alexdesousa / planck / 20c8e1270650dce119e3286c01720e1e57ba049a
84%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 17
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 89.815%. First build
20c8e1270650dce119e3286c01720e1e57ba049a

push

github

alexdesousa 
:sparkles: Add planck_ai — typed LLM provider abstraction

Introduces the first package in the monorepo: `planck_ai`, a typed,
provider-agnostic LLM abstraction built on top of `req_llm`.

Core features:
- Lazy streaming event pipeline (`{:text_delta}`, `{:thinking_delta}`,
  `{:tool_call_complete}`, `{:done}`, `{:error}`) via `Stream.resource`
- Stateful tool call assembly: req_llm streams JSON argument fragments
  across multiple chunks; `Planck.AI.Stream` buffers them by index and
  emits a single assembled `{:tool_call_complete, ...}` per call
- Model catalog for Anthropic, OpenAI, and Google via bundled LLMDB
  snapshot (no network call at startup)
- Runtime model discovery for Ollama (`/api/tags`) and llama.cpp
  (`/models`), plus `model/2` factory for direct struct construction
- `Planck.AI.Config` JSON loader for configuring local servers without
  hardcoding model structs
- Multimodal input: text, image (binary + URL), file, video URL
- `Tool.new/1` for building tool call definitions with JSON Schema params
- `Planck.AI.Adapter` translates `Planck.AI.Context` → `ReqLLM.Context`,
  with per-provider model spec format (string vs map to bypass LLMDB)
- Per-model `default_opts` with call-site override
- 94 tests, full dialyzer and credo compliance
- CI workflow for GitHub Actions

Also updates `specs/planck-ai.md` to reflect the actual implementation.

194 of 216 relevant lines covered (89.81%)

35.78 hits per line

Jobs
ID Job ID Ran Files Coverage
1 20c8e1270650dce119e3286c01720e1e57ba049a.1 17
89.81
 GitHub Action Run
Source Files on build 20c8e1270650dce119e3286c01720e1e57ba049a