Changelog

Direct Sentry SDK envelope parsing and dynamic microservice repo scoping.

• Implemented newline-delimited envelope parsing at /api/{project_id}/envelope/ and /api/{project_id}/store/.
• Dynamic per-service repository overrides and automatic GitHub App token caching.
• Master-detail Fleet Services dashboard cockpit with Framer Motion spring physics.

Per-user SQLite database separation and two-level configuration architecture.

• Automatic personal workspace auto-provisioning with 5-minute metrics caching (workspaces.py).
• Decoupled Platform Fleet settings (severity rules, automation policy) from Microservice settings (repo scoping, Slack alerts).
• Dual-gated automation policy engine supporting full-manual, semi-autonomous, and fully-autonomous modes.

GitHub App authentication eliminating manual Personal Access Tokens.

• Added GitHubAppService with RS256 JWT signing and automatic ephemeral installation token minting.
• Dynamic repository connection and release blame correlation across fleet microservices.
• Supports PEM strings, base64 env vars, file path discovery, and Render /etc/secrets/ mounts.

FastAPI JWT verification middleware and role-based access control.

• Integrated Supabase Auth with HS256 JWT verification middleware in FastAPI (auth_middleware.py).
• Built 4-tier Role-Based Access Control (owner, admin, member, viewer) with team action audit logging (teams.py).
• Edge session protection via @supabase/ssr with unified OAuth callback routing.

Measured deterministic pipeline throughput with zero external IO.

• Evaluated 3 production fixtures with average core latency of 0.09 ms.
• Synchronous regex token scrubbing completed in under 25ms on 50KB payloads.
• Deterministic SHA-256 crash grouping ensures instant deduplication.

188 backend unit tests passing with full pipeline execution.

• Implemented models.py, config.py, parse.py, redact.py, severity.py, and templates.py.
• GraphQL blame module pinned to deployed release commit SHA with 3-commit fallback.
• Bounded LLM explanation integration with structured fallback extraction.
• Offline dry-run replay engine executing via CLI in under 1.5 seconds.

Dashboard interface modeled after modern developer platforms.

• Deploys-style incident stream with status pills and search.
• Deploy detail cockpit: suspect commit metadata, unified diff viewer, logs, and issue draft.
• One-click human sign-off workflow before GitHub publication.

Formal specification defining bounded AI boundaries and deterministic rules.

• Established core rule: Code gathers facts, bounded LLM only explains.
• Eliminated AI hallucination of commit hashes, authors, and severities.
• Specified P1/P2/P3 arithmetic rules based on affected users and event count.