Skip to content

feat(stock-watchlist-agent): support capture --publish (baseline experiment + cost) - #79

Draft
mehulsonowal wants to merge 2 commits into
stock-watchlist-agent-inline-experimentsfrom
stock-watchlist-agent-capture-publish
Draft

feat(stock-watchlist-agent): support capture --publish (baseline experiment + cost)#79
mehulsonowal wants to merge 2 commits into
stock-watchlist-agent-inline-experimentsfrom
stock-watchlist-agent-capture-publish

Conversation

@mehulsonowal

Copy link
Copy Markdown
Contributor

What

Companion demo changes for the ddtrace capture --publish feature (DataDog/dd-trace-py#18984, stacked on #18625).

  • Capture drivers (capture_watchlists.py, capture_quotes.py, capture_one.py) declare SUBJECT (the experiment name) and INPUTS (the boundary's kwargs, one dataset record each), derived from the same watchlists/tickers generate_traffic() uses (env overrides honored). capture --publish reads these to run the real boundary as the baseline experiment.
  • TESTING.md — step-by-step guide for the capture-publish flow (baseline w/ cost + dataset → change → replay + compare view).

No change to local capture behavior.

Testing

See test-apps/stock-watchlist-agent/TESTING.md. Requires a ddtrace build with the capture-publish feature (dd-trace-py#18984). Draft: for testing.

🤖 Generated with Claude Code

mehulsonowal and others added 2 commits July 9, 2026 16:57
…blish`

The new `ddtrace-experiment capture --publish` runs the real boundary as the baseline
experiment (real cost) over declared inputs, so the capture drivers now expose SUBJECT
(the experiment name) and INPUTS (the boundary's kwargs, one dataset record each), derived
from the same watchlists/tickers the local `generate_traffic()` uses (env overrides honored).
No change to local `capture` behavior.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… flow

Step-by-step guide to exercise the baseline-with-cost flow end to end: branches, .env
config, capture --publish (baseline experiment + dataset + cost), replay --publish
(current + compare view), the offline loop, and the backend-coupled things to verify.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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