v0.3.7

v0.3.7 – Test stability fixes

Tests

• fix: stabilize CI by fixing event-loop usage in tests/unit/test_rollback_apply.py and avoiding mixing @pytest.mark.asyncio with TestClient in tests/unit/test_managed_endpoint.py.
• Note: Test-only changes; no runtime behavior change.

Links

• Full Changelog: v0.3.6...v0.3.7

v0.3.6

v0.3.6 – Auto‑unlimit inactive torrents

Added

• feat(config): global.auto_unlimit_on_inactive (default false). When enabled, Qguardarr sets per‑torrent upload limits to unlimited (‑1) as soon as a torrent is not active in the current allocation cycle.

Changed

• feat(allocation): After applying differential updates, auto‑unlimit torrents that are currently inactive. Records rollback entries with reason auto_unlimit_inactive. In dry‑run, persists changes to the dry‑run store only (no qBittorrent calls).

Tests

• Add unit tests covering real mode and dry‑run behavior.

Links

• Full Changelog: v0.3.5...v0.3.6

v0.3.5

What's Changed

Added

• config: global.cache_ttl_seconds to control TorrentCache cleanup TTL (default 1800s).

Changed

• allocation: Only include active torrents; drop cache backfill-by-hashes API calls.
• allocation: Use configured TTL for cache cleanup to avoid stale entries.
• qbit: Add DEBUG logs for tracker URL selection and fallbacks.

Documentation

• README: Simplify Docker Quick Start, clarify config essentials, and document cache TTL.

Tests

• Update active-torrents unit to assert no backfill; add TTL cleanup test.

No breaking changes.

v0.3.4

Changed

• chore(config): Default global.rollout_percentage is now 100 when omitted (previous default was 10). If you don’t specify it in config, Qguardarr manages 100% of eligible torrents.
• chore(matching): Normalize shorthand tracker patterns like .example\..* (e.g., .example\.com.) into .*example\.com.* unless anchored with ^/$.

Added

• feat(logging): DEBUG logs now include the selected tracker URL per torrent and the matched tracker with current upload speed (e.g., up=2.95 MiB/s).
• feat(api): /match/test?url=...&detailed=true endpoint to test a tracker URL against configured patterns.

Tests

• Add unit tests for pattern normalization and the new match-test endpoint.

Full Changelog

• v0.3.3...v0.3.4

v0.3.3

Added

• feat(api): GET /stats/managed returns managed torrents grouped by tracker with hash, current_limit, added_at, last_seen, and age_seconds.

Fixed

• fix(stats): For unlimited trackers (configured max_upload_speed <= 0), report configured_limit_mbps: null instead of -0.

Notes

• Application logs are written to the configured file (e.g., logs/qguardarr.log), while Uvicorn writes console logs. Tail the file for allocator activity.

Full Changelog

• v0.3.2...v0.3.3

v0.3.2

Changed

• perf(qbit): Eliminate full-list queries. Use filter=active + upspeed threshold and backfill only a bounded subset of cached hashes via GET /torrents/info?hashes=... (cap 1000). Trackers are fetched only for the filtered subsets.
• chore(tests): Update unit tests to validate backfill-by-hashes behavior.
• chore(runner): Quick Docker test runner now waits for Qguardarr and executes the config hot-reload test in quick mode.

Added

• feat(config): Automatic config hot-reload via background watcher (mtime-based). Manual POST /config/reload endpoint for ops.
• test(integration): Hot-reload integration test that edits host config/qguardarr.yaml and asserts rollout change.

Security

• No changes since 0.3.1; single-password auth and log redaction remain enforced.

Full Changelog

• v0.3.1...v0.3.2

v0.3.1

Highlights

• Performance: query torrents with filter=active and fetch trackers only for torrents with real upload activity. This massively reduces qBittorrent API calls on large libraries.
• Config: example config now reads global.port from APP_PORT to align with Compose.

Images

• GHCR: ghcr.io/mrinvincible29/qguardarr:latest
• GHCR (pinned): ghcr.io/mrinvincible29/qguardarr:v0.3.1
• Edge (latest commit): ghcr.io/mrinvincible29/qguardarr:edge

Changed

• perf(qbit): Use /api/v2/torrents/info?filter=active (instead of uploading) and then keep only torrents with upspeed >= active_torrent_threshold_kb * 1024. Fetch /api/v2/torrents/trackers only for this filtered subset. Greatly reduces per-cycle API load.

Tests

• Add unit test verifying the active filter is used and tracker lookups occur only for the active subset.
• Docker quick integration tests: green.

Documentation

• Example config reads global.port from APP_PORT for consistency with Compose.

Upgrade Notes

• No breaking changes. Behavior is more selective by default and should lower API traffic.
• If you rely on very low-activity torrents, adjust global.active_torrent_threshold_kb (default example is 10 KB/s). In large environments, 10–50 KB/s is typical; use higher values to be more selective.

Quick Pull

• docker pull ghcr.io/mrinvincible29/qguardarr:v0.3.1
• To pin in Compose: image: ghcr.io/mrinvincible29/qguardarr:v0.3.1

Full Changelog

• v0.3.0...v0.3.1

v0.3.0

Highlights

• Soft per-tracker limits with borrowing + smoothing (Phase 3).
• New admin endpoints: preview next cycle, reset smoothing, reset limits.
• Security: single-password qBittorrent auth, no password logs.
• Docker: use GHCR image by default; simpler Compose with APP_PORT, no version: key.

Images

• GHCR: ghcr.io/mrinvincible29/qguardarr:latest
• GHCR (pinned): ghcr.io/mrinvincible29/qguardarr:v0.3.0
• Edge (latest commit): ghcr.io/mrinvincible29/qguardarr:edge

Added

• Phase 3 “soft” strategy with priority-weighted borrowing and EMA smoothing.
• API: GET /preview/next-cycle for proposed tracker caps and per‑torrent changes.
• API: POST /smoothing/reset to clear Phase 3 smoothing state.
• API: POST /limits/reset to set touched torrents back to unlimited (supports dry‑run).
• Stats: expose current strategy; add effective_cap_mbps, borrowed_mbps per tracker.

Security

• Authenticate to qBittorrent using only configured credentials.
• Remove fallback/default password attempts; redact secrets in logs.

Changed

• Compose: pull from GHCR by default; drop obsolete version: key; expose APP_PORT for container port/health alignment.
• Compose: remove mem_limit; rely on host/container constraints or user-provided limits.
• Logging: centralized setup; ensure log directory exists; avoid file handler permission errors.

Documentation

• Quick Start: run directly from GHCR, download compose and config from repo (no git clone).
• Webhooks: optional but recommended; periodic cycles still manage updates without them. Benefits and examples documented.
• .env.example: add APP_PORT=8089.

Tests

• Unit test: enforce single-password auth and log redaction.
• Docker quick integration tests: passing with Phase 3 and new endpoints.

Upgrade Notes

• Config compatibility: 0.3.0 is backward-compatible; Phase 3 is opt-in via global.allocation_strategy: soft.
• Docker:
  • Compose now expects APP_PORT (defaults 8089). If your config.global.port differs, set APP_PORT in .env to match (e.g., APP_PORT=8189).
  • version: key removed to silence Compose v2 warnings.
  • mem_limit removed; if you previously saw OOM kills (exit 137), consider adding your own limits or ensure adequate host memory.
• Cross-seed:
  • Set cross_seed.enabled: true in config to enable forwarding; environment variables alone do not enable it.

Quick Pull

• docker pull ghcr.io/mrinvincible29/qguardarr:v0.3.0
• In compose, set image: ghcr.io/mrinvincible29/qguardarr:v0.3.0 to pin.

Full Changelog

• v0.2.0...v0.3.0

Qguardarr 0.2.0 — Phase 2: Weighted Allocation + Scoring

Qguardarr 0.2.0 — Phase 2: Weighted Allocation + Scoring

Highlights

• Weighted within-tracker allocation with smart torrent scoring (optional).
• Activity-based selection limits active set to top-N torrents for performance.
• Strategy switch keeps Phase 1 equal-split as default; opt-in to weighted.
• /stats now surfaces managed_torrent_count and score_distribution.

Added

• ActivityScorer: scores torrents by current upload, recent activity, and peers.
• Selection: manages only top global set (global.max_managed_torrents, default 1000).
• Weighted allocator: proportional per-tracker distribution with bounds (min 10 KB/s, max 60% of tracker cap).
• Config keys:
  • global.allocation_strategy: equal (default) or weighted
  • global.max_managed_torrents: integer (default 1000)
• /config and README document new settings; /stats documents new fields.

Migration

• No breaking changes; defaults remain Phase 1 equal-split.
• To enable Phase 2, add under global:
  • allocation_strategy: weighted
  • max_managed_torrents: 1000
• Verify via /config and observe /stats for managed_torrent_count and score_distribution.

Notes

• Per-torrent bounds: min 10 KB/s, max 60% of tracker cap.
• Differential updates and batching preserved to keep API calls low.
• Borrowing across trackers remains out-of-scope for 0.2.0 (planned for Phase 3).

Links

• Compare: v0.1.0...v0.2.0
• Tag: https://github.com/mrInvincible29/Qguardarr/releases/tag/v0.2.0

Qguardarr 0.1.0 — Phase 1 MVP

Qguardarr 0.1.0 — Phase 1 MVP

Highlights

• Hard per-tracker upload limits with equal distribution.
• Fast webhook handler (<10ms) using a queue pattern.
• Gradual rollout and differential updates to reduce API churn.
• Full rollback capability with SQLite-backed change tracking.

Added

• Per-tracker upload speed limits (equal split within trackers).
• Webhook handler for qBittorrent events with background processing.
• Allocation engine with rollout percentage and differential updates.
• SQLite rollback system with REST /rollback endpoint.
• Docker deployment, Docker-based integration tests, and CI setup.
• Multi-arch Docker images (linux/amd64, linux/arm64).
• Unlimited semantics: set max_upload_speed: -1 for no cap (catch-all supported).

Notes

• Catch-all tracker should be last (pattern: .*).
• Matching precedence: specific tracker patterns take priority over catch-all.
• Multi-tracker handling (Phase 1): we use a single primary tracker per torrent (first working URL, else first non-error).
• Endpoints: /health, /stats, /stats/trackers, POST /webhook, POST /cycle/force, POST /rollback, POST /rollout.

Links

• Tag: https://github.com/mrInvincible29/Qguardarr/releases/tag/v0.1.0