Static method was referencing bare gameState which is always null.
Now accepts gameState as parameter, and all 5 callers pass it through.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Was reading gameState.battlefield (doesn't exist) causing AI to
receive empty board state. Now reads gameState.battlefieldCards
and matches by ownerIndex.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add bright silver 3D smoky border around commander card zones
- Wrap all 4 life panels in a single bordered container
- Add drag-and-drop from hand zone to battlefield with ghost preview
- Update card placement to center of battlefield, avoiding overlaid zones
- Fix card dimensions in placement calc (110x150 -> 63x84)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Move life/cmd-dmg/poison into 3-box life panels as a top-level
overlay (z-index 50) centered at quadrant intersections
- Remove center emblem
- Match battlefield card size to hand cards (63x84px)
- Overlay zone-row and commander-row on battlefield edges to
eliminate dead space and maximize battlefield area
- Restructure poison badges from per-player absolute positioning
into life-panel counters box
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Scale .hand-thumb from 32x45px to 63x84px so cards fill the
hand zone vertically while maintaining MTG card proportions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Forge's sim mode does not honour --policy/--policyServer CLI flags, so
the Python brain was never called during live games. AI only played lands
because Forge ran its built-in AI for every decision.
Fix: when parseForgeLine() detects a Priority event for an AI seat, the
session now immediately calls policyClient.decide() with the current
state snapshot and POSTs the result to /api/policy/override before Forge
reads the next line. This mirrors what handleHumanTurn() already does
for seat 0. The synthetic fallback path is unchanged.
- Add _export_forge_episode() adapter: translates ForgeEpisode transitions
back into DecisionExporter.record_step() calls so Forge-sourced wins flow
into the distillation JSONL pipeline (closes Phase 1 for real games)
- ppo_consumer() now accepts an optional `exporter: DecisionExporter` arg;
begin_episode / record_step / end_episode are called around each episode
as it is dequeued
- run_forge_pipeline() forwards the exporter kwarg down to ppo_consumer and
calls exporter.flush() after the pipeline completes, returning the output
path in the stats dict under "exporter_file"
- Fix snapshots_to_episode(): compute_terminal_reward() was called with
non-existent kwargs `turns=` and `life_remaining=`; compute_intermediate_reward()
was called with non-existent kwargs `life_delta=`, `cards_drawn=`,
`creatures_played=`. Both now use the correct signatures from reward.py
(prev_state/curr_state/action_taken for intermediate; won/draw/timeout for
terminal). Forge snapshots carry `players` arrays so we reconstruct minimal
prev/curr dicts for the intermediate path.
Ref: docs/CLOSED-LOOP-DISTILLATION.md Phase 1
The Forge DecisionExporter writes action={"type":"cast"} without a card
field — action.get("card") and action.get("raw") always returned "",
causing every cast to fall through to the CAST_CREATURE default.
Fix: when card_name is empty, scan the active player's hand against
the removal / draw / ramp / commander name databases and pick the
highest-priority match found. Priority order mirrors cast importance:
removal > ramp > draw > commander > creature (default)
A hand-level match is a reasonable proxy for the decision made:
Forge's AI picks the highest-priority action from available options,
so if the hand contains a removal spell during main phase, the cast
action almost certainly represents that removal spell being played.
Also fixes print_label_distribution: was sorting by dict key (string)
instead of by index, causing the display order to be alphabetical
rather than matching the MacroAction enum order — making cast_creature
appear at position 0 in the sorted output even when it had counts.
Now prints in descending count order with correct enum-keyed lookup.
_discover_jsonl_files now recognises ml-decisions-ds-*.jsonl as its own
"ds" source key instead of falling through to legacy/forge.
DatasetConfig.source_weights gains ds=4.0 as a default so the 8 ds files
(1775 records, ~5.8% raw) are oversampled to ~20% of effective training
tokens, matching their higher signal quality.
CLI gains --ds-weight argument (default 4.0) to override without code changes.
The training_manifest.json produced by PowerShell already encodes weight=4.0
per ds file; this change makes dataset_builder.py honour the same ratio
natively so the manifest and the builder stay in sync.
Adds a single INFO log line after each successful inference showing:
- chosen action + confidence
- top-5 actions sorted by probability
- turn, phase, vector_source
This lets us diagnose the 'only plays lands' behaviour without a Python
REPL. Remove or downgrade to DEBUG once the training issue is resolved.
Forge sends cmdr_dmg as a dict keyed by opponent seat (e.g. {"0": 14,
"2": 7}). _encode_global read it raw and crashed on `dict / 21.0`.
Fix: resolve cmdr_dmg to a scalar at the read site — sum dict values if
it's a dict, cast to int if it's already a number, default to 0.
This is defensive for all call paths (live Forge IPC, JSONL training,
_build_encoder_snapshot pre-collapsed int) so no upstream change needed.
Both 'Inference error' log lines use identical logger strings, making it
impossible to tell which except block is catching the crash. Also, broad
except blocks were swallowing the full traceback so the real crash site
and cause were hidden.
Changes:
- policy_server.py: logger renamed to 'ml.serving.policy_server';
predict() and predict_batch() error handlers add exc_info=True
- routes/policy.py: logger renamed to 'routes.policy.decide';
_infer_from_vector() error handler adds exc_info=True;
decide() fallback warning adds exc_info=True
The next crash will print the full stack trace, pinpointing the exact
line and call chain causing the TypeError.
The crash `unsupported operand type(s) for /: 'dict' and 'float'` was
firing from _infer_from_vector(), which is the code path taken when
has_precomputed_vector() is True (GameSession 1.1.0+ snapshots with
a precomputed state_vector float[29]).
Unlike PolicyInferenceService.predict() — which was fixed in the
previous commit via _unwrap_logits — this function calls
policy_service.model(state_tensor) directly and immediately uses the
return value in arithmetic (logits / max(temperature, 0.01)), so any
tuple/dict return (DataParallel, torch.compile, PolicyValueNetwork
swap) causes the TypeError.
Fix: import _unwrap_logits from ml.serving.policy_server and apply it
to the raw model output in _infer_from_vector before the temperature
division and argmax calls. All three return shapes are now handled:
- plain tensor → passthrough (normal case, zero overhead)
- tuple/list → raw[0] (PolicyValueNetwork (logits, value))
- dict → first recognised key or first tensor value
If self.model is wrapped (DataParallel, torch.compile, or future
PolicyValueNetwork path), forward() can return a tuple (logits, value)
or even a dict. Add an explicit unwrap immediately after the forward
call in both predict() and predict_batch() so every downstream
temperature/softmax operation always receives a plain tensor.
The root cause of the persistent /api/policy/health 404:
- app.mount("/", StaticFiles(...)) is called at module load time
- StaticFiles acts as a sub-app and consumes matching requests,
returning its own 404 for unknown files rather than falling through
- register_policy_routes() was called in lifespan (runtime), AFTER
the "/" mount was already in the route list — so the policy router
was effectively unreachable
Fix: move route registration to module scope right after all other
app.include_router() calls and BEFORE any app.mount() calls. Only
the model load (_POLICY_SVC.load()) stays in lifespan.
Previously /api/policy/health returned 404 when the model wasn't loaded
because register_policy_routes() was gated behind _policy_svc.load().
The health endpoint itself already handles the unloaded state gracefully
(returns {ready: false, status: degraded}), so the router must be mounted
unconditionally so ai-bridge.js can poll it during startup.
Fixes 404 errors from lan-client ai-bridge.js which calls these routes
to populate the deck selector. Routes were missing from lab.py — only
precon_router (routes/precon.py) existed but served different paths.
Also added PRECON_DECKS_DIR env var alias in lab_api.py _parse_args()
to match the .env variable name users commonly set.
- Mount lan-client/ as static files at /lan with html=True
- Mount precon-decks/ at /precon-decks/ for direct .dck access
- ai-bridge.js: use same-origin (empty URLs, auto-detect from window.location)
- engine.js: simplify deck fetch to use /api/lab/precons/deck endpoint + /precon-decks/ static fallback
- Eliminates cross-origin issues by serving everything from localhost:8080
Access the LAN battlefield at http://localhost:8080/lan/
Three fixes that prevented precon decks from loading into actual gameplay:
1. ui.js: window.startGame() now calls startGameWithDecks() when
setupDeckChoices has entries (the 3-tab selector stores choices there).
Previously it called initGame() which looked for old deck-sel-{i}
dropdowns that no longer exist.
2. engine.js: loadPreconForPlayer() now fetches .dck file content from
the backend (/api/lab/precons/deck?fileName=...) when precon.decklist
is empty. Falls back to fetching precon-decks/{file} directly.
PRECON_DATA entries only have metadata (name, commander, set) — the
actual card list lives in the .dck files.
3. engine.js: parseDeckList() now handles .dck bracket-style section
headers like [Commander], [Main], [metadata]. Previously only matched
bare text like 'COMMANDER:' without brackets. Also skips [metadata]
lines (e.g. 'Name=Elven Empire').
4. routes/precon.py: Added GET /api/lab/precons/deck?fileName=... endpoint
to serve raw .dck file content for the LAN client.
- Add max-height: 92vh and overflow-y: auto to .setup-card
- Widen .setup-card to 860px to fit 2-column deck grid
- Change .deck-assignment-grid from vertical stack to 2-column grid
so 4 player panels render in 2 rows instead of 4
- Tighten .deck-tab-content (170px) and .deck-precon-list (140px)
- Make top gradient bar sticky on scroll
- Responsive fallback to single column below 640px
- Generate lan-client/precons.js from precon-decks/precon-index.json so
PRECON_DATA is always available for local/offline mode (163 precons)
- Add scripts/generate-precons-js.js for regeneration when precons update
- Fix _loadBackendPrecons() to re-render the precon tab UI when backend
precons load successfully (was only updating status indicator text,
not populating the dropdown)
Without precons.js, PRECON_DATA was undefined and the dropdown showed
'No precon data loaded' even though 163 .dck files exist in the repo.
* feat: LAN Battlefield App integration with Commander-AI-Lab AI pipeline
Add the LAN Battlefield App as a browser client (lan-client/) that
connects to the Commander-AI-Lab backend for ML-powered AI opponents,
3-category deck selection, and enhanced player panels.
Phase 1 - Foundation:
- Remove floating mana pool display (keep internal tracking)
- Add energy, experience, rad, emblems fields to game state
- No-op renderManaPool(), showManaPool(), setupManaPoolDrag()
Phase 2 - Backend Connection (ai-bridge.js):
- BackendConnector: REST + WebSocket connectivity with retry/timeout
- GameStateBridge: Serializes engine.js state for /api/policy/decide
- MacroActionMapper: Translates 8 ML macro-actions to engine.js calls
- DeckSourceClient: Fetches precon/EDHREC decks from backend API
Phase 3 - Deck System:
- 3-tab deck selector (Precons / EDHREC / My Decks) replaces dropdown
- Precon grouping by set with collapsible sections and filtering
- EDHREC search integration (stub, ready for backend implementation)
- My Decks textarea import placeholder
Phase 4 - AI Integration:
- runAiTurn() priority chain: ML Pipeline > Forge > Local Heuristic
- _runMlAiTurn() async turn loop with multi-phase decision requests
- Backend session initialization on game start
- Automatic fallback to local AI on any ML pipeline error
Phase 5 - Enhanced Player Panels (player-panels.js):
- Collapsible panels with localStorage persistence
- Commander damage progress bars with warning/critical thresholds
- Poison, energy, experience, rad counter widgets
- Emblem tracking with add/remove
Phase 6 - Polish:
- Script loading order: precons > engine > ai-bridge > panels > ui
- Backend auto-detection on page load with status indicator
- Connection status display on setup screen
Closes#184, Closes#185, Closes#186, Closes#187, Closes#188, Closes#189
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: refine LAN battlefield integration with full JSDoc, proper phase IDs
- Add comprehensive JSDoc comments to all ai-bridge.js classes
- Use optional chaining for PHASES lookup in GameStateBridge
- Direct addLogEntry calls in player-panels.js (loaded after engine.js)
- Remove leftover mana pool CSS that was missed in initial pass
- Minor style.css cleanup for consistency
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
When running plain `python script.py` or bare uvicorn, .env is never
loaded automatically. Added a dotenv bootstrap block at the top of
state.py that finds and loads the nearest .env file (walks up from
models/ to the project root). Uses load_dotenv with override=False so
explicitly set environment variables always win over .env values.
Gracefully skips if python-dotenv is not installed.
CFG.forge_dir, precon_dir, forge_decks_dir, forge_jar, and lab_jar
were only set via CLI args. When the server starts via uvicorn directly
(no CLI), all paths stayed empty and forge_card_loader could not find
the cardsfolder, causing 0 lands in every deck.
Fix: read FORGE_DIR, FORGE_JAR, PRECON_DIR, PRECON_DECKS_DIR,
FORGE_DECKS_DIR, and LAB_JAR from environment at import time, mirroring
the existing ANTHROPIC_API_KEY / XIMILAR_API_KEY pattern.
_find_dck_file only searched CFG.precon_dir / CFG.forge_decks_dir, both
of which are empty when the server is started without CLI args. This
caused all precon deck lookups to return 0 cards.
Fix: add _REPO_ROOT and _BUILTIN_DECK_DIRS constants, then append
precon-decks/, sample-decks/, imported-decks/ to the search list
whenever they exist on disk, regardless of whether CFG paths are set.
_load_deck_cards_from_dck was hardcoding type_line="" for every card,
causing Card.is_land() to return False for all lands when a deck is
loaded via the .dck file path (i.e. not in the SQLite DB). This meant
the engine never recognised lands as lands, breaking mana production.
Fix: call lookup_forge_card() per card and populate type_line,
oracle_text, cmc, mana_cost, power, toughness, and keywords from the
Forge cardsfolder cache. The cache is lazy-loaded and module-level so
there is no per-game overhead after the first deck load.
Precon decks (Elven_Empire, Blood_Rites, etc.) are stored as .dck files
on disk but not inserted into the SQLite decks/deck_cards tables.
The DeepSeek batch engine calls _load_deck_cards_by_name() which
previously returned [] for any deck not in the DB, causing
'No decks could be loaded.' errors.
Adds _find_dck_file() and _load_deck_cards_from_dck() helpers.
_find_dck_file searches CFG.precon_dir, CFG.forge_decks_dir, and
FORGE_DIR/res/decks/commander with exact + fuzzy matching.
_load_deck_cards_by_name now tries DB first, then .dck fallback.
- start-lab.bat: Set FORGE_DECKS_DIR to D:\ForgeCommander\commander-ai-lab\precon-decks
- BatchRunner.java: resolveForgeDecksPath() now checks PRECON_DECKS_DIR env var
first, then falls back to the hardcoded precon-decks path, before trying
%APPDATA%\Forge\decks\commander. Also pass full absolute path to Forge -d flag.
- .env.example: document PRECON_DECKS_DIR variable
Captures life-at-timeout snapshots for every draw game via [DRAW] [PARSE-SUMMARY] log lines. Adds services/draw_game_analyzer.py with DrawGameSnapshot + summarize_draw_games(). Wires drawGames summary into GET /api/lab/result response. Lowers batch clock default from 6000 to 3000 (30s per-game). Closes#176.
- Add validate_card_names() helper: checks collection_entries DB first,
then falls back to Scryfall /cards/named?fuzzy= for unknowns.
Drops any name that cannot be resolved (hallucinated cards).
- Inject validation in Step 4 immediately after each suggest_cards()
call so hallucinated names never reach ratio enforcement or assembly.
- Log dropped hallucinations at WARNING level with count + examples.
- Resolves the 'Sonic Hedgehog Charger' / 'Sonic Boom' class of
LLM-invented card names that were silently entering the deck.
- Add _parse_card_list_fallback() to recover plain-text CSV output from
the LLM when json_mode fails to enforce JSON (e.g. qwen2.5:7b drift)
- Harden suggest_cards system prompt with a one-shot JSON example so
small models are less likely to return prose in the first place
- Log recovered names at WARNING level so the degradation is visible
without being silent (previously returned [] with no recovery attempt)
- No behaviour changes to any other function
dict() on a types.SimpleNamespace raises TypeError: object is not iterable.
vars() correctly extracts the __dict__ attribute map. Fixes
TestBuildCollectionFailureTracking::test_high_failure_rate_raises_runtime_error.
The three split-combat helpers were accidentally defined as nested
functions inside _resolve_combat(), making them unreachable as
GameEngine.assign_attackers() etc. from tests and turn_manager.
Move them out to class-level methods with correct 4-space indentation.
_resolve_combat() now remains as the backward-compatible atomic path for
run_n(); the split methods are available for turn_manager priority
windows and direct testing.
begin_combat/end_combat are dispatched as named priority windows inside
_phase_combat(), not as entries in RESPONSE_PHASES. The constant is
correctly {"upkeep", "end_step"} — test was asserting stale intent.
Delta matches \bdelta\b in the nonbasic land guard regex, causing enrich_card
to classify it as a land (cmc=0). Zeta and Theta have no land keyword matches.
Surfaces previously-computed-but-unused stats into the Claude prompt:
1. USER_PROMPT_TEMPLATE gains a new '### Sim Insights' section that shows:
- Deck-level opening hand quality: avg keptInOpeningHandRate across all
cards, plus the 3 worst offenders with their actual rates
- Curve velocity: avg avgTurnCast weighted by castRate, flagging cards
cast on avg turn > 6 as 'late'
- Synergy outliers: bottom-3 synergyScore cards flagged as 'isolated'
- Archetype weak spots: archetypes with winRate < 40% flagged explicitly
2. _format_archetypes() now annotates weak spots with [WEAK SPOT] and
strong matchups with [STRENGTH] so Claude sees them immediately.
3. _format_top_cards() adds keptInOpeningHandRate to each line so
the 'good' cards also show their hand-keepability signal.
4. New helper _format_sim_insights() computes all deck-level aggregates
from CardPerformance list — no new fields on DeckMeta needed.
No schema changes. Fully backward-compatible with existing DeckReport.
New test file covering:
- TestGlobalFeatureDimensions: output shape (29,), all values in [0,1]
- TestTotalPowerRealValue: real total_power used at index 7 when present
- TestTotalPowerHeuristic: creatures*3 fallback when total_power absent/zero
- TestLandCount: real lands used at index 13 when present; defaults to 0
- TestCreatureCount: real creatures used at index 6; fallback to bf len
- TestManaFeature: mana normalised and clamped to [0,1]
- TestPhaseOneHot: exactly one hot in indices 9-12 per player block
- TestIsActivePlayer: binary 0/1 at index 8
- TestLifeNormalisation: 40 life → 1.0, 0 life → 0.0
- TestTurnFeature: turn clamped to [0,1]
- TestFullEncode: total state vector shape (6177,), dtype float32
- TestEmptyPlayer: _empty_player() has all Step 4 keys
- TestLegacySnapshot: old JSONL snapshot without Step 4 keys still encodes"
Previously _encode_global() always computed:
total_power = creatures * 3 (heuristic)
Now reads p.get('total_power', 0) first, falling back to the heuristic
only when the key is absent (legacy JSONL snapshots without Step 4 data).
Also wires total_toughness, artifacts, and enchantments into _empty_player()
so padding players have consistent keys.
- TestBoardStatsWiring: verifies creaturesOnField, landCount,
totalPowerOnBoard, totalToughnessOnBoard, artifactsOnField,
enchantmentsOnField are accepted without 422 and survive
a round-trip through the fallback encoder path
- TestBoardStatsFallback: verifies resolver fallbacks when Forge
fields are absent (creatures inferred from bf length,
total_power falls back to creatures * 3 heuristic,
lands defaults to 0)
- TestSchemaAlias: verifies snapshot_schema Field(alias='schema')
still accepts JSON key 'schema' without 422
- Removes now-irrelevant UserWarning (schema shadow fixed)
- ml/eval/forge_evaluator.py: reusable eval harness wrapping
MultiThreadBatchRunner + PolicyInferenceService; collects win rate,
avg turns, life differential, commander damage, illegal action rate,
policy entropy per game; writes results/eval-{run_id}.json
- scripts/eval_policy.py: CLI to run N Forge eval games with any
checkpoint vs forge-builtin or policy-server opponent; supports
--threads, --clock, --seed, --greedy, --temperature
- scripts/compare_eval.py: loads two result JSONs, computes absolute
win-rate delta + 95% CI, runs Fisher's exact test for significance,
prints formatted comparison table and pass/fail verdict
Previously _build_encoder_snapshot() stubbed:
- lands = 0 (always zero)
- creatures = len(bf_names) (name-list length, ignores tokens/etc)
- mana = mana_available (was already correct)
- total_power / total_toughness / artifacts / enchantments = missing
Now reads real pre-computed values from PlayerZoneState:
- creatures ← creaturesOnField (Forge board stat)
- lands ← landCount (Forge board stat)
- total_power ← totalPowerOnBoard (fed to global feature index 7)
- total_toughness← totalToughnessOnBoard
- artifacts ← artifactsOnField
- enchantments ← enchantmentsOnField
- mana ← manaAvailable (unchanged)
Also replaces the `creatures * 3` proxy in _build_encoder_snapshot with
the real totalPowerOnBoard value so index-7 of the global feature vector
tracks actual power rather than a heuristic.
PlayerZoneState gains 5 new optional int fields (default 0) matching
DecisionSnapshot.PlayerSnapshot field names so JSON from both the live
IPC path and the JSONL training path deserialise without changes.
Covers:
- health endpoint returns {ready, status, model_loaded}
- precomputed fast path: state_vector float[29] → vector_source=precomputed
- correct 6177-dim tensor shape reaches the mock model
- all 8 MacroAction values are valid responses
- encoder fallback: missing state_vector → vector_source=encoder
- encoder fallback: wrong-length state_vector → vector_source=encoder
- GameSession field name aliases (turnNumber, activePlayer,
commanderTax, handCount, commandZone) are accepted without 422
- 503 when model not loaded
- /api/policy/stats tracks precomputed_vector_hits vs encoder_fallbacks
- both phase strings (MAIN1, main_1) round-trip correctly
Uses a MockPolicyInferenceService that captures the tensor shape passed
to the model so we can assert on dim without needing a real checkpoint.
DecideRequest now accepts an optional state_vector field (float[29]).
When present, /api/policy/decide uses it directly for the global scalar
block instead of re-deriving it from nested player fields.
StateEncoder gets a new encode_with_precomputed_global() path that:
- Accepts the precomputed float[29] global block from Java
- Still runs full zone embedding pooling (hand/battlefield/graveyard/
command_zone) from the players[] zone contents
- Appends the playstyle one-hot as normal
- Returns the same 6177-dim tensor the policy network expects
Fallback: if state_vector is absent or wrong length, encode() is called
as before so all existing callers (batch predict, JSONL training) are
unaffected.
Also maps GameSession player field names to encoder field names:
handCount -> hand size proxy when hand[] is empty (AI seats)
battlefield list -> creatures count proxy
commanderTax -> cmdr_tax
turnNumber -> turn
activePlayer -> active_seat
Replaces synthetic stub loop with a live Forge process:
- Spawns Forge with --policy + --policyServer flags so Forge calls
back to /api/policy/decide at each AI decision point
- Reads Forge stdout line-by-line and parses zone/phase/life updates
via the same regex patterns used in BatchRunner
- Syncs full board state (battlefield, hand, graveyard, exile,
command zone, stack) on every Forge state-change line
- Routes human input from humanActionQueue via a side-channel
HTTP POST to /api/policy/override on the policy server
- Handles activePlayer + priorityPlayer correctly for 3-4 player pods
- Falls back to synthetic stub loop if Forge subprocess fails to start
- Watchdog thread kills stalled Forge process after 10 min
- _write_build_meta: after writing build_meta.json, also stamps built_at
onto the collection via col.modify() (best-effort, swallows exceptions)
- _collection_is_current: if sidecar built_at is missing/zero, falls back
to collection.metadata["built_at"] so a deleted sidecar doesn't force
an unnecessary full rebuild when the ChromaDB data is still valid
- Add BATCH_RETRY_LIMIT and BATCH_RETRY_DELAY constants after DEFAULT_BATCH_SIZE
- Add _insert_batch_with_retry() helper with exponential backoff
- Replace both bare try/except insert blocks in _build_collection_from_bulk
with calls to the retry helper so transient ChromaDB errors are retried
up to BATCH_RETRY_LIMIT times before a batch is counted as failed
Adds a buildStateVector(int seat) method to GameSession that produces
a float[29] global scalar vector matching STATE_DIMS layout in
ml/config/scope.py:
Per-player block (14 features × 2 players = 28):
[0] life_total / 40.0
[1] commander_damage / 21.0 (stub 0.0 until real Forge tracks it)
[2] mana_available / 20.0 (stub 0.0 until mana pool wired)
[3] commander_tax / 10.0
[4] cards_in_hand / 15.0
[5] cards_in_graveyard / 100.0
[6] creatures_on_battlefield / 30.0
[7] total_power_on_board / 100.0 (stub = creature count until P/T data)
[8] is_active_player (0 or 1)
[9] phase one-hot[0] main1
[10] phase one-hot[1] combat
[11] phase one-hot[2] main2
[12] phase one-hot[3] end
[13] (padding / reserved)
Global:
[28] turn_number / max_turns (25)
buildStateSnapshot() now includes "state_vector": float[] so that
PolicyClient.decide(JsonObject) sends a machine-readable vector the
Python /api/policy/decide endpoint can pass directly to the encoder
without re-parsing the nested human-readable fields.
Stub fields (mana_available, commander_damage, total_power_on_board)
are zero until real Forge board state is wired in Step 4.
Wires forge_episode_generator into the PPO training loop:
- New --data-source flag: synthetic (default) | forge | mixed
- New --forge-results-dir, --forge-batch-id, --forge-weight flags
- ForgeDataSource class: runs run_forge_pipeline() in background thread,
drains episodes into RolloutBuffer via episodes_to_buffer()
- Mixed mode: blends Forge episodes + synthetic self-play per iteration
using --forge-weight ratio (0.0–1.0)
- Synthetic path unchanged — no regression for existing training runs
