Pull Request Explorer

Summary

Health Assessment

Summary

Show filtered unread count badges

Adds a new feature that displays filtered unread count badges in the dashboard, giving users clearer visibility of conversation status and improving engagement tracking.

Health Assessment

• • The PR had a long cycle time (202h) largely due to automated AI review comments, but no human review delays; the change is large but non‑breaking, resulting in moderate risk.

Summary

Add filtered unread count snapshot builder

Enables filtered unread counts for mentions, participating, unattended, and saved‑folder counts behind a feature flag, improving accuracy and performance for filtered views.

Health Assessment

• • The PR received extensive AI‑driven review with multiple iterations but no human review before merge, indicating heavy reliance on automated checks.
• • Cycle time of 211 hours and 5 review rounds suggest moderate complexity and potential blockers.

Summary

Add filtered unread count invalidation

This feature ensures that unread message counts for filtered conversations are accurate by invalidating stale snapshots. It improves notification reliability and user engagement.

Health Assessment

• • The PR had a long cycle time of over 8 days but the first review was almost immediate. Multiple automated review comments from chatgpt-codex-connector led to a single large commit, indicating significant complexity and potential risk.

Summary

Extend account feature flag storage

Adds a second feature flag column to support more flags, enabling future scalability without breaking existing APIs.

Health Assessment

• • PR completed in under a day with a single review, indicating efficient collaboration.

Summary

Re-enable Instagram inbox creation

Restores Instagram channel in inbox creation and onboarding, allowing users to add Instagram contacts while keeping other channels disabled.

Health Assessment

• • Fast cycle time and single review indicate a smooth, low-risk change.

Summary

Add filtered unread count Redis store

Introduces Redis caching for filtered unread counts, boosting sidebar notification performance and reducing database load.

Health Assessment

• • PR merged after 8.5 days with a single review cycle and a large code addition, indicating significant effort and potential risk. The automated AI review helped streamline the process.

Summary

Enforce inbox limits at model level

Adds model‑level validation to prevent inbox limit abuse, removes duplicate controller checks, and preserves the 402 API response while allowing updates when limits are reached.

Health Assessment

• • Fast initial review (0.3h) but multiple reverts and reapplications indicate iterative refinement; moderate risk due to medium scope and several review rounds.

Summary

Fix sanitization of control characters in team names

Ensures team names are cleaned of control characters, allowing deletion via UI and preventing malformed names. This improves data integrity and user experience.

Health Assessment

• • The PR took over 11 days to merge, largely due to integration merges, but the change itself was small and straightforward.

Summary

Fix Captain Hours Saved Metric

Reworks the Captain assistant hours saved metric to produce realistic values, replacing inflated calculations with a fixed per-reply effort estimate and rendering the metric in days after 100 hours.

Health Assessment

• • Quick fix with minimal changes, fast review and merge, low risk.

Summary

Add Captain Assistant Overview Page

Enables users to view key performance metrics and receive contextual banners, improving assistant adoption and coverage.

Health Assessment

• • The PR required extensive changes across frontend and backend, with a long cycle time and multiple automated review comments, indicating significant effort and potential risk.

Summary

Fix HMAC identity hash comparison

Securely compare HMAC identity hashes to prevent timing attacks, improving authentication security.

Health Assessment

• • Quick fix with minimal changes, resolved in under 5 hours, indicating efficient review and low complexity.

Summary

Validate identifier hash on widget contact update

Ensures consistent identity validation for widget contact updates, preventing unauthorized changes and improving security for user data.

Health Assessment

• • The PR received a rapid initial review and AI‑powered review suggestions, but required several merge cycles with the develop branch, leading to a longer overall cycle time. Minimal rework and a small code footprint keep risk low.

Summary

Default new accounts to Captain V2

Enables new paid accounts to automatically use the advanced Captain V2 AI feature, improving user experience and feature rollout speed.

Health Assessment

• • The PR had a long cycle time of 122 hours and involved many changes, but the first review was rapid thanks to automated AI reviews. The large scope and extended timeline suggest moderate to high risk for production impact.

Summary

Sanitize query canceled API errors

Message creation API failures now return a generic retryable error instead of exposing raw database internals, protecting customer-facing error messages.

Health Assessment

• • Fast review and single commit indicate low complexity and minimal risk.

Summary

Optimize Agent Bot Assignment Dropdown

Adds Agent Bots to the conversation assignee dropdown, enabling automated assignment and improving workflow efficiency for support teams.

Health Assessment

• • Long cycle time (264h) and multiple review rounds indicate significant complexity and stakeholder involvement. The PR touches many files across backend and frontend, increasing integration risk.

Summary

Disable Instagram and WhatsApp embedded signup

Temporarily disables inbox creation for Instagram and WhatsApp embedded signup, reducing unnecessary onboarding steps.

Health Assessment

• • Quick approval with minimal changes indicates low risk and straightforward disabling of embedded signup channels.

Summary

Fix OpenAI key usage in article search terms

Ensures article search term generation uses the correct OpenAI key from InstallationConfig, preventing authentication errors and improving reliability.

Health Assessment

• • Fast review cycle with AI assistance; minimal rework and low risk.

Summary

Fix report drilldown drawer RTL/LTR positioning

Corrects the drawer opening side for right-to-left and left-to-right languages, improving UI consistency for users.

Health Assessment

• • Quick fix with minimal changes and fast review, indicating low complexity and high confidence in the patch.

Summary

Preserve link URLs on non-hyperlink channels

Ensures links remain clickable in channels that lack Markdown support, improving user experience and reducing broken links.

Health Assessment

• • Rapid AI-assisted review cycle enabled quick resolution; moderate rework but minimal scope and fast merge.

Summary

Add catch-all failure handling to article writer

Ensures onboarding jobs complete reliably, preventing UI stalls and improving user experience.

Health Assessment

• • Fast review and single rework commit, but long overall cycle likely due to merge timing rather than code complexity.

Summary

Assign outbound call conversation to calling agent

Ensures outbound calls remain assigned to the agent who placed them, preventing misassignment by auto‑assignment and improving call tracking accuracy.

Health Assessment

• • The PR received rapid AI‑driven review comments, followed by several commits to address feedback, resulting in a moderate cycle time and moderate risk.

Summary

Add drag‑to‑reorder help center articles

Users can now drag articles across pages, streamlining content organization. This reduces manual page switching and speeds up editorial workflows.

Health Assessment

• • The PR was reviewed quickly but required several commits to address review feedback, indicating moderate friction. Despite the large code change, the cycle time remained under 24 hours.

Summary

Fix SLA handling for blocked contacts

Ensures blocked contacts are excluded from SLA processing, preventing false breaches and improving reporting accuracy.

Health Assessment

• • The PR required extensive review and multiple iterations, indicating complex SLA logic changes and a high risk of regression. The use of AI review tools helped surface issues early but the overall cycle time remained long.

Summary

Fix WhatsApp call termination race condition

Ensures calls that terminate before connect are correctly marked missed, preventing stale ringing states and improving agent experience.

Health Assessment

• • PR took 8 days to merge with a single review cycle and moderate changes; the use of an automated review bot helped catch issues early.

Summary

Fix SLA breach calculation with business hours

Ensures SLA deadlines respect business hours, improving accuracy of response time metrics for users and agents.

Health Assessment

• • Long cycle time (~135h) with many merges and commits indicates complex integration and high risk; large scope and multiple changes suggest significant review friction.

Summary

Default temperature to 0.5 and remove UI control

Changes the default temperature setting and removes the UI control for it, preserving existing account settings.

Health Assessment

• • The PR was reviewed and approved within 24 hours, indicating a smooth review process.
• • The changes are relatively small and focused, suggesting low risk.

Summary

Update msgpack gem to address CVE-2026-54522

This PR updates the msgpack dependency to patch a security vulnerability, ensuring the application remains secure without breaking functionality.

Health Assessment

• • Quick, single-commit security fix with minimal review, indicating low risk and high confidence in the change.

Summary

Broaden help center search scope

Expands Firecrawl search terms to include more help site paths, reducing onboarding skips for new accounts.

Health Assessment

• • Minimal changes and rapid review led to a quick merge with low risk.

Summary

Show clear error for failed message translation

This fix surfaces translation errors instead of generic 500, improving agent troubleshooting.

Health Assessment

• • Quick fix with minimal changes and fast review, indicating low complexity and high confidence.

Summary

Tighten Captain v2 Prompt and Config

Improves prompt consistency and behavior for Captain v2, aligning it with v1, adding datetime awareness, reducing max turns, and ensuring language consistency for users.

Health Assessment

• • PR was merged within 23 hours, with a quick first review and only a few commits after review, indicating efficient resolution of prompt logic and test adjustments.

Summary

Fix auto‑association when Companies feature disabled

Prevents unintended company creation and overwriting of contact names in accounts lacking the Companies feature, improving data integrity.

Health Assessment

• • Quick review and minimal changes indicate a straightforward bug fix with low risk.

Summary

Recover stale LeadSquared lead IDs on sync

Fixes failures when a contact's cached LeadSquared ID is stale by clearing it, re-resolving or recreating the lead, and retrying the sync, improving reliability.

Health Assessment

• • PR had a long cycle time due to merges but minimal code changes and no review, indicating low complexity and low risk.

Summary

Add trace level attributes to generations

Enables detailed tracing for AI generation events, improving observability and debugging.

Health Assessment

• • Fast cycle, minimal rework, quick approval

Summary

Update crass to 1.0.7

Fixes a security vulnerability by updating the crass dependency to 1.0.7, removing denial‑of‑service advisories.

Health Assessment

• • Security patch applied quickly with minimal code changes and fast review, indicating an efficient process.