Skip to content

feat(energy): visible identity and inline task specs in notebook 03 - #196

Merged
ethancjackson merged 4 commits into
mainfrom
feat/nb3-visible-identity-task-specs
Aug 26, 2026
Merged

feat(energy): visible identity and inline task specs in notebook 03#196
ethancjackson merged 4 commits into
mainfrom
feat/nb3-visible-identity-task-specs

Conversation

@ethancjackson

Copy link
Copy Markdown
Collaborator

Summary

  • One shared, editable analyst identity (system instruction + search sub-agent + toolbelt inventory) at the top of 03_one_agent_three_tasks.ipynb.
  • Each of the three streams assigns its role with an inline task spec and a user-payload preview before the model run (trajectory, binary shock, scenario).
  • Factory matches the notebook: all task kinds use build_wti_multitask_news_config + WtiMultitaskPromptBuilder (payload now includes horizons and standard_quantiles).

Test plan

  • Run notebook 03 setup + identity cell: system prompt and toolbelt table render; search_web shown on.
  • Stream 1: task spec visible; payload preview has no model call; trajectory run still produces fan chart + MAE.
  • Stream 2: shock spec + calibration bullets visible; origin cards and Brier chart still work.
  • Stream 3: default origin is 2026-03-02; editing SCENARIO_AS_OF / the task spec and setting USE_CACHE = False changes the ask.
  • uv run pytest implementations/tests/energy_oil_forecasting/test_tasks.py

Made with Cursor

ethancjackson and others added 4 commits August 26, 2026 09:32
One shared multitask analyst identity with editable system and search
instructions, then three inline task specs (trajectory, shock, scenario)
with payload previews before each run. Unify the factory so all task kinds
use WtiMultitaskPromptBuilder and build_wti_multitask_news_config.

Co-authored-by: Cursor <cursoragent@cursor.com>
Capture the executed Stream 3 cells after running with a past as_of
(2026-02-02) so the identity/task-spec work is preserved independently
of the cutoff-verifier tracing follow-up.

Co-authored-by: Cursor <cursoragent@cursor.com>
Keep the pedagogical as_of on 2026-03-02; pre-shock and live dates stay as
commented alternatives. Drop the local USE_CACHE override from that cell.

Co-authored-by: Cursor <cursoragent@cursor.com>
@ethancjackson
ethancjackson merged commit 6873c0b into main Aug 26, 2026
1 of 2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant