minder-cli 0.3.4__tar.gz → 0.3.8__tar.gz

{minder_cli-0.3.4 → minder_cli-0.3.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: minder-cli
-Version: 0.3.4
+Version: 0.3.8
 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
@@ -37,83 +37,99 @@ Requires-Dist: zipp>=3.21.0; extra == 'server'
 Requires-Dist: zstandard>=0.25.0; extra == 'server'
 Description-Content-Type: text/markdown
 
-# minder-cli
+# Minder
 
 [![PyPI version](https://img.shields.io/pypi/v/minder-cli.svg)](https://pypi.org/project/minder-cli/)
 [![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
 
-**minder-cli** is the command-line interface for [Minder](https://github.com/hiimtrung/minder) — a self-hosted MCP platform for repository-aware engineering intelligence.
+**Minder** is a self-hosted MCP (Model Context Protocol) platform for repository-aware engineering intelligence.
 
-The CLI handles IDE scaffolding, repository sync, authentication, and self-updates. It connects to a **Minder Server** which runs the MCP gateway, RAG pipeline, memory engine, and admin console.
+It combines a local-first inference stack, a persistent memory and workflow engine, and a developer-facing CLI into a single deployable unit.
 
-## Installation
+## What's in this repo
 
-```bash
-# Recommended
-uv tool install minder-cli
+| Component | Description |
+|-----------|-------------|
+| **Minder Server** | MCP gateway — SSE + stdio transport, RAG pipeline, workflow engine, memory, admin HTTP |
+| **Minder CLI** (`minder-cli` on PyPI) | Edge CLI — IDE scaffold, repo sync, login, self-update |
+| **Minder Dashboard** | Astro admin console — client management, onboarding, skill catalog |
 
-# Alternative
-pipx install minder-cli
-```
+## Architecture
 
-## Requirements
+```
+Developer → minder-cli → Minder Server ←→ AI agents (Codex / Copilot / Claude)
+                              │
+               ┌──────────────┼──────────────┐
+               │              │              │
+           MongoDB          Redis         Milvus
+         (graph/memory)   (cache)      (vector search)
+               │
+          ┌─────┴──────┐
+          │            │
+       LiteRT-LM    FastEmbed
+       (LLM gen)   (embedding)
+```
 
-A running [Minder Server](https://github.com/hiimtrung/minder) — see the [server setup guide](https://github.com/hiimtrung/minder/blob/main/docs/minder-server.md) to get one running in minutes with Docker.
+- **LLM inference**: LiteRT-LM (Google AI Edge) — on-device, hardware-accelerated, no HTTP overhead
+- **Embedding inference**: FastEmbed running natively in-process (`mxbai-embed-large-v1`) — zero-dependency ONNX acceleration
 
 ## Quick Start
 
-### 1. Connect to your server
+### Run the server
 
 ```bash
-minder login --client-key mkc_your_client_key --server-url http://localhost:8800/sse
+# 1. Download the LiteRT-LM model
+./scripts/download_models.sh
+
+# 2. Start infra (MongoDB + Redis + Milvus)
+docker compose -f docker/docker-compose.local.yml up -d
+
+# 3. Run Minder Server
+uv run python -m minder.server
 ```
 
-### 2. Set up IDE integration
+Or use the one-command release installer (Docker only, no local uv needed):
 
 ```bash
-# In your project root — scaffolds MCP config for your editor
-minder install-ide --target vscode --target claude-code
+curl -fsSL https://raw.githubusercontent.com/hiimtrung/minder/main/scripts/release/install-minder-release.sh | bash
 ```
 
-Supported targets: `vscode`, `cursor`, `claude-code`.
-
-### 3. Sync a repository
+### Connect the CLI
 
 ```bash
-minder sync --repo-id <repository-uuid>
-```
+# Install
+uv tool install minder-cli
 
-Indexes code and documentation so AI agents can use semantic search and RAG tools.
+# Log in with your client key
+minder login --client-key mkc_your_key --server-url http://localhost:8800/sse
 
-## Commands
+# Set up IDE integration (VS Code, Cursor, Claude Code)
+minder install-ide --target vscode --target claude-code
 
-| Command | Description |
-|---------|-------------|
-| `minder login` | Authenticate the CLI against a Minder Server |
-| `minder install-ide` | Scaffold MCP assets for VS Code / Cursor / Claude Code |
-| `minder sync` | Index a repository into the Minder Server |
-| `minder check-update` | Check for available CLI and server updates |
-| `minder self-update` | Update CLI or server in place |
+# Sync a repository
+minder sync --repo-id <uuid>
+```
 
-## MCP Tools (once connected)
+## MCP Tools
 
-Your AI agents will have access to:
+When connected, Minder exposes these tools to your AI agents:
 
 | Tool | Description |
 |------|-------------|
 | `minder_query` | Full RAG pipeline: retrieve → reason → verify → respond |
-| `minder_search_code` | Semantic code search across indexed repositories |
+| `minder_search_code` | Semantic code search across indexed repos |
 | `minder_memory_recall` | Retrieve persisted engineering memory |
 | `minder_workflow_get` | Read current workflow state |
 | `minder_session_restore` | Restore session continuity across context windows |
 
-## Links
+## Documentation
 
-- [GitHub Repository](https://github.com/hiimtrung/minder)
-- [Server Setup Guide](https://github.com/hiimtrung/minder/blob/main/docs/minder-server.md)
-- [Local Dev Setup](https://github.com/hiimtrung/minder/blob/main/docs/guides/local-setup.md)
-- [System Design](https://github.com/hiimtrung/minder/blob/main/docs/system-design.md)
+- [System Design](docs/system-design.md)
+- [Server Setup](docs/minder-server.md)
+- [Local Dev Setup](docs/guides/local-setup.md)
+- [Production Deployment](docs/guides/production-deployment.md)
+- [Admin & Client Onboarding](docs/guides/admin-client-onboarding.md)
 
 ## License
 
-Apache License 2.0. See [LICENSE](https://github.com/hiimtrung/minder/blob/main/LICENSE) for details.
+Apache License 2.0. See [LICENSE](LICENSE) for details.

{minder_cli-0.3.4 → minder_cli-0.3.8}/pyproject.toml

@@ -4,9 +4,9 @@ build-backend = "hatchling.build"
 
 [project]
 name = "minder-cli"
-version = "0.3.4"
+version = "0.3.8"
 description = "Minder CLI is the command-line interface for the Minder self-hosted MCP platform."
-readme = "README-pypi.md"
+readme = "README.md"
 requires-python = ">=3.14"
 keywords = ["mcp", "cli", "ai", "code-search", "workflow"]
 classifiers = [

{minder_cli-0.3.4 → minder_cli-0.3.8}/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-0.3.8}/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-0.3.8}/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-0.3.8}/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,