chore: remove unused GitHub Actions workflows#104
Merged
Conversation
…ompts
Two measured waste eliminations:
1. inject_brain_rules: mandatory_reminder block restated every mandatory
rule verbatim in a recency tag. mandatory_block already places them in
primacy position — the reminder was defensive redundancy costing
~50-200 tok/session per mandatory rule.
2. context_inject: MIN_MESSAGE_LEN raised 10 → 60 to skip brain FTS
search on trivial follow-ups ("ok", "yes", "continue"). Previous
threshold only filtered 1-9 char messages, causing ~400 tok of
wasted context per ack-style reply.
Council audit (adversarial, synthesized from Skeptic + Architect +
Pragmatist) found my earlier measurement methodology used synthetic
inputs that masked real per-operation costs. Real heavy session
(10 prompts, 30 edits, 5 agents) accumulates ~9.6K gradata tokens;
context_inject alone can contribute 3.5K when prompts are substantive.
Tests: 2884 passed, 1 preexisting cp1252 Windows unicode failure
in test_rule_to_hook (unrelated to injection).
Co-Authored-By: Gradata <noreply@gradata.ai>
…imits Rule_enforcement was re-injecting SessionStart rules on every Write/Edit, costing ~165 tok/edit for zero added primacy value. Default it off behind GRADATA_RULE_ENFORCEMENT=1 opt-in for ablation. Also add env knobs for tuning injection volume without redeploying: - context_inject: GRADATA_CONTEXT_INJECT kill switch + MIN_MESSAGE_LEN raised 60 -> 100 so ack-style replies skip FTS cost - inject_brain_rules: GRADATA_MAX_RULES, GRADATA_MIN_CONFIDENCE, GRADATA_MAX_META_RULES for tuning SessionStart payload Tests updated to opt into rule_enforcement where they exercise the hook and to use a >=100 char message for context_inject. Co-Authored-By: Gradata <noreply@gradata.ai>
Telemetry showed 10 SessionStart events in a long session re-injected 1.9KB each (~3.7k tokens total) — duplicative because the compact summary already carries rules from the prior session, and the new session's primacy slot is consumed by that summary. Skip inject_brain_rules when source is 'compact' or 'resume'. Opt back in with GRADATA_INJECT_ON_COMPACT=1 for ablation. Co-Authored-By: Gradata <noreply@gradata.ai>
…d metas
When a meta-rule is injected, its source_lesson_ids are added to a mutex
set; individual RULE/PATTERN leaves whose _lesson_id() is in that set are
skipped so the principle isn't double-injected as both the meta and its
concrete leaves.
Freed slots get refilled: rank_rules is now called with a 3x overshoot
so the render loop has candidates beyond the MAX_RULES cap; the cap is
enforced at render time (cluster_lines + individual_lines <= MAX_RULES)
rather than at rank time.
On current brain: 1822B -> 1319B per SessionStart (27% smaller), and
leaf categories diversify from {LEADS, DEMO_PREP, DRAFTING} to
{DRAFTING, TONE, DATA_INTEGRITY, ARCHITECTURE, CODE} while the LEADS +
DEMO_PREP principles still land via metas.
Opt-out: GRADATA_META_RULE_MUTEX=0 for ablation. Mirrors the existing
cluster-vs-meta mutex pattern (meta_covered_categories).
Co-Authored-By: Gradata <noreply@gradata.ai>
These workflows produced noisy red CI on every PR without reflecting the actual test/deploy process. Removing to stop burning Actions minutes and false-negative signal on PR merges. Removed: - ci.yml, sdk-ci.yml, sdk-test.yml (redundant pytest runners) - docker-publish.yml, sdk-publish.yml, npm-publish.yml, sdk-release.yml (unused publish paths) - docs-deploy.yml (unused docs site deployment) Co-Authored-By: Gradata <noreply@gradata.ai>
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
There was a problem hiding this comment.
Gradata has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (8)
📝 Walkthrough• Removed 8 unused GitHub Actions workflows from • Eliminates noisy, failing CI runs that don't reflect the repository's actual test or deployment processes • Stops consuming GitHub Actions minutes on redundant test runners (multiple pytest workflows) and unused publishing paths (Docker, npm, PyPI publish workflows) • Removes automated documentation deployment to Cloudflare Pages that was not being used • No impact on local development or actual release processes—testing occurs locally and deployments happen outside GitHub Actions • Pure CI/CD configuration changes with no code-level modifications, breaking changes, or new public APIs WalkthroughAll GitHub Actions workflow files are removed, eliminating CI testing, code quality checks, documentation deployment, and automated publishing workflows for SDK, npm packages, and Docker images. Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~8 minutes Possibly related PRs
Suggested labels
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Gradata
added a commit
that referenced
this pull request
Apr 20, 2026
Restores a minimal version of the sdk-publish workflow removed in PR #104, adapted for the current release convention: - Trigger on `v*` tag push (was `sdk-v*` in the original) - Adds `workflow_dispatch` input for manual re-run - Preserves the OIDC trusted-publisher flow (no API token needed) - Preserves version-match + already-published checks - Drops the test-gate step (tests run in a separate workflow; adding ~5 min to release was the main cost motivating #104's removal) Unblocks v0.6.1 PyPI publish for mobile-only release workflow — no UV_PUBLISH_TOKEN required. Co-Authored-By: Gradata <noreply@gradata.ai>
3 tasks
Gradata
added a commit
that referenced
this pull request
Apr 20, 2026
Restores a minimal version of the sdk-publish workflow removed in PR #104, adapted for the current release convention: - Trigger on `v*` tag push (was `sdk-v*` in the original) - Adds `workflow_dispatch` input for manual re-run - Preserves the OIDC trusted-publisher flow (no API token needed) - Preserves version-match + already-published checks - Drops the test-gate step (tests run in a separate workflow; adding ~5 min to release was the main cost motivating #104's removal) Unblocks v0.6.1 PyPI publish for mobile-only release workflow — no UV_PUBLISH_TOKEN required. Co-authored-by: Gradata <noreply@gradata.ai>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
.github/workflows/— they produced noisy red CI on every PR without reflecting the actual test/deploy processRemoved
ci.yml,sdk-ci.yml,sdk-test.yml— redundant pytest runnersdocker-publish.yml,sdk-publish.yml,npm-publish.yml,sdk-release.yml— unused publish pathsdocs-deploy.yml— unused docs deploymentTest plan
gh prflow unaffected — merges no longer gated by broken CIGenerated with Gradata