feat: batch observations consolidation

hindsight-api/hindsight_api/config.py

@@ -285,6 +285,7 @@ def normalize_config_dict(config: dict[str, Any]) -> dict[str, Any]:
 # Observations settings (consolidated knowledge from facts)
 ENV_ENABLE_OBSERVATIONS = "HINDSIGHT_API_ENABLE_OBSERVATIONS"
 ENV_CONSOLIDATION_BATCH_SIZE = "HINDSIGHT_API_CONSOLIDATION_BATCH_SIZE"
+ENV_CONSOLIDATION_LLM_BATCH_SIZE = "HINDSIGHT_API_CONSOLIDATION_LLM_BATCH_SIZE"
 ENV_CONSOLIDATION_MAX_TOKENS = "HINDSIGHT_API_CONSOLIDATION_MAX_TOKENS"
 ENV_OBSERVATIONS_MISSION = "HINDSIGHT_API_OBSERVATIONS_MISSION"
 
@@ -426,7 +427,8 @@ def normalize_config_dict(config: dict[str, Any]) -> dict[str, Any]:
 # 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
+DEFAULT_CONSOLIDATION_LLM_BATCH_SIZE = 8  # Facts per LLM call (1 = no batching; >1 = batch mode)
+DEFAULT_CONSOLIDATION_MAX_TOKENS = 512  # Max tokens for recall when finding related observations
 DEFAULT_OBSERVATIONS_MISSION = None  # Declarative spec of what observations are for this bank
 
 # Database migrations
@@ -679,6 +681,7 @@ class HindsightConfig:
     # Observations settings (consolidated knowledge from facts)
     enable_observations: bool
     consolidation_batch_size: int
+    consolidation_llm_batch_size: int
     consolidation_max_tokens: int
     observations_mission: str | None
 
@@ -1097,6 +1100,9 @@ def from_env(cls) -> "HindsightConfig":
             consolidation_batch_size=int(
                 os.getenv(ENV_CONSOLIDATION_BATCH_SIZE, str(DEFAULT_CONSOLIDATION_BATCH_SIZE))
             ),
+            consolidation_llm_batch_size=int(
+                os.getenv(ENV_CONSOLIDATION_LLM_BATCH_SIZE, str(DEFAULT_CONSOLIDATION_LLM_BATCH_SIZE))
+            ),
             consolidation_max_tokens=int(
                 os.getenv(ENV_CONSOLIDATION_MAX_TOKENS, str(DEFAULT_CONSOLIDATION_MAX_TOKENS))
             ),