v0.1.4
- Version bump to stay in sync with the monorepo release; no functional changes.
v0.1.3
- Version bump to stay in sync with the monorepo release; no functional changes.
v0.1.2
- Version bump to stay in sync with the monorepo release; no functional changes.
v0.1.1
ex_docbumped to~> 0.40.2; no functional changes.
v0.1.0
First release.
- Provider-agnostic streaming and completion API over
req_llm - Lazy event-tuple stream (
{:text_delta, _},{:thinking_delta, _},{:tool_call_complete, _},{:done, _},{:error, _}) - Tool calling with streaming argument assembly
- Model catalog for Anthropic, OpenAI, and Google via LLMDB
- Local server support for Ollama and llama.cpp with runtime model discovery
- JSON config loader (
Planck.AI.Config) - Multimodal input: text, image, image_url, file, video_url
parse_provider/1(private) —String.to_existing_atom→String.to_atom;@valid_providersderived fromPlanck.AI.list_providers()at compile time (single source of truth). Private function specs added.