfx (/v1/sound-generation), and text->music (/v1/music/compose), xi-api-key auth, ELEVENLABS_API_KEY env fallback; registered in the default media_providers list. README refreshed (provider list, media generation shipped, ElevenLabs usage example).