Conversation
…ompts - packages/agent/src/api/registry-service.ts: comment 'Babylon-compatible' → 'Feed-compatible' - packages/cloud-shared/src/lib/services/referrals.ts: comment 'Babylon's pattern' → 'Feed's pattern' - packages/core/src/features/trajectories/export.ts: comment 'Babylon-specific' → 'Feed-specific' - packages/docs/changelog.mdx: 'Babylon agent terminal' / 'Babylon prediction market game' → Feed - packages/training/scripts/rl/attacker_trainer.py: DEFENDER_SYSTEM_PROMPT 'Babylon agent' → 'Feed agent' Remaining 'babylon' refs left intentionally: - plugins/plugin-babylon/* and apps/babylon.json (per user instruction: Babylon.js 3D plugin shell) - BabylonOperatorSurface/BabylonTuiView (Babylon.js 3D UI) - elizaOS/prr@babylon branch ref in .github/workflows/run-prr.yml (real upstream branch) - BabylonSocial/scambench import URLs in packages/benchmarks/scambench/ (external dataset) - 'babylon layout' historical comment in tests/rl/conftest.py Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The snap build was failing with 'No package found with name @elizaos/scenario-runner/schema'. That's a TypeScript subpath import — not a workspace package name. Use the base name '@elizaos/scenario-runner' which actually resolves in the workspace. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| export function getAllRubricsHash(): string { | ||
| return createHash("sha256") | ||
| .update(`${Object.values(RUBRICS).sort().join("::")}${DEFAULT_RUBRIC}`) | ||
| .digest("hex") | ||
| .substring(0, 16); | ||
| } |
There was a problem hiding this comment.
getAllRubricsHash() appends DEFAULT_RUBRIC twice: once because every archetype in RUBRICS maps to DEFAULT_RUBRIC (so Object.values(RUBRICS).sort().join("::") already contains it 12 times), and again via the explicit concatenation. If per-archetype rubrics are ever added, the appended DEFAULT_RUBRIC would still inflate the hash, making it inconsistent with the intent of hashing "all unique rubric content".
| export function getAllRubricsHash(): string { | |
| return createHash("sha256") | |
| .update(`${Object.values(RUBRICS).sort().join("::")}${DEFAULT_RUBRIC}`) | |
| .digest("hex") | |
| .substring(0, 16); | |
| } | |
| export function getAllRubricsHash(): string { | |
| return createHash("sha256") | |
| .update([...new Set(Object.values(RUBRICS))].sort().join("::")) | |
| .digest("hex") | |
| .substring(0, 16); | |
| } |
| "content": turn["content"], | ||
| } | ||
| for turn in recent_history | ||
| if turn["role"] in {"system", "user", "assistant", "agent"} | ||
| if turn["role"] in {"user", "assistant", "agent"} | ||
| and str(turn["content"]).strip() != system_hint |
There was a problem hiding this comment.
"system" turns in conversation history
The filter now excludes turns with role == "system" from recent_history. Previously, system-role turns that didn't match system_hint would pass through to the bridge. If any adapter or history builder emits mid-conversation system turns (e.g., injected context updates), they will now be silently dropped without warning, potentially leaving the agent with incomplete context. Worth confirming this is intentional for all adapters feeding recent_history.
|
Claude encountered an error after 0s —— View job I'll analyze this and get back to you. |
LifeOps Multi-Tier BenchmarkSuite:
|
1 participant
Cherry-pick of snap turbo filter fix to main for v2.0.2 release builds.
Greptile Summary
This PR cherry-picks the snap turbo filter fix (removing the incorrect
/schemasubpath from--filter=@elizaos/scenario-runner) to main for the v2.0.2 release, bundled with a broad post-release follow-up touching benchmarks, training scripts, QA tooling, and generated app-core files.snapcraft.yamlcorrects the turbo workspace filter from@elizaos/scenario-runner/schema(invalid) to@elizaos/scenario-runner, ensuring the package is included in snap builds.adapters.pyextracts a retrying_docker_info_available()helper (3 attempts, 20 s timeout each), expands vision-language bundle manifest detection to three schema locations, andrunner.pyaddsfailed_scenarios/interrupted_runquarantine reasons plusoverall_scoreto the score-extraction lookup chain.run.mjsfixes the repo-root path (2→3 levels up) and renamesapp-lifeops/app-trainingreferences toplugin-lifeops/plugin-training;check-mobile-artifacts.test.tsupdates Android paths frompackages/apptopackages/app-core/platforms/android.Confidence Score: 3/5
The snap fix is minimal and correct, but the breadth of post-release follow-up across 65 files warrants careful review before merging to main.
The e2e test file calls discoverDocsRoutes() at module load time with no guard around readdirSync — if the content/ directory is absent in any CI job, every test in the file fails as a collection error rather than being skipped. The remaining changes are well-scoped with accompanying tests.
packages/cloud-frontend/tests/e2e/cloud-routes.spec.ts needs a try/catch or existence check around walk(CONTENT_DIR) before the call at module scope.
Important Files Changed
/schemasubpath from turbo filter so@elizaos/scenario-runnerworkspace is correctly included in snap builds.discoverDocsRoutes()called at module scope without error handling — ENOENT on missing content dir fails all tests._docker_info_available()with 3-attempt retry logic; expands vision-language bundle detection to acceptkernelsandfilesschema fields in addition toruntime.failed_scenariosandinterrupted_runquarantine reasons; addsoverall_scorekey to score extraction lookups in three places.getAllRubricsHashdouble-countsDEFAULT_RUBRICin its hash input.app-lifeops/app-trainingplugin paths toplugin-lifeops/plugin-training.w_active[m].slavedirectly into signal assignments, removing theautomaticlocal variable; semantically equivalent."system"from the role allowlist when building bridge message history; system-role turns inrecent_historyare now silently excluded.Flowchart
%%{init: {'theme': 'neutral'}}%% flowchart TD A[_has_terminal_bench_docker_backend] --> B[_docker_info_available] C[_has_hermes_sandbox_backend] -->|no MODAL tokens| B B --> D{docker in PATH?} D -->|no| E[return False] D -->|yes| F[attempt loop\nmax 3 tries] F --> G[docker info --format ServerVersion\ntimeout=20s] G -->|returncode == 0| H[return True] G -->|error / timeout| I{more attempts?} I -->|yes| J[sleep 0.25s] J --> F I -->|no| K[return False]Comments Outside Diff (1)
packages/cloud-frontend/tests/e2e/cloud-routes.spec.ts, line 76-80 (link)discoverDocsRoutes()called at module scope without error handlingwalk(CONTENT_DIR)callsreaddirSyncwith no try-catch. IfCONTENT_DIR(../../contentrelative to the test file) doesn't exist in the CI environment — e.g., docs aren't checked out in an agent-focused workflow —readdirSyncthrowsENOENTat module initialization, preventing the entire test file from loading and failing every test in it as a collection error rather than a skipped test.Reviews (1): Last reviewed commit: "fix(snap): use base package name instead..." | Re-trigger Greptile