feat(R4): make LLM rate limit configurable via LLM_MAX_CALLS_PER_MIN

README.md

@@ -163,6 +163,7 @@ All keys live in `.repi/config.json` (see `config.example.json` for the full sch
 | `UI_PORT` | `3000` | Port the web UI binds to (read by `repi ui`) |
 | `WATCHER_CONFIG_REFRESH_SECS` | `30` | How often the worker polls for config changes |
 | `OLLAMA_BASE_URL` | `http://localhost:11434` | Ollama endpoint |
+| `LLM_MAX_CALLS_PER_MIN` | `10` | Maximum LLM API calls per minute in the ReAct loop. Increase for high-tier API plans. |
 
 ## Development
 

repi/core/config.py

@@ -57,6 +57,7 @@ class Settings(BaseSettings):
     OLLAMA_BASE_URL: str = "http://localhost:11434"
 
     WATCHER_CONFIG_REFRESH_SECS: int = 30
+    LLM_MAX_CALLS_PER_MIN: int = Field(default=60, ge=1)
 
     # "fastembed" (ONNX Runtime, ~50 MB) or "torch" via sentence-transformers
     # (~790 MB). Vectors are byte-identical; the choice is image size / RSS.

repi/core/container.py

@@ -3,8 +3,10 @@
 from sqlalchemy.orm import sessionmaker
 from sqlmodel import select, func
 from sqlmodel.ext.asyncio.session import AsyncSession
+import importlib
 
 from repi.core.config import settings
+
 from repi.core.cache import cache
 from repi.retrieval.pgvector_store import PgVectorStore
 from repi.retrieval.fts_factory import create_fts_retriever
@@ -237,7 +239,8 @@ async def cached_find_logs_by_id(**kwargs):
             store=store,
             enable_reflection=settings.ENABLE_REFLECTION,
             reflection_interval=settings.REFLECTION_INTERVAL,
-        )
+            llm_max_calls_per_min=settings.LLM_MAX_CALLS_PER_MIN,
+            )
 
     def get_investigation_store(self, session: AsyncSession) -> InvestigationStore:
         return InvestigationStore(session)

repi/investigation/react_loop.py

@@ -107,6 +107,7 @@ def __init__(
         enable_reflection: bool = True,
         reflection_interval: int = 3,
         max_reflections: int = 2,
+        llm_max_calls_per_min: int = 60,
     ) -> None:
         self.llm = llm
         self.tools = tools
@@ -120,6 +121,7 @@ def __init__(
         self.enable_reflection = enable_reflection
         self.reflection_interval = reflection_interval
         self.max_reflections = max_reflections
+        self.llm_max_calls_per_min = llm_max_calls_per_min
         self._llm_call_timestamps: list[float] = []
 
     @staticmethod
@@ -145,7 +147,7 @@ def _ledger_summary(ledger: dict[str, dict]) -> str:
     async def _wait_for_rate_limit(self):
         now = time.time()
         self._llm_call_timestamps = [t for t in self._llm_call_timestamps if now - t < 60]
-        while len(self._llm_call_timestamps) >= 3:
+        while len(self._llm_call_timestamps) >= self.llm_max_calls_per_min:
             wait_time = 60 - (now - self._llm_call_timestamps[0]) + 1
             logger.warning(f"Rate limit: Waiting {wait_time:.1f}s...")
             await asyncio.sleep(wait_time)