hindsight-api 0.3.0__py3-none-any.whl → 0.4.1__py3-none-any.whl

hindsight_api/api/mcp.py

@@ -1,4 +1,4 @@
-"""Hindsight MCP Server implementation using FastMCP."""
+"""Hindsight MCP Server implementation using FastMCP (HTTP transport)."""
 
 import json
 import logging
@@ -8,8 +8,7 @@ from contextvars import ContextVar
 from fastmcp import FastMCP
 
 from hindsight_api import MemoryEngine
-from hindsight_api.engine.response_models import VALID_RECALL_FACT_TYPES
-from hindsight_api.models import RequestContext
+from hindsight_api.mcp_tools import MCPToolsConfig, register_mcp_tools
 
 # Configure logging from HINDSIGHT_API_LOG_LEVEL environment variable
 _log_level_str = os.environ.get("HINDSIGHT_API_LOG_LEVEL", "info").lower()
@@ -52,194 +51,15 @@ def create_mcp_server(memory: MemoryEngine) -> FastMCP:
     # Use stateless_http=True for Claude Code compatibility
     mcp = FastMCP("hindsight-mcp-server", stateless_http=True)
 
-    @mcp.tool()
-    async def retain(
-        content: str,
-        context: str = "general",
-        async_processing: bool = True,
-        bank_id: str | None = None,
-    ) -> str:
-        """
-        Store important information to long-term memory.
-
-        Use this tool PROACTIVELY whenever the user shares:
-        - Personal facts, preferences, or interests
-        - Important events or milestones
-        - User history, experiences, or background
-        - Decisions, opinions, or stated preferences
-        - Goals, plans, or future intentions
-        - Relationships or people mentioned
-        - Work context, projects, or responsibilities
-
-        Args:
-            content: The fact/memory to store (be specific and include relevant details)
-            context: Category for the memory (e.g., 'preferences', 'work', 'hobbies', 'family'). Default: 'general'
-            async_processing: If True, queue for background processing and return immediately. If False, wait for completion. Default: True
-            bank_id: Optional bank to store in (defaults to session bank). Use for cross-bank operations.
-        """
-        try:
-            target_bank = bank_id or get_current_bank_id()
-            if target_bank is None:
-                return "Error: No bank_id configured"
-            contents = [{"content": content, "context": context}]
-            if async_processing:
-                # Queue for background processing and return immediately
-                result = await memory.submit_async_retain(
-                    bank_id=target_bank, contents=contents, request_context=RequestContext()
-                )
-                return f"Memory queued for background processing (operation_id: {result.get('operation_id', 'N/A')})"
-            else:
-                # Wait for completion
-                await memory.retain_batch_async(
-                    bank_id=target_bank,
-                    contents=contents,
-                    request_context=RequestContext(),
-                )
-                return f"Memory stored successfully in bank '{target_bank}'"
-        except Exception as e:
-            logger.error(f"Error storing memory: {e}", exc_info=True)
-            return f"Error: {str(e)}"
-
-    @mcp.tool()
-    async def recall(query: str, max_tokens: int = 4096, bank_id: str | None = None) -> str:
-        """
-        Search memories to provide personalized, context-aware responses.
-
-        Use this tool PROACTIVELY to:
-        - Check user's preferences before making suggestions
-        - Recall user's history to provide continuity
-        - Remember user's goals and context
-        - Personalize responses based on past interactions
-
-        Args:
-            query: Natural language search query (e.g., "user's food preferences", "what projects is user working on")
-            max_tokens: Maximum tokens in the response (default: 4096)
-            bank_id: Optional bank to search in (defaults to session bank). Use for cross-bank operations.
-        """
-        try:
-            target_bank = bank_id or get_current_bank_id()
-            if target_bank is None:
-                return "Error: No bank_id configured"
-            from hindsight_api.engine.memory_engine import Budget
-
-            recall_result = await memory.recall_async(
-                bank_id=target_bank,
-                query=query,
-                fact_type=list(VALID_RECALL_FACT_TYPES),
-                budget=Budget.HIGH,
-                max_tokens=max_tokens,
-                request_context=RequestContext(),
-            )
-
-            # Use model's JSON serialization
-            return recall_result.model_dump_json(indent=2)
-        except Exception as e:
-            logger.error(f"Error searching: {e}", exc_info=True)
-            return f'{{"error": "{e}", "results": []}}'
-
-    @mcp.tool()
-    async def reflect(query: str, context: str | None = None, budget: str = "low", bank_id: str | None = None) -> str:
-        """
-        Generate thoughtful analysis by synthesizing stored memories with the bank's personality.
-
-        WHEN TO USE THIS TOOL:
-        Use reflect when you need reasoned analysis, not just fact retrieval. This tool
-        thinks through the question using everything the bank knows and its personality traits.
-
-        EXAMPLES OF GOOD QUERIES:
-        - "What patterns have emerged in how I approach debugging?"
-        - "Based on my past decisions, what architectural style do I prefer?"
-        - "What might be the best approach for this problem given what you know about me?"
-        - "How should I prioritize these tasks based on my goals?"
-
-        HOW IT DIFFERS FROM RECALL:
-        - recall: Returns raw facts matching your search (fast lookup)
-        - reflect: Reasons across memories to form a synthesized answer (deeper analysis)
-
-        Use recall for "what did I say about X?" and reflect for "what should I do about X?"
-
-        Args:
-            query: The question or topic to reflect on
-            context: Optional context about why this reflection is needed
-            budget: Search budget - 'low', 'mid', or 'high' (default: 'low')
-            bank_id: Optional bank to reflect in (defaults to session bank). Use for cross-bank operations.
-        """
-        try:
-            target_bank = bank_id or get_current_bank_id()
-            if target_bank is None:
-                return "Error: No bank_id configured"
-            from hindsight_api.engine.memory_engine import Budget
-
-            # Map string budget to enum
-            budget_map = {"low": Budget.LOW, "mid": Budget.MID, "high": Budget.HIGH}
-            budget_enum = budget_map.get(budget.lower(), Budget.LOW)
-
-            reflect_result = await memory.reflect_async(
-                bank_id=target_bank,
-                query=query,
-                budget=budget_enum,
-                context=context,
-                request_context=RequestContext(),
-            )
-
-            return reflect_result.model_dump_json(indent=2)
-        except Exception as e:
-            logger.error(f"Error reflecting: {e}", exc_info=True)
-            return f'{{"error": "{e}", "text": ""}}'
-
-    @mcp.tool()
-    async def list_banks() -> str:
-        """
-        List all available memory banks.
-
-        Use this tool to discover what memory banks exist in the system.
-        Each bank is an isolated memory store (like a separate "brain").
-
-        Returns:
-            JSON list of banks with their IDs, names, dispositions, and backgrounds.
-        """
-        try:
-            banks = await memory.list_banks(request_context=RequestContext())
-            return json.dumps({"banks": banks}, indent=2)
-        except Exception as e:
-            logger.error(f"Error listing banks: {e}", exc_info=True)
-            return f'{{"error": "{e}", "banks": []}}'
-
-    @mcp.tool()
-    async def create_bank(bank_id: str, name: str | None = None, background: str | None = None) -> str:
-        """
-        Create a new memory bank or get an existing one.
-
-        Memory banks are isolated stores - each one is like a separate "brain" for a user/agent.
-        Banks are auto-created with default settings if they don't exist.
-
-        Args:
-            bank_id: Unique identifier for the bank (e.g., 'user-123', 'agent-alpha')
-            name: Optional human-friendly name for the bank
-            background: Optional background context about the bank's owner/purpose
-        """
-        try:
-            # get_bank_profile auto-creates bank if it doesn't exist
-            profile = await memory.get_bank_profile(bank_id, request_context=RequestContext())
-
-            # Update name/background if provided
-            if name is not None or background is not None:
-                await memory.update_bank(
-                    bank_id,
-                    name=name,
-                    background=background,
-                    request_context=RequestContext(),
-                )
-                # Fetch updated profile
-                profile = await memory.get_bank_profile(bank_id, request_context=RequestContext())
-
-            # Serialize disposition if it's a Pydantic model
-            if "disposition" in profile and hasattr(profile["disposition"], "model_dump"):
-                profile["disposition"] = profile["disposition"].model_dump()
-            return json.dumps(profile, indent=2)
-        except Exception as e:
-            logger.error(f"Error creating bank: {e}", exc_info=True)
-            return f'{{"error": "{e}"}}'
+    # Configure and register tools using shared module
+    config = MCPToolsConfig(
+        bank_id_resolver=get_current_bank_id,
+        include_bank_id_param=True,  # HTTP MCP supports multi-bank via parameter
+        tools=None,  # All tools
+        retain_fire_and_forget=False,  # HTTP MCP supports sync/async modes
+    )
+
+    register_mcp_tools(mcp, memory, config)
 
     return mcp
 

hindsight_api/config.py

@@ -4,9 +4,12 @@ Centralized configuration for Hindsight API.
 All environment variables and their defaults are defined here.
 """
 
+import json
 import logging
 import os
+import sys
 from dataclasses import dataclass
+from datetime import datetime, timezone
 
 from dotenv import find_dotenv, load_dotenv
 
@@ -17,6 +20,7 @@ logger = logging.getLogger(__name__)
 
 # Environment variable names
 ENV_DATABASE_URL = "HINDSIGHT_API_DATABASE_URL"
+ENV_DATABASE_SCHEMA = "HINDSIGHT_API_DATABASE_SCHEMA"
 ENV_LLM_PROVIDER = "HINDSIGHT_API_LLM_PROVIDER"
 ENV_LLM_API_KEY = "HINDSIGHT_API_LLM_API_KEY"
 ENV_LLM_MODEL = "HINDSIGHT_API_LLM_MODEL"
@@ -36,8 +40,14 @@ ENV_REFLECT_LLM_API_KEY = "HINDSIGHT_API_REFLECT_LLM_API_KEY"
 ENV_REFLECT_LLM_MODEL = "HINDSIGHT_API_REFLECT_LLM_MODEL"
 ENV_REFLECT_LLM_BASE_URL = "HINDSIGHT_API_REFLECT_LLM_BASE_URL"
 
+ENV_CONSOLIDATION_LLM_PROVIDER = "HINDSIGHT_API_CONSOLIDATION_LLM_PROVIDER"
+ENV_CONSOLIDATION_LLM_API_KEY = "HINDSIGHT_API_CONSOLIDATION_LLM_API_KEY"
+ENV_CONSOLIDATION_LLM_MODEL = "HINDSIGHT_API_CONSOLIDATION_LLM_MODEL"
+ENV_CONSOLIDATION_LLM_BASE_URL = "HINDSIGHT_API_CONSOLIDATION_LLM_BASE_URL"
+
 ENV_EMBEDDINGS_PROVIDER = "HINDSIGHT_API_EMBEDDINGS_PROVIDER"
 ENV_EMBEDDINGS_LOCAL_MODEL = "HINDSIGHT_API_EMBEDDINGS_LOCAL_MODEL"
+ENV_EMBEDDINGS_LOCAL_FORCE_CPU = "HINDSIGHT_API_EMBEDDINGS_LOCAL_FORCE_CPU"
 ENV_EMBEDDINGS_TEI_URL = "HINDSIGHT_API_EMBEDDINGS_TEI_URL"
 ENV_EMBEDDINGS_OPENAI_API_KEY = "HINDSIGHT_API_EMBEDDINGS_OPENAI_API_KEY"
 ENV_EMBEDDINGS_OPENAI_MODEL = "HINDSIGHT_API_EMBEDDINGS_OPENAI_MODEL"
@@ -57,6 +67,7 @@ ENV_RERANKER_LITELLM_MODEL = "HINDSIGHT_API_RERANKER_LITELLM_MODEL"
 
 ENV_RERANKER_PROVIDER = "HINDSIGHT_API_RERANKER_PROVIDER"
 ENV_RERANKER_LOCAL_MODEL = "HINDSIGHT_API_RERANKER_LOCAL_MODEL"
+ENV_RERANKER_LOCAL_FORCE_CPU = "HINDSIGHT_API_RERANKER_LOCAL_FORCE_CPU"
 ENV_RERANKER_LOCAL_MAX_CONCURRENT = "HINDSIGHT_API_RERANKER_LOCAL_MAX_CONCURRENT"
 ENV_RERANKER_TEI_URL = "HINDSIGHT_API_RERANKER_TEI_URL"
 ENV_RERANKER_TEI_BATCH_SIZE = "HINDSIGHT_API_RERANKER_TEI_BATCH_SIZE"
@@ -68,6 +79,7 @@ ENV_RERANKER_FLASHRANK_CACHE_DIR = "HINDSIGHT_API_RERANKER_FLASHRANK_CACHE_DIR"
 ENV_HOST = "HINDSIGHT_API_HOST"
 ENV_PORT = "HINDSIGHT_API_PORT"
 ENV_LOG_LEVEL = "HINDSIGHT_API_LOG_LEVEL"
+ENV_LOG_FORMAT = "HINDSIGHT_API_LOG_FORMAT"
 ENV_WORKERS = "HINDSIGHT_API_WORKERS"
 ENV_MCP_ENABLED = "HINDSIGHT_API_MCP_ENABLED"
 ENV_GRAPH_RETRIEVER = "HINDSIGHT_API_GRAPH_RETRIEVER"
@@ -76,18 +88,21 @@ ENV_RECALL_MAX_CONCURRENT = "HINDSIGHT_API_RECALL_MAX_CONCURRENT"
 ENV_RECALL_CONNECTION_BUDGET = "HINDSIGHT_API_RECALL_CONNECTION_BUDGET"
 ENV_MCP_LOCAL_BANK_ID = "HINDSIGHT_API_MCP_LOCAL_BANK_ID"
 ENV_MCP_INSTRUCTIONS = "HINDSIGHT_API_MCP_INSTRUCTIONS"
-
-# Observation thresholds
-ENV_OBSERVATION_MIN_FACTS = "HINDSIGHT_API_OBSERVATION_MIN_FACTS"
-ENV_OBSERVATION_TOP_ENTITIES = "HINDSIGHT_API_OBSERVATION_TOP_ENTITIES"
+ENV_MENTAL_MODEL_REFRESH_CONCURRENCY = "HINDSIGHT_API_MENTAL_MODEL_REFRESH_CONCURRENCY"
 
 # Retain settings
 ENV_RETAIN_MAX_COMPLETION_TOKENS = "HINDSIGHT_API_RETAIN_MAX_COMPLETION_TOKENS"
 ENV_RETAIN_CHUNK_SIZE = "HINDSIGHT_API_RETAIN_CHUNK_SIZE"
 ENV_RETAIN_EXTRACT_CAUSAL_LINKS = "HINDSIGHT_API_RETAIN_EXTRACT_CAUSAL_LINKS"
 ENV_RETAIN_EXTRACTION_MODE = "HINDSIGHT_API_RETAIN_EXTRACTION_MODE"
+ENV_RETAIN_CUSTOM_INSTRUCTIONS = "HINDSIGHT_API_RETAIN_CUSTOM_INSTRUCTIONS"
 ENV_RETAIN_OBSERVATIONS_ASYNC = "HINDSIGHT_API_RETAIN_OBSERVATIONS_ASYNC"
 
+# Observations settings (consolidated knowledge from facts)
+ENV_ENABLE_OBSERVATIONS = "HINDSIGHT_API_ENABLE_OBSERVATIONS"
+ENV_CONSOLIDATION_BATCH_SIZE = "HINDSIGHT_API_CONSOLIDATION_BATCH_SIZE"
+ENV_CONSOLIDATION_MAX_TOKENS = "HINDSIGHT_API_CONSOLIDATION_MAX_TOKENS"
+
 # Optimization flags
 ENV_SKIP_LLM_VERIFICATION = "HINDSIGHT_API_SKIP_LLM_VERIFICATION"
 ENV_LAZY_RERANKER = "HINDSIGHT_API_LAZY_RERANKER"
@@ -101,13 +116,20 @@ ENV_DB_POOL_MAX_SIZE = "HINDSIGHT_API_DB_POOL_MAX_SIZE"
 ENV_DB_COMMAND_TIMEOUT = "HINDSIGHT_API_DB_COMMAND_TIMEOUT"
 ENV_DB_ACQUIRE_TIMEOUT = "HINDSIGHT_API_DB_ACQUIRE_TIMEOUT"
 
-# Background task processing
-ENV_TASK_BACKEND = "HINDSIGHT_API_TASK_BACKEND"
-ENV_TASK_BACKEND_MEMORY_BATCH_SIZE = "HINDSIGHT_API_TASK_BACKEND_MEMORY_BATCH_SIZE"
-ENV_TASK_BACKEND_MEMORY_BATCH_INTERVAL = "HINDSIGHT_API_TASK_BACKEND_MEMORY_BATCH_INTERVAL"
+# Worker configuration (distributed task processing)
+ENV_WORKER_ENABLED = "HINDSIGHT_API_WORKER_ENABLED"
+ENV_WORKER_ID = "HINDSIGHT_API_WORKER_ID"
+ENV_WORKER_POLL_INTERVAL_MS = "HINDSIGHT_API_WORKER_POLL_INTERVAL_MS"
+ENV_WORKER_MAX_RETRIES = "HINDSIGHT_API_WORKER_MAX_RETRIES"
+ENV_WORKER_BATCH_SIZE = "HINDSIGHT_API_WORKER_BATCH_SIZE"
+ENV_WORKER_HTTP_PORT = "HINDSIGHT_API_WORKER_HTTP_PORT"
+
+# Reflect agent settings
+ENV_REFLECT_MAX_ITERATIONS = "HINDSIGHT_API_REFLECT_MAX_ITERATIONS"
 
 # Default values
 DEFAULT_DATABASE_URL = "pg0"
+DEFAULT_DATABASE_SCHEMA = "public"
 DEFAULT_LLM_PROVIDER = "openai"
 DEFAULT_LLM_MODEL = "gpt-5-mini"
 DEFAULT_LLM_MAX_CONCURRENT = 32
@@ -115,11 +137,13 @@ DEFAULT_LLM_TIMEOUT = 120.0  # seconds
 
 DEFAULT_EMBEDDINGS_PROVIDER = "local"
 DEFAULT_EMBEDDINGS_LOCAL_MODEL = "BAAI/bge-small-en-v1.5"
+DEFAULT_EMBEDDINGS_LOCAL_FORCE_CPU = False  # Force CPU mode for local embeddings (avoids MPS/XPC issues on macOS)
 DEFAULT_EMBEDDINGS_OPENAI_MODEL = "text-embedding-3-small"
 DEFAULT_EMBEDDING_DIMENSION = 384
 
 DEFAULT_RERANKER_PROVIDER = "local"
 DEFAULT_RERANKER_LOCAL_MODEL = "cross-encoder/ms-marco-MiniLM-L-6-v2"
+DEFAULT_RERANKER_LOCAL_FORCE_CPU = False  # Force CPU mode for local reranker (avoids MPS/XPC issues on macOS)
 DEFAULT_RERANKER_LOCAL_MAX_CONCURRENT = 4  # Limit concurrent CPU-bound reranking to prevent thrashing
 DEFAULT_RERANKER_TEI_BATCH_SIZE = 128
 DEFAULT_RERANKER_TEI_MAX_CONCURRENT = 8
@@ -138,6 +162,7 @@ DEFAULT_RERANKER_LITELLM_MODEL = "cohere/rerank-english-v3.0"
 DEFAULT_HOST = "0.0.0.0"
 DEFAULT_PORT = 8888
 DEFAULT_LOG_LEVEL = "info"
+DEFAULT_LOG_FORMAT = "text"  # Options: "text", "json"
 DEFAULT_WORKERS = 1
 DEFAULT_MCP_ENABLED = True
 DEFAULT_GRAPH_RETRIEVER = "link_expansion"  # Options: "link_expansion", "mpfp", "bfs"
@@ -145,19 +170,22 @@ DEFAULT_MPFP_TOP_K_NEIGHBORS = 20  # Fan-out limit per node in MPFP graph traver
 DEFAULT_RECALL_MAX_CONCURRENT = 32  # Max concurrent recall operations per worker
 DEFAULT_RECALL_CONNECTION_BUDGET = 4  # Max concurrent DB connections per recall operation
 DEFAULT_MCP_LOCAL_BANK_ID = "mcp"
-
-# Observation thresholds
-DEFAULT_OBSERVATION_MIN_FACTS = 5  # Min facts required to generate entity observations
-DEFAULT_OBSERVATION_TOP_ENTITIES = 5  # Max entities to process per retain batch
+DEFAULT_MENTAL_MODEL_REFRESH_CONCURRENCY = 8  # Max concurrent mental model refreshes
 
 # Retain settings
 DEFAULT_RETAIN_MAX_COMPLETION_TOKENS = 64000  # Max tokens for fact extraction LLM call
 DEFAULT_RETAIN_CHUNK_SIZE = 3000  # Max chars per chunk for fact extraction
 DEFAULT_RETAIN_EXTRACT_CAUSAL_LINKS = True  # Extract causal links between facts
-DEFAULT_RETAIN_EXTRACTION_MODE = "concise"  # Extraction mode: "concise" or "verbose"
-RETAIN_EXTRACTION_MODES = ("concise", "verbose")  # Allowed extraction modes
+DEFAULT_RETAIN_EXTRACTION_MODE = "concise"  # Extraction mode: "concise", "verbose", or "custom"
+RETAIN_EXTRACTION_MODES = ("concise", "verbose", "custom")  # Allowed extraction modes
+DEFAULT_RETAIN_CUSTOM_INSTRUCTIONS = None  # Custom extraction guidelines (only used when mode="custom")
 DEFAULT_RETAIN_OBSERVATIONS_ASYNC = False  # Run observation generation async (after retain completes)
 
+# Observations defaults (consolidated knowledge from facts)
+DEFAULT_ENABLE_OBSERVATIONS = True  # Observations enabled by default
+DEFAULT_CONSOLIDATION_BATCH_SIZE = 50  # Memories to load per batch (internal memory optimization)
+DEFAULT_CONSOLIDATION_MAX_TOKENS = 1024  # Max tokens for recall when finding related observations
+
 # Database migrations
 DEFAULT_RUN_MIGRATIONS_ON_STARTUP = True
 
@@ -167,10 +195,16 @@ DEFAULT_DB_POOL_MAX_SIZE = 100
 DEFAULT_DB_COMMAND_TIMEOUT = 60  # seconds
 DEFAULT_DB_ACQUIRE_TIMEOUT = 30  # seconds
 
-# Background task processing
-DEFAULT_TASK_BACKEND = "memory"  # Options: "memory", "noop"
-DEFAULT_TASK_BACKEND_MEMORY_BATCH_SIZE = 10
-DEFAULT_TASK_BACKEND_MEMORY_BATCH_INTERVAL = 1.0  # seconds
+# Worker configuration (distributed task processing)
+DEFAULT_WORKER_ENABLED = True  # API runs worker by default (standalone mode)
+DEFAULT_WORKER_ID = None  # Will use hostname if not specified
+DEFAULT_WORKER_POLL_INTERVAL_MS = 500  # Poll database every 500ms
+DEFAULT_WORKER_MAX_RETRIES = 3  # Max retries before marking task failed
+DEFAULT_WORKER_BATCH_SIZE = 10  # Tasks to claim per poll cycle
+DEFAULT_WORKER_HTTP_PORT = 8889  # HTTP port for worker metrics/health
+
+# Reflect agent settings
+DEFAULT_REFLECT_MAX_ITERATIONS = 10  # Max tool call iterations before forcing response
 
 # Default MCP tool descriptions (can be customized via env vars)
 DEFAULT_MCP_RETAIN_DESCRIPTION = """Store important information to long-term memory.
@@ -196,6 +230,36 @@ Use this tool PROACTIVELY to:
 EMBEDDING_DIMENSION = DEFAULT_EMBEDDING_DIMENSION
 
 
+class JsonFormatter(logging.Formatter):
+    """JSON formatter for structured logging.
+
+    Outputs logs in JSON format with a 'severity' field that cloud logging
+    systems (GCP, AWS CloudWatch, etc.) can parse to correctly categorize log levels.
+    """
+
+    SEVERITY_MAP = {
+        logging.DEBUG: "DEBUG",
+        logging.INFO: "INFO",
+        logging.WARNING: "WARNING",
+        logging.ERROR: "ERROR",
+        logging.CRITICAL: "CRITICAL",
+    }
+
+    def format(self, record: logging.LogRecord) -> str:
+        log_entry = {
+            "severity": self.SEVERITY_MAP.get(record.levelno, "DEFAULT"),
+            "message": record.getMessage(),
+            "timestamp": datetime.now(timezone.utc).isoformat(),
+            "logger": record.name,
+        }
+
+        # Add exception info if present
+        if record.exc_info:
+            log_entry["exception"] = self.formatException(record.exc_info)
+
+        return json.dumps(log_entry)
+
+
 def _validate_extraction_mode(mode: str) -> str:
     """Validate and normalize extraction mode."""
     mode_lower = mode.lower()
@@ -214,6 +278,7 @@ class HindsightConfig:
 
     # Database
     database_url: str
+    database_schema: str
 
     # LLM (default, used as fallback for per-operation config)
     llm_provider: str
@@ -234,9 +299,15 @@ class HindsightConfig:
     reflect_llm_model: str | None
     reflect_llm_base_url: str | None
 
+    consolidation_llm_provider: str | None
+    consolidation_llm_api_key: str | None
+    consolidation_llm_model: str | None
+    consolidation_llm_base_url: str | None
+
     # Embeddings
     embeddings_provider: str
     embeddings_local_model: str
+    embeddings_local_force_cpu: bool
     embeddings_tei_url: str | None
     embeddings_openai_base_url: str | None
     embeddings_cohere_base_url: str | None
@@ -244,6 +315,8 @@ class HindsightConfig:
     # Reranker
     reranker_provider: str
     reranker_local_model: str
+    reranker_local_force_cpu: bool
+    reranker_local_max_concurrent: int
     reranker_tei_url: str | None
     reranker_tei_batch_size: int
     reranker_tei_max_concurrent: int
@@ -254,6 +327,7 @@ class HindsightConfig:
     host: str
     port: int
     log_level: str
+    log_format: str
     mcp_enabled: bool
 
     # Recall
@@ -261,18 +335,21 @@ class HindsightConfig:
     mpfp_top_k_neighbors: int
     recall_max_concurrent: int
     recall_connection_budget: int
-
-    # Observation thresholds
-    observation_min_facts: int
-    observation_top_entities: int
+    mental_model_refresh_concurrency: int
 
     # Retain settings
     retain_max_completion_tokens: int
     retain_chunk_size: int
     retain_extract_causal_links: bool
     retain_extraction_mode: str
+    retain_custom_instructions: str | None
     retain_observations_async: bool
 
+    # Observations settings (consolidated knowledge from facts)
+    enable_observations: bool
+    consolidation_batch_size: int
+    consolidation_max_tokens: int
+
     # Optimization flags
     skip_llm_verification: bool
     lazy_reranker: bool
@@ -286,10 +363,16 @@ class HindsightConfig:
     db_command_timeout: int
     db_acquire_timeout: int
 
-    # Background task processing
-    task_backend: str
-    task_backend_memory_batch_size: int
-    task_backend_memory_batch_interval: float
+    # Worker configuration (distributed task processing)
+    worker_enabled: bool
+    worker_id: str | None
+    worker_poll_interval_ms: int
+    worker_max_retries: int
+    worker_batch_size: int
+    worker_http_port: int
+
+    # Reflect agent settings
+    reflect_max_iterations: int
 
     @classmethod
     def from_env(cls) -> "HindsightConfig":
@@ -297,6 +380,7 @@ class HindsightConfig:
         return cls(
             # Database
             database_url=os.getenv(ENV_DATABASE_URL, DEFAULT_DATABASE_URL),
+            database_schema=os.getenv(ENV_DATABASE_SCHEMA, DEFAULT_DATABASE_SCHEMA),
             # LLM
             llm_provider=os.getenv(ENV_LLM_PROVIDER, DEFAULT_LLM_PROVIDER),
             llm_api_key=os.getenv(ENV_LLM_API_KEY),
@@ -313,15 +397,30 @@ class HindsightConfig:
             reflect_llm_api_key=os.getenv(ENV_REFLECT_LLM_API_KEY) or None,
             reflect_llm_model=os.getenv(ENV_REFLECT_LLM_MODEL) or None,
             reflect_llm_base_url=os.getenv(ENV_REFLECT_LLM_BASE_URL) or None,
+            consolidation_llm_provider=os.getenv(ENV_CONSOLIDATION_LLM_PROVIDER) or None,
+            consolidation_llm_api_key=os.getenv(ENV_CONSOLIDATION_LLM_API_KEY) or None,
+            consolidation_llm_model=os.getenv(ENV_CONSOLIDATION_LLM_MODEL) or None,
+            consolidation_llm_base_url=os.getenv(ENV_CONSOLIDATION_LLM_BASE_URL) or None,
             # Embeddings
             embeddings_provider=os.getenv(ENV_EMBEDDINGS_PROVIDER, DEFAULT_EMBEDDINGS_PROVIDER),
             embeddings_local_model=os.getenv(ENV_EMBEDDINGS_LOCAL_MODEL, DEFAULT_EMBEDDINGS_LOCAL_MODEL),
+            embeddings_local_force_cpu=os.getenv(
+                ENV_EMBEDDINGS_LOCAL_FORCE_CPU, str(DEFAULT_EMBEDDINGS_LOCAL_FORCE_CPU)
+            ).lower()
+            in ("true", "1"),
             embeddings_tei_url=os.getenv(ENV_EMBEDDINGS_TEI_URL),
             embeddings_openai_base_url=os.getenv(ENV_EMBEDDINGS_OPENAI_BASE_URL) or None,
             embeddings_cohere_base_url=os.getenv(ENV_EMBEDDINGS_COHERE_BASE_URL) or None,
             # Reranker
             reranker_provider=os.getenv(ENV_RERANKER_PROVIDER, DEFAULT_RERANKER_PROVIDER),
             reranker_local_model=os.getenv(ENV_RERANKER_LOCAL_MODEL, DEFAULT_RERANKER_LOCAL_MODEL),
+            reranker_local_force_cpu=os.getenv(
+                ENV_RERANKER_LOCAL_FORCE_CPU, str(DEFAULT_RERANKER_LOCAL_FORCE_CPU)
+            ).lower()
+            in ("true", "1"),
+            reranker_local_max_concurrent=int(
+                os.getenv(ENV_RERANKER_LOCAL_MAX_CONCURRENT, str(DEFAULT_RERANKER_LOCAL_MAX_CONCURRENT))
+            ),
             reranker_tei_url=os.getenv(ENV_RERANKER_TEI_URL),
             reranker_tei_batch_size=int(os.getenv(ENV_RERANKER_TEI_BATCH_SIZE, str(DEFAULT_RERANKER_TEI_BATCH_SIZE))),
             reranker_tei_max_concurrent=int(
@@ -333,6 +432,7 @@ class HindsightConfig:
             host=os.getenv(ENV_HOST, DEFAULT_HOST),
             port=int(os.getenv(ENV_PORT, DEFAULT_PORT)),
             log_level=os.getenv(ENV_LOG_LEVEL, DEFAULT_LOG_LEVEL),
+            log_format=os.getenv(ENV_LOG_FORMAT, DEFAULT_LOG_FORMAT).lower(),
             mcp_enabled=os.getenv(ENV_MCP_ENABLED, str(DEFAULT_MCP_ENABLED)).lower() == "true",
             # Recall
             graph_retriever=os.getenv(ENV_GRAPH_RETRIEVER, DEFAULT_GRAPH_RETRIEVER),
@@ -341,14 +441,12 @@ class HindsightConfig:
             recall_connection_budget=int(
                 os.getenv(ENV_RECALL_CONNECTION_BUDGET, str(DEFAULT_RECALL_CONNECTION_BUDGET))
             ),
+            mental_model_refresh_concurrency=int(
+                os.getenv(ENV_MENTAL_MODEL_REFRESH_CONCURRENCY, str(DEFAULT_MENTAL_MODEL_REFRESH_CONCURRENCY))
+            ),
             # Optimization flags
             skip_llm_verification=os.getenv(ENV_SKIP_LLM_VERIFICATION, "false").lower() == "true",
             lazy_reranker=os.getenv(ENV_LAZY_RERANKER, "false").lower() == "true",
-            # Observation thresholds
-            observation_min_facts=int(os.getenv(ENV_OBSERVATION_MIN_FACTS, str(DEFAULT_OBSERVATION_MIN_FACTS))),
-            observation_top_entities=int(
-                os.getenv(ENV_OBSERVATION_TOP_ENTITIES, str(DEFAULT_OBSERVATION_TOP_ENTITIES))
-            ),
             # Retain settings
             retain_max_completion_tokens=int(
                 os.getenv(ENV_RETAIN_MAX_COMPLETION_TOKENS, str(DEFAULT_RETAIN_MAX_COMPLETION_TOKENS))
@@ -361,10 +459,19 @@ class HindsightConfig:
             retain_extraction_mode=_validate_extraction_mode(
                 os.getenv(ENV_RETAIN_EXTRACTION_MODE, DEFAULT_RETAIN_EXTRACTION_MODE)
             ),
+            retain_custom_instructions=os.getenv(ENV_RETAIN_CUSTOM_INSTRUCTIONS) or DEFAULT_RETAIN_CUSTOM_INSTRUCTIONS,
             retain_observations_async=os.getenv(
                 ENV_RETAIN_OBSERVATIONS_ASYNC, str(DEFAULT_RETAIN_OBSERVATIONS_ASYNC)
             ).lower()
             == "true",
+            # Observations settings (consolidated knowledge from facts)
+            enable_observations=os.getenv(ENV_ENABLE_OBSERVATIONS, str(DEFAULT_ENABLE_OBSERVATIONS)).lower() == "true",
+            consolidation_batch_size=int(
+                os.getenv(ENV_CONSOLIDATION_BATCH_SIZE, str(DEFAULT_CONSOLIDATION_BATCH_SIZE))
+            ),
+            consolidation_max_tokens=int(
+                os.getenv(ENV_CONSOLIDATION_MAX_TOKENS, str(DEFAULT_CONSOLIDATION_MAX_TOKENS))
+            ),
             # Database migrations
             run_migrations_on_startup=os.getenv(ENV_RUN_MIGRATIONS_ON_STARTUP, "true").lower() == "true",
             # Database connection pool
@@ -372,14 +479,15 @@ class HindsightConfig:
             db_pool_max_size=int(os.getenv(ENV_DB_POOL_MAX_SIZE, str(DEFAULT_DB_POOL_MAX_SIZE))),
             db_command_timeout=int(os.getenv(ENV_DB_COMMAND_TIMEOUT, str(DEFAULT_DB_COMMAND_TIMEOUT))),
             db_acquire_timeout=int(os.getenv(ENV_DB_ACQUIRE_TIMEOUT, str(DEFAULT_DB_ACQUIRE_TIMEOUT))),
-            # Background task processing
-            task_backend=os.getenv(ENV_TASK_BACKEND, DEFAULT_TASK_BACKEND),
-            task_backend_memory_batch_size=int(
-                os.getenv(ENV_TASK_BACKEND_MEMORY_BATCH_SIZE, str(DEFAULT_TASK_BACKEND_MEMORY_BATCH_SIZE))
-            ),
-            task_backend_memory_batch_interval=float(
-                os.getenv(ENV_TASK_BACKEND_MEMORY_BATCH_INTERVAL, str(DEFAULT_TASK_BACKEND_MEMORY_BATCH_INTERVAL))
-            ),
+            # Worker configuration
+            worker_enabled=os.getenv(ENV_WORKER_ENABLED, str(DEFAULT_WORKER_ENABLED)).lower() == "true",
+            worker_id=os.getenv(ENV_WORKER_ID) or DEFAULT_WORKER_ID,
+            worker_poll_interval_ms=int(os.getenv(ENV_WORKER_POLL_INTERVAL_MS, str(DEFAULT_WORKER_POLL_INTERVAL_MS))),
+            worker_max_retries=int(os.getenv(ENV_WORKER_MAX_RETRIES, str(DEFAULT_WORKER_MAX_RETRIES))),
+            worker_batch_size=int(os.getenv(ENV_WORKER_BATCH_SIZE, str(DEFAULT_WORKER_BATCH_SIZE))),
+            worker_http_port=int(os.getenv(ENV_WORKER_HTTP_PORT, str(DEFAULT_WORKER_HTTP_PORT))),
+            # Reflect agent settings
+            reflect_max_iterations=int(os.getenv(ENV_REFLECT_MAX_ITERATIONS, str(DEFAULT_REFLECT_MAX_ITERATIONS))),
         )
 
     def get_llm_base_url(self) -> str:
@@ -410,16 +518,32 @@ class HindsightConfig:
         return log_level_map.get(self.log_level.lower(), logging.INFO)
 
     def configure_logging(self) -> None:
-        """Configure Python logging based on the log level."""
-        logging.basicConfig(
-            level=self.get_python_log_level(),
-            format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
-            force=True,  # Override any existing configuration
-        )
+        """Configure Python logging based on the log level and format.
+
+        When log_format is "json", outputs structured JSON logs with a severity
+        field that GCP Cloud Logging can parse for proper log level categorization.
+        """
+        root_logger = logging.getLogger()
+        root_logger.setLevel(self.get_python_log_level())
+
+        # Remove existing handlers
+        for handler in root_logger.handlers[:]:
+            root_logger.removeHandler(handler)
+
+        # Create handler writing to stdout (GCP treats stderr as ERROR)
+        handler = logging.StreamHandler(sys.stdout)
+        handler.setLevel(self.get_python_log_level())
+
+        if self.log_format == "json":
+            handler.setFormatter(JsonFormatter())
+        else:
+            handler.setFormatter(logging.Formatter("%(asctime)s - %(levelname)s - %(name)s - %(message)s"))
+
+        root_logger.addHandler(handler)
 
     def log_config(self) -> None:
         """Log the current configuration (without sensitive values)."""
-        logger.info(f"Database: {self.database_url}")
+        logger.info(f"Database: {self.database_url} (schema: {self.database_schema})")
         logger.info(f"LLM: provider={self.llm_provider}, model={self.llm_model}")
         if self.retain_llm_provider or self.retain_llm_model:
             retain_provider = self.retain_llm_provider or self.llm_provider
@@ -429,6 +553,10 @@ class HindsightConfig:
             reflect_provider = self.reflect_llm_provider or self.llm_provider
             reflect_model = self.reflect_llm_model or self.llm_model
             logger.info(f"LLM (reflect): provider={reflect_provider}, model={reflect_model}")
+        if self.consolidation_llm_provider or self.consolidation_llm_model:
+            consolidation_provider = self.consolidation_llm_provider or self.llm_provider
+            consolidation_model = self.consolidation_llm_model or self.llm_model
+            logger.info(f"LLM (consolidation): provider={consolidation_provider}, model={consolidation_model}")
         logger.info(f"Embeddings: provider={self.embeddings_provider}")
         logger.info(f"Reranker: provider={self.reranker_provider}")
         logger.info(f"Graph retriever: {self.graph_retriever}")