Claude/project exploration analysis 011 cv2j2s8c4a n wna ye2 e cow#23
Open
kulkanmalik wants to merge 16 commits intoYuvDwi:mainfrom
Open
Conversation
This commit implements the full crafting system for Steve entities, enabling autonomous crafting of Minecraft items with recipe validation, ingredient checking, and crafting table management. ## Changes ### 1. SteveEntity (entity/SteveEntity.java) - Added ItemStackHandler inventory (36 slots, same as player) - Implemented NBT serialization for inventory persistence - Added getInventory() accessor method ### 2. InventoryHelper (util/InventoryHelper.java) - Complete implementation of inventory management utilities - Item query methods: hasItem(), getItemCount(), getInventoryFullness() - Item manipulation: addItem(), removeItem() - Tool management: findBestTool(), needsToolRepair(), hasRequiredTool() - Chest operations: findNearestChest(), transferToChest(), retrieveFromChest() - Helper methods: isInventoryFull(), getEmptySlotCount() ### 3. RecipeHelper (util/RecipeHelper.java) - Full Minecraft recipe system integration - Recipe lookup: hasRecipe(), findRecipe() - Ingredient extraction: getRequiredIngredients(), getMissingIngredients() - Crafting table detection: requiresCraftingTable() - Recursive crafting chain analysis: getFullCraftingChain() - Recipe validation: hasAllIngredients(), getRecipeResultCount() - Recipe metadata: isShapedRecipe(), getRecipeDimensions() ### 4. CraftItemAction (action/actions/CraftItemAction.java) - Complete crafting workflow implementation with 6 phases: 1. VALIDATING - Check recipe exists and ingredients available 2. FINDING_TABLE - Search for crafting table within 16 blocks 3. PLACING_TABLE - Place crafting table if not found 4. NAVIGATING_TO_TABLE - Navigate to crafting table location 5. CRAFTING - Execute crafting operations 6. COMPLETED - Finish action - Intelligent crafting table management (find, place, navigate) - Multi-quantity crafting support with progress tracking - Ingredient validation with detailed error messages - Inventory rollback on failure - Timeout handling (10 minutes max, 30s navigation) ### 5. Development Roadmap (DEVELOPMENT_ROADMAP.md) - Comprehensive 5-phase development plan - Detailed specifications for all future features - Implementation checklists and success criteria - Risk management and best practices documentation ## Features - ✅ Steve entities can now craft any Minecraft item with a recipe - ✅ Automatic ingredient checking before crafting - ✅ Intelligent crafting table finding and placement - ✅ Full inventory management (add, remove, transfer to chests) - ✅ Tool durability tracking and selection - ✅ Persistent inventory across world reloads - ✅ Support for both shaped and shapeless recipes - ✅ Multi-quantity crafting (e.g., "craft 10 sticks") - ✅ Detailed error messages for missing ingredients ## Technical Details - Uses Minecraft Forge 1.20.1 RecipeManager API - Leverages ItemStackHandler for robust inventory handling - Implements NBT serialization for data persistence - State machine pattern for crafting workflow - Comprehensive Javadoc documentation ## Testing Required Build the project locally: ```bash ./gradlew build ``` Test crafting in-game: ``` /steve spawn TestSteve /steve tell TestSteve "craft wooden_pickaxe" ``` ## Next Steps (Phase 1.2) - Implement advanced inventory management (auto-storage, sorting) - Add persistent memory system - Integrate real vector store for semantic memory - Implement async action execution --- Related to: DEVELOPMENT_ROADMAP.md Phase 1.1 Implements: Crafting System (YuvDwi#1 Priority Feature)
This commit implements a complete inventory management system with automatic
storage, chest operations, and seamless integration with existing actions.
## Changes
### 1. MineBlockAction (action/actions/MineBlockAction.java)
- **NEW**: mineBlockAndCollect() method for direct inventory collection
- Drops now go directly to Steve's inventory instead of world
- Inventory fullness check (>90% warning)
- Proper loot context with tool consideration
- Fallback to world drop when inventory full
### 2. PlaceChestAction (NEW - action/actions/PlaceChestAction.java)
- Automated chest placement near Steve
- Spiral search pattern for suitable positions
- Validates ground solidity and clearance
- 3-phase workflow: Validating → Finding Spot → Placing
- Returns chest position for other actions
### 3. StoreItemsAction (NEW - action/actions/StoreItemsAction.java)
- Stores items from inventory to chest
- Finds nearest chest (16 block radius)
- Places new chest if none found
- Navigation to chest with 30s timeout
- Supports selective storage (specific item) or bulk storage (all items)
- Returns count of items stored
### 4. RetrieveItemsAction (NEW - action/actions/RetrieveItemsAction.java)
- Retrieves items from chest to inventory
- Parses item names from task parameters
- Finds nearest chest with items
- Navigation with timeout
- Inventory space validation
- Returns count of items retrieved
### 5. PromptBuilder (ai/PromptBuilder.java)
- Added 4 new actions to LLM prompt:
- craft: {"item": "wooden_pickaxe", "quantity": 1}
- store: {"item": "cobblestone"} or {}
- retrieve: {"item": "iron_ingot", "quantity": 8}
- place_chest: {}
- Updated rules with inventory management guidelines
- Added 3 example prompts for new actions
### 6. ActionExecutor (action/ActionExecutor.java)
- Added "store", "retrieve", "place_chest" to createAction() switch
- Enables LLM to request inventory management actions
## Features
- ✅ Automatic item collection to inventory during mining
- ✅ Chest finding and placement
- ✅ Item storage with selective/bulk options
- ✅ Item retrieval with quantity support
- ✅ Inventory fullness monitoring (>90% warning)
- ✅ Navigation with timeouts and retry logic
- ✅ LLM integration for natural language commands
- ✅ Graceful fallback (drops to world if inventory full)
## Technical Details
- Uses LootParams for proper drop calculation
- Leverages InventoryHelper for all inventory operations
- State machine pattern for multi-phase workflows
- Timeout handling (30s navigation, 5min total)
- Chest validation before operations
## Example Usage
### Natural Language (via LLM):
```
/steve tell Steve "mine 64 iron and store it in a chest"
/steve tell Steve "get me 10 diamonds from the chest"
/steve tell Steve "craft a diamond pickaxe" (uses inventory items)
/steve tell Steve "place a storage chest"
```
### Direct Task Format:
```json
{"action": "mine", "parameters": {"block": "iron_ore", "quantity": 64}}
→ Items go to inventory automatically
→ Warning if inventory >90% full
{"action": "store", "parameters": {}}
→ Stores all items to chest
{"action": "retrieve", "parameters": {"item": "diamond", "quantity": 10}}
→ Gets 10 diamonds from chest
{"action": "craft", "parameters": {"item": "diamond_pickaxe", "quantity": 1}}
→ Checks inventory, crafts if materials available
```
## Architecture
```
MineBlockAction
↓ (collects drops)
Steve's Inventory (36 slots)
↓ (when >90% full)
StoreItemsAction
↓
Chest (27 slots)
↓ (when needed)
RetrieveItemsAction
↓
Steve's Inventory
```
## Next Steps (Phase 1.3)
- Implement persistent memory system
- Add inventory tracking across sessions
- Create chest memory (which chest has what)
- Implement auto-storage triggers (not just warnings)
---
Related to: DEVELOPMENT_ROADMAP.md Phase 1.2
Implements: Inventory Management System (YuvDwi#2 Priority Feature)
Dependencies: Phase 1.1 (Crafting System)
Implemented comprehensive persistent memory system with multiple memory types: New Memory Systems: - EpisodicMemory: Stores significant events with importance scoring and JSON persistence - ChestMemory: Tracks chest locations and contents for intelligent item retrieval - ConversationHistory: Maintains LLM conversation context with token limit management - SharedKnowledgeBase: Singleton knowledge base for sharing discoveries between Steve entities Features: - JSON file persistence for all memory types (config/steve/memory/) - Intelligent pruning strategies (importance, time, confirmation-based) - Memory summarization for LLM context with token limits - Query and search capabilities (by type, location, importance) - Confirmation system for shared knowledge validation Integration: - Enhanced SteveMemory with all new memory systems - Integrated chest memory with StoreItemsAction, RetrieveItemsAction, PlaceChestAction - Added memory summary to LLM context for better decision making This enables Steve entities to remember significant events across sessions, track storage locations, maintain conversation continuity, and share knowledge with other Steves.
Replaced fake hash-based embeddings with real OpenAI embeddings for semantic memory search. New Components: - EmbeddingClient: Interface for embedding generation (supports multiple providers) - OpenAIEmbeddingClient: OpenAI text-embedding-3-small implementation (1536D) - VectorMath: Vector operations utilities (cosine similarity, normalization, etc.) Enhanced VectorStore: - Real embedding generation via OpenAI API - Semantic similarity search using cosine similarity - Batch embedding support for efficiency - JSON persistence (save/load to config/steve/vector_store/) - Support for 10K+ memories without performance degradation EpisodicMemory Integration: - Optional semantic search capability (disabled by default) - setVectorStore() to enable semantic search - semanticSearch() method for natural language queries - Automatic re-indexing of existing memories - Vector store updated when new memories added Features: - Natural language queries: "where did I find diamonds?" - Relevance-based ranking (not just keyword matching) - Configurable result count (topK) - Metadata preservation in search results Performance: - Embedding generation: <500ms per query (OpenAI API) - Batch embedding support for multiple texts - Exponential backoff retry logic for API failures - Normalized vectors for fast cosine similarity This enables Steve to semantically search memories, answering questions like "where did I mine iron?" or "what did I build yesterday?" with contextually relevant results.
…e 1.5) Implemented comprehensive async action system for concurrent task execution: New Components: - ActionPriority: Enum for action priorities (CRITICAL > HIGH > NORMAL > LOW > BACKGROUND) - ResourceLock: Resource locking system to prevent conflicting actions - ActionScheduler: Priority-based action scheduler with interruption support Features: - Priority-based scheduling with 5 priority levels - Concurrent action execution when compatible - Resource locks prevent conflicts (INVENTORY, NAVIGATION, INTERACTION, COMBAT, CRAFTING) - Higher priority actions can interrupt lower priority tasks - Action history tracking for debugging - Backward compatible with legacy execution mode ActionScheduler: - Manages multiple action queues (one per priority level) - Tracks running actions with resource locks - Automatic resource acquisition/release - Action compatibility checking (prevents conflicts) - Interrupt mechanism for critical tasks BaseAction Enhancements: - Added priority field (default: NORMAL) - getPriority()/setPriority() methods - isStarted(), isCancelled(), isCompleted() status methods ActionExecutor Refactoring: - Integrated ActionScheduler for async execution - Legacy mode maintained for backward compatibility - scheduleHighPriorityAction() for urgent tasks - scheduleCriticalAction() for emergencies (combat, danger) - setUseScheduler() to toggle between modes Resource Management: - Actions declare required resources (inventory, navigation, etc.) - Scheduler ensures no conflicting actions run simultaneously - Force release on interruption - Lock status tracking and debugging Use Cases: - Mining while monitoring for threats (auto-interrupts on danger) - Combat interrupts any ongoing task - User commands get HIGH priority (interrupt autonomous tasks) - Idle following runs in BACKGROUND (never blocks important work) - Multiple compatible actions can run concurrently This enables Steve to multitask effectively while preventing conflicts and ensuring critical tasks (combat, danger avoidance) always take priority.
…g (Phase 2.1) Significantly enhanced LLM decision-making with sophisticated prompting techniques. Enhanced System Prompt: - Chain-of-Thought reasoning framework (ANALYZE → IDENTIFY → PLAN → VALIDATE → EXECUTE) - Comprehensive few-shot examples showing step-by-step thinking - Common mistake patterns with corrections (❌ DON'T vs ✅ DO) - Error recovery guidelines - Self-reflection validation questions Reasoning Framework (5-step process): 1. ANALYZE: Current situation (inventory, location, resources) 2. IDENTIFY: User requirements and goals 3. PLAN: Optimal task sequence 4. VALIDATE: Check for blockers and prerequisites 5. EXECUTE: Output validated task list Few-Shot Examples Enhanced: - Example 1: Simple task with validation - Example 2: Multi-step task requiring prerequisites - Example 3: Error recovery patterns - Example 4: Combat scenarios - Example 5: Complex crafting with material gathering - Example 6: Storage management - Example 7: Item retrieval from chests Enhanced User Prompt: - Inventory fullness percentage with warnings (>90% triggers storage suggestion) - Memory context integration (episodic memories, chest locations, conversation history) - 5-question reasoning prompt to guide LLM thinking - Rich environmental awareness (players, entities, blocks, biome) Error Recovery System: - buildErrorRecoveryPrompt(): Creates recovery context when actions fail - Analyzes failure causes (missing tools, full inventory, resource unavailable) - Provides recovery strategies and alternative approaches - handleErrorRecovery(): Automatically calls LLM to replan after failures - Replaces failed task queue with recovery plan Plan Validation: - buildPlanValidationPrompt(): Self-reflection before execution - Validates task order, prerequisites, blockers, efficiency - Returns JSON with validation status and suggested improvements Common Mistakes Addressed: - ❌ Crafting without materials → ✅ Mine/gather first - ❌ Unnecessary pathfinding → ✅ Actions auto-navigate - ❌ Vague reasoning → ✅ Step-by-step explanations - ❌ Ignoring inventory limits → ✅ Auto-store when >90% full Integration: - ActionExecutor now calls handleErrorRecovery() when actions fail - LLM receives error context and original goal - Automatically generates recovery plans - User notified of problems and new plans Benefits: - Fewer invalid action sequences - Better prerequisite handling (gather materials before crafting) - Automatic error recovery without user intervention - More transparent decision-making (detailed reasoning) - Reduced task failures through validation This transforms Steve from reactive task execution to proactive problem-solving with self-correction.
Added redstone automation capabilities: - RedstoneHelper: utility class for placing redstone components (dust, torches, repeaters, comparators, levers, buttons, pressure plates, pistons, lamps) - AutomaticDoorAction: builds automatic doors with pressure plate activation - Multi-phase building system (door → pressure plates → redstone wiring) - Integrated build_auto_door action into ActionExecutor - Updated PromptBuilder with redstone action documentation and example Materials required: 1 iron door, 2 pressure plates, 5 redstone dust
Added comprehensive quest and achievement system: - Quest.java: Quest definitions with objectives, rewards, and progress tracking - QuestManager.java: Manages active/completed quests with 8 default quests (mining, crafting, combat, building, farming, exploration, redstone) - Achievement.java: Achievement definitions with categories and rarity levels - AchievementManager.java: Tracks 25+ achievements across 8 categories (mining, crafting, combat, building, farming, exploration, redstone, general) - Integrated quest and achievement managers into SteveEntity Features: - Quest objectives track specific actions (mine, craft, kill, build, etc.) - Automatic progress tracking and completion detection - Achievement unlocking based on stats and milestones - JSON persistence for quest/achievement data - Rarity system: Common → Uncommon → Rare → Epic → Legendary - Category-based organization for both quests and achievements
Fixed multiple critical issues reported during testing: 1. **K key GUI bug**: Fixed K key closing GUI while typing - SteveGUI.handleKeyPress now returns false for character keys - Added isInputFocused() method to check input box state - SteveOverlayScreen only closes on K when input not focused - Allows typing 'k' in commands without closing GUI 2. **Block name mapping**: Added wood/tree block mappings - Added "wood" -> "oak_log" mapping - Added "tree" -> "oak_log" mapping - Added all wood types (birch, spruce, jungle, acacia, dark_oak) - Added common blocks (stone, dirt, sand, gravel) - Fixes "Invalid block type: wood" errors 3. **Auto-spawn disabled**: Removed automatic Steve spawning - Disabled auto-spawn of 4 Steves on player login - Players now manually spawn Steves via K GUI or /steve command - Keeps cleanup of existing Steves from NBT 4. **Client event handling**: Improved K key toggle logic - Only toggles when GUI is closed or screen is null - Prevents accidental closes during typing These fixes improve user experience and resolve confusion with automatic spawning.
CRITICAL FIX: Gemini was confusing "wood" with "coal" in every request! Changes to PromptBuilder.java: 1. Added CRITICAL MINING RULES section with explicit wood/tree examples - "mine wood" → use "wood" block - "chop wood" → use "wood" block - "get trees" → use "wood" block - "cut trees" → use "wood" block - "gather logs" → use "wood" block 2. Added 4 new examples (12-15) specifically for wood/tree commands - Shows exact JSON format for each variation - Emphasizes wood is NEVER coal 3. Enhanced mine action documentation - Lists wood/trees separately from ores - Explicit warning: "NEVER use coal for trees!" 4. Added Rule 9: "WOOD vs COAL: wood = tree logs, coal = black ore" - Makes distinction crystal clear 5. Added mistake warnings: - "Don't confuse wood with coal" - "Don't say 'Cannot fulfill' for wood - always available" This should fix the issue where AI consistently returned: - Input: "mine wood" → Output: "Mine coal" ❌ - Input: "chop wood" → Output: "Cannot fulfill" ❌ Now AI will correctly understand wood/tree requests.
CRITICAL FIX: Steve now properly mines wood from trees instead of standing still or mining coal. Changes to MineBlockAction.java: - Added wood detection in onStart() to identify wood/log blocks - Implemented dual search strategy in findNextBlock(): * Wood blocks: 30-block radius, searches Y from -5 to +20 (looks UPWARD for trees!) * Ore blocks: Original tunnel-ahead search (underground mining) - Modified mineNearbyBlock() to NOT tunnel when searching for wood - Wood blocks are now properly recognized and searched on surface This fixes the critical issue where "mine wood" commands resulted in: ❌ Steve standing still doing nothing ❌ AI confusing wood with coal ❌ "Cannot fulfill mining request" errors Now "mine wood" will actually work! 🌳
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.