From 2692dc5b96595b97b2c70c7a9e136d42130737e9 Mon Sep 17 00:00:00 2001 From: Inyxiel Date: Thu, 20 Aug 2026 16:33:55 +0300 Subject: [PATCH 1/4] Add CP07 About & Help R3 release notes --- docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md | 64 ++++++++++++++++++++++++++ 1 file changed, 64 insertions(+) create mode 100644 docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md diff --git a/docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md b/docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md new file mode 100644 index 0000000..e689a1a --- /dev/null +++ b/docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md @@ -0,0 +1,64 @@ +# AMP Lab v8.0.0 — CP07 About & Help R3 Public Test + +Release date: 2026-08-20 +Status: **Public Test — current checkpoint** + +## Release summary + +CP07 builds on the v8 public-test line with a new in-app **About / Help** surface and the expanded **Benchmark Tier List** work. The optimizer, Party Forge, save reader, and physical-item allocation remain separate from the guide/presentation layer. + +## About / Help + +- Adds a dedicated **About AMP Lab** page and Help entry point. +- Explains the end-to-end flow from ruleset selection and save reading through Build Recommendations, Synergies, progression, Respec Suggestions, and Full Party Rebuild. +- Documents key safety guarantees: **owned items only**, **one item / one owner**, **ruleset isolation**, and **fail-closed unknown mechanics**. +- Adds plain-language explanations for labels such as **FORGE VALIDATED**, **BENCHMARK**, **RECOMMENDED PICK**, and **TRADE-OFF / ALTERNATIVE**. +- Includes troubleshooting guidance for ruleset mismatches, missing items, slow rebuilds, and unexpected scores. + +## Benchmark Tier List + +### Rulesets + +- Restores a clear split between **Pure Vanilla** and **D&D 5.5e Mod** subclass benchmarking. +- The D&D 5.5e Mod benchmark remains explicitly marked **WIP / not finalized**. +- Vanilla uses the isolated BG3 Vanilla + GustavX progression model without D&D Beyond progression. + +### Vanilla scoring modes + +The Vanilla subclass benchmark supports separate states rather than locking every comparison to one equipment mode: + +- **Naked** +- **Orb Only** +- **Ring Only** +- **Orb + Ring** + +The Vanilla matrix covers **58 subclasses**, Levels **1–12**, all **14 AMP Orb families with dedicated matching Rings**, and the fixed Orb/Ring benchmark states used by AMP Lab. + +### Ranking views + +- **Selected Family** +- **Overall Across Families** +- **Best Compatibility** + +Compatibility is based on source-proven subclass ↔ AMP cross-activation rather than name/theme matching. + +### Rarity labels + +The player-facing benchmark UI now uses the actual rarity names: + +- **Uncommon** +- **Rare** +- **Epic** +- **Legendary** + +Internal numeric tier indices remain implementation details and are no longer the primary UI labels. + +## Safety / optimizer integrity + +CP07 presentation and benchmark work does not replace the underlying Party Forge / optimizer decision engines. The application remains read-only with respect to BG3 saves. + +## Windows binary + +- **Asset:** `AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.exe` +- **Size:** `87,361,536` bytes +- **SHA-256:** `8210e950bc218b8dc9afbe5f4d2dd8d303c5fa531d498080b4810594b20c308a` From 90acd36585430f6d9756c11db1aee9d58862f50a Mon Sep 17 00:00:00 2001 From: Inyxiel Date: Thu, 20 Aug 2026 16:34:25 +0300 Subject: [PATCH 2/4] Update README for CP07 public-test checkpoint --- README.md | 56 ++++++++++++++++++++++++++++++++++++++----------------- 1 file changed, 39 insertions(+), 17 deletions(-) diff --git a/README.md b/README.md index 6b45519..7586251 100644 --- a/README.md +++ b/README.md @@ -4,36 +4,58 @@ Baldur's Gate 3 build and inventory optimizer focused on real party/inventory da ## Current public-test checkpoint -**AMP Lab v8.0.0 — CP05 Unified Respec R3** +**AMP Lab v8.0.0 — CP07 About & Help R3** -CP05 is the current verified public-test checkpoint. Development continues separately on later checkpoints, so this release is intentionally marked as a **public test** rather than a final stable build. +CP07 is the current public-test checkpoint. It builds on the v8 optimizer baseline with a new in-app About / Help guide and the expanded Benchmark Tier List for Pure Vanilla and D&D 5.5e Mod subclass comparisons. The Windows application reads real BG3 `.lsv` saves directly, detects the active party and owned physical item instances, and presents its results through a BG3-inspired interface. It is read-only and does not modify the save or the game. -Current verified public-test binary: +Current public-test binary: -- File: `AMP_Lab_v8.0.0_CP05_UNIFIED_RESPEC_R3.exe` -- SHA-256: `a060656d0875534602a16bb80722a2cf32c971e59d6a6cb0863278113ff35116` +- File: `AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.exe` +- SHA-256: `8210e950bc218b8dc9afbe5f4d2dd8d303c5fa531d498080b4810594b20c308a` - Size: `87,361,536` bytes -- Status: **Public Test — QA checkpoint passed** +- Status: **Public Test — current checkpoint** Windows builds are distributed through [GitHub Releases](https://github.com/Inyxiel/AMP-Lab/releases). +See [CP07 release notes](docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md) for the current checkpoint details. + ## Current features - **Party Forge** — optimized four-character equipment allocation using the physical item instances actually owned by the party; unique items are not duplicated across characters. - **Unified Respec Suggestions** — combines the source-aware v8 Coach shortlist, deep structural search, and final whole-party Forge validation into one surface. - **Recommended / Single-Class / Alternative** — presents one validated recommendation, one single-class reference, and one alternative trade-off per character while suppressing component-identical duplicate cards. - **Level Progression** — displays class/subclass level splits and canonical Level 1–12 feature progression for Vanilla and D&D 5.5e & Beyond rulesets. +- **Benchmark Tier List** — separates **Pure Vanilla** from the **D&D 5.5e Mod** WIP benchmark. Vanilla supports **Naked**, **Orb Only**, **Ring Only**, and **Orb + Ring** scoring, plus Selected Family, Overall Across Families, and Best Compatibility views. +- **AMP rarity labels** — benchmark Orb/Ring controls use **Uncommon / Rare / Epic / Legendary** instead of internal T1–T4 labels. +- **About / Help** — plain-language guide to the app flow, each major result type, benchmark labels, safety guarantees, and troubleshooting. - **Archetypes** — shows inventory-supported archetype readiness, tiers, available and missing mechanics, supporting items, subclass fit, and best available builds. - **Synergies** — identifies item-to-item synergy chains and explains why the pieces work together. - **Arsenal** — analyzes unused owned gear, alternative synergy directions, future Orb pieces, and compatibility. - **Shards** — canonical shard recommendations with clearer AMP-facing presentation. - **Data Sources** — AMP, AMP+, D&D 5.5e & Beyond, and compatibility-source status and management. -- **Responsive deep search** — the heavy Step9 structural search runs in a background Worker with progress stages, live counters, cache controls, and real cancellation that discards partial results. +- **Responsive deep search** — the heavy structural search runs in a background Worker with progress stages, live counters, cache controls, and real cancellation that discards partial results. - **Dark / Light themes** — including the v8 light-theme contrast QA pass. -## CP05 verification +## Benchmark Tier List notes + +The Pure Vanilla subclass benchmark uses the isolated BG3 Vanilla + GustavX progression model without D&D Beyond progression. + +Vanilla benchmark modes: + +- **Naked** +- **Orb Only** +- **Ring Only** +- **Orb + Ring** + +The Vanilla matrix covers **58 subclasses**, Levels **1–12**, and all **14 AMP Orb families with dedicated matching Rings**. Compatibility scoring is based on source-proven subclass ↔ AMP cross-activation rather than name/theme matching. + +The **D&D 5.5e Mod** subclass benchmark remains explicitly marked **WIP / not finalized**. + +## Public-test validation notes + +The previous CP05 Unified Respec R3 checkpoint established the verified optimizer/runtime baseline, including: - Vanilla real-browser regression: **12/12 PASS** - D&D 5.5e & Beyond real-browser regression with Artificer data: **12/12 PASS** @@ -41,19 +63,18 @@ Windows builds are distributed through [GitHub Releases](https://github.com/Inyx - Global ruleset router regression: **20/20 PASS** - Icon pipeline browser regression: **10/10 PASS** - Level Progression browser regression: **13/13 PASS** -- CP05 semantic/frozen-boundary regression: **6/6 PASS** - Deep Worker cancellation: **PASS** - Console errors, page errors, failed requests, and 404s: **0** -- Extracted executable identity: **848/848 files identical** to the tested build -See [CP05 QA](docs/CP05_UNIFIED_RESPEC_R3_QA.md) and [CP05 release notes](docs/RELEASE_NOTES_CP05_PUBLIC_TEST.md) for details. +CP07 adds the current About / Help and Benchmark Tier List presentation/data work on top of the v8 public-test line. See [CP05 QA](docs/CP05_UNIFIED_RESPEC_R3_QA.md) for the historical baseline and [CP07 release notes](docs/RELEASE_NOTES_CP07_PUBLIC_TEST.md) for current changes. ## Current public-test limitations -- CP05 identifies canonical Feat/ASI opportunities and other player-choice points, but it does not yet select a named Feat, ASI distribution, or Eldritch Invocation. Source-aware named-choice recommendations are continuing after this checkpoint. +- The **D&D 5.5e Mod Benchmark Tier List** is still WIP and should not be treated as a frozen final ranking. +- Named Feat/ASI and Eldritch Invocation selection remains ongoing work where the source does not support a validated named pick. - Some items may display a generic slot symbol when a packaged/local item icon is unavailable. This is a presentation limitation only; it does not change item identity, scoring, allocation, mechanics, or recommendations. - When the current save does not expose a complete equipped-item baseline, Party Forge shows **SAVE BASELINE INCOMPLETE** instead of inventing an inaccurate `% vs save` value. -- Full Party Rebuild is preserved but is not invoked by the CP05 presentation; its complete level-progression and global-inventory work belongs to a later checkpoint. +- Full Party Rebuild remains intentionally heavy and is separate from the normal lightweight recommendation flow. ## Reporting public-test feedback @@ -61,11 +82,12 @@ When reporting an issue, please include: 1. The selected ruleset. 2. The character/build involved. -3. A screenshot of the result or error. -4. What you expected to happen. -5. Any error message shown by the application. +3. The tab or benchmark mode involved. +4. A screenshot of the result or error. +5. What you expected to happen. +6. Any error message shown by the application. -Useful reports include incorrect item or mechanic interpretations, unreasonable equipment/build choices, missing icons, class/subclass/multiclass recommendations, ruleset accuracy, crashes, performance problems, and confusing UI explanations. +Useful reports include incorrect item or mechanic interpretations, unreasonable equipment/build choices, missing icons, class/subclass/multiclass recommendations, benchmark/ruleset accuracy, crashes, performance problems, and confusing UI explanations. ## Release model From 8a8fd44a8797efb547837ad8cb70fd05abd8b530 Mon Sep 17 00:00:00 2001 From: Inyxiel Date: Thu, 20 Aug 2026 16:34:32 +0300 Subject: [PATCH 3/4] Add CP07 binary checksum --- checksums/AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.sha256 | 1 + 1 file changed, 1 insertion(+) create mode 100644 checksums/AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.sha256 diff --git a/checksums/AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.sha256 b/checksums/AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.sha256 new file mode 100644 index 0000000..3bd53d2 --- /dev/null +++ b/checksums/AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.sha256 @@ -0,0 +1 @@ +8210e950bc218b8dc9afbe5f4d2dd8d303c5fa531d498080b4810594b20c308a AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.exe From 5f0a8881ce36b89fb3e6e622d09628f60b5de53f Mon Sep 17 00:00:00 2001 From: Inyxiel Date: Thu, 20 Aug 2026 16:34:40 +0300 Subject: [PATCH 4/4] Register CP07 checksum --- checksums/SHA256SUMS.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/checksums/SHA256SUMS.txt b/checksums/SHA256SUMS.txt index 36a2fec..2325656 100644 --- a/checksums/SHA256SUMS.txt +++ b/checksums/SHA256SUMS.txt @@ -1,3 +1,4 @@ b4367fc3ab6c46123c31b19d2f795e5b765e079179d618b954dcc19ca1ccd571 AMP20Lab_v7.4.8_Raw-Arsenal-Direct-Shards.exe bee3bd10c42dc0d3f5019471bf6aaecd12e9f470755f87bdb840cdf90f36d556 AMP_Lab_v8.0.0.exe a060656d0875534602a16bb80722a2cf32c971e59d6a6cb0863278113ff35116 AMP_Lab_v8.0.0_CP05_UNIFIED_RESPEC_R3.exe +8210e950bc218b8dc9afbe5f4d2dd8d303c5fa531d498080b4810594b20c308a AMP_Lab_v8.0.0_CP07_ABOUT_HELP_R3.exe