minder-cli 0.3.4__tar.gz → 0.3.6__tar.gz

{minder_cli-0.3.4 → minder_cli-0.3.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: minder-cli
-Version: 0.3.4
+Version: 0.3.6
 Summary: Minder CLI is the command-line interface for the Minder self-hosted MCP platform.
 Project-URL: Homepage, https://github.com/hiimtrung/minder
 Project-URL: Repository, https://github.com/hiimtrung/minder

{minder_cli-0.3.4 → minder_cli-0.3.6}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "minder-cli"
-version = "0.3.4"
+version = "0.3.6"
 description = "Minder CLI is the command-line interface for the Minder self-hosted MCP platform."
 readme = "README-pypi.md"
 requires-python = ">=3.14"

{minder_cli-0.3.4 → minder_cli-0.3.6}/src/minder/config.py

@@ -46,7 +46,7 @@ class LLMConfig(BaseModel):
     litert_model_path: str = "~/.minder/models/gemma-4-E2B-it.litertlm"
     litert_backend: str = "auto"  # "auto" (GPU on Mac, CPU elsewhere) | "cpu" | "gpu"
     litert_cache_dir: str = "~/.minder/cache/litert"
-    context_length: int = 131072
+    context_length: int = 32768
     temperature: float = 0.1
     openai_api_key: Optional[str] = None
     openai_model: str = "gpt-4o-mini"

{minder_cli-0.3.4 → minder_cli-0.3.6}/src/minder/graph/state.py

@@ -24,3 +24,4 @@ class GraphState(BaseModel):
     transition_log: list[dict[str, Any]] = Field(default_factory=list)
     retry_count: int = 0
     metadata: dict[str, Any] = Field(default_factory=dict)
+    chat_history: list[dict[str, Any]] = Field(default_factory=list)

{minder_cli-0.3.4 → minder_cli-0.3.6}/src/minder/llm/litert.py

@@ -28,7 +28,7 @@ logger = logging.getLogger(__name__)
 
 _ENGINE_CACHE: dict[str, Any] = {}
 _CONVERSATION_CACHE: OrderedDict[uuid.UUID, Any] = OrderedDict()
-MAX_CACHED_CONVERSATIONS = 5
+MAX_CACHED_CONVERSATIONS = 3
 
 
 class LiteRTModelLLM:
@@ -39,7 +39,7 @@ class LiteRTModelLLM:
         model_path: str = "~/.minder/models/gemma-4-E2B-it.litertlm",
         backend: str = "cpu",
         cache_dir: str = "~/.minder/cache/litert",
-        context_length: int = 131072,
+        context_length: int = 32768,
     ) -> None:
         self._model_path = str(Path(model_path).expanduser())
         self._backend = backend
@@ -189,12 +189,14 @@ class LiteRTModelLLM:
 
         reasoning_output = getattr(state, "reasoning_output", {}) or {}
         prompt = reasoning_output.get("prompt") or state.query
+        chat_history = getattr(state, "chat_history", []) or []
         text = self.complete_text(
             str(prompt),
             max_tokens=256,
             temperature=0.1,
             fallback=fallback,
             session_id=state.session_id,
+            chat_history=chat_history,
         )
 
         return {
@@ -235,11 +237,12 @@ class LiteRTModelLLM:
         try:
             reasoning_output = getattr(state, "reasoning_output", {}) or {}
             prompt = str(reasoning_output.get("prompt") or state.query)
+            chat_history = getattr(state, "chat_history", []) or []
             
             # Use session cache if available. Reusing the conversation preserves the KV-cache state.
             # For linear conversations, we don't pass 'messages' to send_message_async because
             # it builds on top of the existing state in the Conversation object.
-            conv = self._get_conversation(state.session_id, initial_messages=[])
+            conv = self._get_conversation(state.session_id, initial_messages=chat_history)
             
             for chunk in conv.send_message_async(prompt):
                 for item in chunk.get("content", []):
@@ -267,13 +270,14 @@ class LiteRTModelLLM:
         temperature: float = 0.1,
         fallback: str = "",
         session_id: uuid.UUID | None = None,
+        chat_history: list[dict[str, Any]] | None = None,
     ) -> str:
         """Simple text-in/text-out completion."""
         if self.runtime != "litert":
             return fallback
 
         try:
-            conv = self._get_conversation(session_id, initial_messages=[])
+            conv = self._get_conversation(session_id, initial_messages=chat_history or [])
             response = conv.send_message(prompt)
             return cast(str, response["content"][0]["text"])
         except Exception as e:

{minder_cli-0.3.4 → minder_cli-0.3.6}/src/minder/tools/query.py

@@ -86,6 +86,21 @@ class QueryTools:
             "query_reasoning",
             self._store,
         )
+        chat_history = []
+        if session_id:
+            try:
+                history_docs = await self._store.list_history_for_session(session_id)
+                chat_history = [
+                    {
+                        "role": str(getattr(doc, "role", "")).replace("assistant", "model"),
+                        "content": str(getattr(doc, "content", "")),
+                    }
+                    for doc in history_docs
+                    if getattr(doc, "role", "") and getattr(doc, "content", "")
+                ]
+            except Exception:
+                pass
+
         state = GraphState(
             query=query,
             session_id=session_id,
@@ -93,6 +108,7 @@ class QueryTools:
             repo_id=repo_id,
             repo_path=repo_path,
             workflow_context=workflow_context,
+            chat_history=chat_history,
             metadata={
                 "verification_payload": verification_payload,
                 "max_attempts": max_attempts,
@@ -181,6 +197,21 @@ class QueryTools:
             "query_reasoning",
             self._store,
         )
+        chat_history = []
+        if session_id:
+            try:
+                history_docs = await self._store.list_history_for_session(session_id)
+                chat_history = [
+                    {
+                        "role": str(getattr(doc, "role", "")).replace("assistant", "model"),
+                        "content": str(getattr(doc, "content", "")),
+                    }
+                    for doc in history_docs
+                    if getattr(doc, "role", "") and getattr(doc, "content", "")
+                ]
+            except Exception:
+                pass
+
         state = GraphState(
             query=query,
             session_id=session_id,
@@ -188,6 +219,7 @@ class QueryTools:
             repo_id=repo_id,
             repo_path=repo_path,
             workflow_context=workflow_context,
+            chat_history=chat_history,
             metadata={
                 "verification_payload": verification_payload,
                 "max_attempts": max_attempts,