alexdesousa / planck / 4664cc62f7143d47d23676b41bb9f47c459781d5
84%

DEFAULT BRANCH: main
Ran
Jobs 2
Files 31
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: 90.608% (+1.2%) from 89.401%
4664cc62f7143d47d23676b41bb9f47c459781d5

push

github

alexdesousa 
:sparkles: Session lifecycle, dynamic worker reconstruction, JsonBinding config

planck_headless — session lifecycle:

- start_session/1: resolves team (alias via ResourceStore, path, or nil for
  default dynamic team), generates <adjective>-<noun> session name, materialises
  agents with BuiltinTools + external tools, saves metadata to SQLite.
  base_url pulled from available_models for correct local server routing.
- resume_session/1: accepts id or name; reconstructs base team from metadata;
  replays completed spawn_agent calls from the previous orchestrator's history
  to restore dynamic workers — deduped by {type, name} so multiple builders
  with different names (e.g. "Bob", "Charlie") are both reconstructed; detects
  in-flight ask_agent calls (unresolved tool_result) and unfinished workers
  (last assistant turn lacks send_response); injects recovery context under
  the new orchestrator if needed.
- close_session/1, prompt/2, list_sessions/0. No SessionTracker — team_id
  read from SQLite metadata; active status from Session.whereis/1.
- start_session(template: alias) exercised via ResourceStore in tests.

planck_headless — config:

- JsonBinding (Skogsra.Binding) reads ~/.planck/config.json and
  .planck/config.json at resolution time; cached in persistent_term.
- config_files app_env (PLANCK_CONFIG_FILES); skip_json_config for tests.
- models app_env replaces local_servers: Planck.AI.Config-format declarations;
  zero network at boot. base_url flows through AgentSpec for local providers.
- Provider atoms pre-loaded before Skogsra preloads to prevent
  String.to_existing_atom failures.
- Fixed in-flight detection and spawn_agent matching to use MapSet.member?/2
  instead of is_map_key/2 guard (MapSet is not a plain map).

planck_ai:

- parse_provider/1: String.to_existing_atom -> String.to_atom; @valid_providers
  derived from list_providers() at compile time. Private specs added.

820 of 905 relevant lines covered (90.61%)

20.03 hits per line

Coverage Regressions

Lines Coverage ∆ File
2
93.55
 0.0% lib/planck/ai/config.ex
Jobs
ID Job ID Ran Files Coverage
1 4664cc62f7143d47d23676b41bb9f47c459781d5.1 17
89.4
 GitHub Action Run
2 4664cc62f7143d47d23676b41bb9f47c459781d5.2 14
90.99
 GitHub Action Run
Source Files on build 4664cc62f7143d47d23676b41bb9f47c459781d5