• v0.3.1 0c366180fa

    TFTSR v0.3.1
    Some checks failed
    Test / rust-fmt-check (push) Successful in 1m40s
    Test / frontend-typecheck (push) Successful in 2m15s
    Test / rust-tests (push) Has been cancelled
    Test / frontend-tests (push) Has been cancelled
    Test / rust-clippy (push) Has been cancelled
    Auto Tag / autotag (push) Successful in 6s
    Auto Tag / wiki-sync (push) Successful in 7s
    Auto Tag / changelog (push) Failing after 35s
    Auto Tag / build-macos-arm64 (push) Successful in 3m23s
    Auto Tag / build-linux-amd64 (push) Successful in 8m36s
    Auto Tag / build-windows-amd64 (push) Successful in 10m10s
    Auto Tag / build-linux-arm64 (push) Successful in 10m24s

    sarman released this 2026-05-23 22:39:41 +00:00 | 375 commits to beta since this release

    What's New in v0.3.1

    feat(mcp): MCP Server Support

    Adds full Model Context Protocol (MCP) server management, enabling the AI assistant to discover and call tools from external MCP servers during triage conversations.

    • Connect external tool servers via stdio (local process) or http (Streamable HTTP) transports
    • Auth types: none, api_key, bearer, oauth2 — credentials encrypted at rest (AES-256-GCM)
    • Auto-discovery of enabled servers at app startup
    • MCP tools injected into every AI triage chat session automatically
    • Mandatory audit log entry before every external tool call
    • PII scan on all tool call arguments
    • New Settings page: Settings → MCP Servers
    • 185/185 Rust tests · 94/94 Vitest tests

    fix(mcp): timeouts, audit on delete, OAuth state nonce

    • 30s hard timeout on MCP tool calls; 60s on discovery — prevents hung AI sessions
    • delete_mcp_server writes audit event (server name, tool count, resource count) before cascade delete
    • initiate_mcp_oauth includes cryptographically random state nonce for CSRF protection

    fix(ci): PR review model and pipeline fixes

    • PR review automation now uses qwen3-coder-next (always-running on DGX Spark) instead of qwen2.5-72b (on-demand)
    • Improved review prompt requires line-quoted evidence for every finding
    • auto-tag now reads Cargo.toml version — major/minor bumps in Cargo.toml are honoured directly
    • Replaced tea CLI calls with curl in changelog steps (tea not available in alpine:latest)
    Downloads