v0.4.13: bugfixes for dev containers, posthog redirect, helm updates

[icecrasher321]

Summary

fix(devcontainer): fixed devcontainers (#1614)

fix(posthog): prevent redirects by whitelisting ingest route (#1615)

fix(kb): added tiktoken for embedding token estimation (#1616)

feat(helm): added pdb to helm (#1617)

fix(schedules): migrate to trigger dev (#1618)

fix(sdk): fixed TS SDK publishing job to use later Node version (#1619)

feat(helm): added cert-postgresql template to helm (#1620)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
docs	Ready	Preview	Comment	Oct 13, 2025 9:16pm

[greptile-apps]

Greptile Overview

Summary

This PR consolidates 7 bug fixes and feature additions across infrastructure, performance, and developer tooling:

Major Changes:

• Schedule execution migration to trigger.dev: Refactored schedule execution from monolithic API route to task-based architecture with executeScheduleJob function, adding proper authentication via verifyCronAuth and supporting both trigger.dev and direct execution modes
• Token-aware embedding batching: Replaced fixed 50-item batching with accurate token counting using tiktoken library, batching by 8000 tokens for more efficient API usage and preventing request failures
• PostHog optimization: Disabled heavy features (pageleave, performance tracking, heatmaps, dead clicks) and narrowed autocapture scope; prevented redirect loops by excluding /ingest path from domain redirects
• DevContainer improvements: Complete rewrite with proper docker-compose setup, non-root user configuration, automated dependency installation and database migrations
• Helm enhancements: Added PodDisruptionBudget for high availability and PostgreSQL TLS certificate template for secure connections
• SDK publishing fix: Upgraded Node.js to v22 in CI workflow to support File type

Key Improvements:

• Better separation of concerns with schedule execution logic extracted to dedicated background task file
• More accurate token counting prevents API errors and optimizes costs
• Reduced PostHog overhead and eliminated redirect issues
• Production-ready Kubernetes configurations for availability and security

Confidence Score: 4/5

• This PR is safe to merge with minimal risk - changes are well-isolated, focused bug fixes with one minor concern around error handling.
• The PR demonstrates solid engineering practices with good separation of concerns, comprehensive error handling, and appropriate testing. The schedule execution refactoring properly extracts complex logic, the tiktoken integration uses established libraries correctly, and infrastructure changes follow Kubernetes best practices. Score is 4/5 rather than 5/5 due to one observation: the direct execution path uses fire-and-forget pattern (void executeScheduleJob) which could mask failures, though this appears intentional for the fallback mode.
• No files require special attention - all changes are straightforward and well-implemented

Important Files Changed

File Analysis

Filename	Score	Overview
apps/sim/app/api/schedules/execute/route.ts	4/5	Refactored to delegate schedule execution to trigger.dev tasks or direct execution via executeScheduleJob. Added verifyCronAuth check for security.
apps/sim/background/schedule-execution.ts	4/5	New file containing extracted schedule execution logic with comprehensive error handling, rate limiting, usage checks, and trigger.dev integration.
apps/sim/lib/tokenization/estimators.ts	5/5	Added tiktoken integration for accurate token counting with OpenAI models, including batching utilities and truncation functions.
apps/sim/lib/embeddings/utils.ts	5/5	Migrated from fixed-size batching (50 items) to token-aware batching (8000 tokens) using tiktoken for more efficient API usage.
apps/sim/lib/posthog/provider.tsx	5/5	Reduced PostHog feature set (disabled pageleave, performance tracking, heatmaps, dead clicks) and narrowed autocapture scope to improve performance.
apps/sim/next.config.ts	5/5	Added /ingest path to redirect exclusions to prevent PostHog analytics calls from being redirected to sim.ai domain.
.devcontainer/post-create.sh	5/5	Comprehensive setup script that handles dependency installation, database migrations, and shell configuration for development environment.
.github/workflows/publish-ts-sdk.yml	5/5	Updated Node.js version from earlier version to 22 to support File type used in SDK.
helm/sim/templates/poddisruptionbudget.yaml	5/5	New PodDisruptionBudget template for both app and realtime services to ensure high availability during cluster operations.
helm/sim/templates/certificate-postgresql.yaml	5/5	New certificate template for PostgreSQL TLS with cert-manager integration for secure database connections.

Sequence Diagram

sequenceDiagram
    participant Cron as Cron Job
    participant API as /api/schedules/execute
    participant TriggerDev as Trigger.dev
    participant Executor as Schedule Execution
    participant DB as Database
    participant RateLimit as Rate Limiter
    participant Billing as Billing Service
    participant Workflow as Workflow Executor
    
    Cron->>API: GET (scheduled trigger)
    API->>API: verifyCronAuth()
    API->>DB: Query due schedules
    DB-->>API: Return schedules list
    
    alt TRIGGER_DEV_ENABLED
        loop For each schedule
            API->>TriggerDev: tasks.trigger('schedule-execution', payload)
            TriggerDev-->>API: Task handle
        end
        API-->>Cron: 200 OK (queued)
        
        TriggerDev->>Executor: executeScheduleJob(payload)
        Executor->>DB: Load workflow & user data
        Executor->>RateLimit: Check rate limits
        Executor->>Billing: Check usage limits
        
        alt Limits OK
            Executor->>Workflow: Execute workflow
            Workflow-->>Executor: Execution result
            Executor->>DB: Update schedule (nextRunAt, lastRanAt)
            Executor->>DB: Update user stats
        else Rate limited
            Executor->>DB: Set nextRunAt (retry in 5 min)
        else Usage exceeded
            Executor->>DB: Calculate and set nextRunAt
        end
    else Direct execution
        loop For each schedule
            API->>Executor: executeScheduleJob(payload)
            Note over API,Executor: Fire and forget (void)
        end
        API-->>Cron: 200 OK (queued)
    end

Loading

_{30 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}