PyPI - mango-brain - Versions diffs - 3.2.1__tar.gz → 3.2.3__tar.gz - Mend

mango-brain 3.2.1tar.gz → 3.2.3tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (53) hide show

{mango_brain-3.2.1 → mango_brain-3.2.3}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mango-brain
-Version: 3.2.1
+Version: 3.2.3
 Summary: The learning layer for Claude Code — persistent associative memory + development workflow. Claude Code gets smarter the longer you use it.
 Author-email: Federico Anastasi <federico.anastasi@outlook.com>
 License: MIT
@@ -42,6 +42,7 @@ Requires-Dist: scipy>=1.12; extra == "dev"
 <p align="center">
   <a href="https://pypi.org/project/mangobrain/"><img src="https://img.shields.io/pypi/v/mangobrain?style=flat-square&color=7c3aed&label=PyPI" /></a>
+  <img src="https://img.shields.io/badge/v3.2.2-7c3aed?style=flat-square" />
   <img src="https://img.shields.io/badge/python-3.11+-blue?style=flat-square&logo=python&logoColor=white" />
   <img src="https://img.shields.io/badge/MCP-compatible-8A2BE2?style=flat-square" />
   <img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" />
@@ -76,33 +77,18 @@ After 500+ memories across two real projects, session 50 is genuinely better tha
 ## How it works
-```mermaid
-flowchart LR
-    DISCUSS["/discuss\nPlan with memory"] -->|task.md| TASK["/task"]
-    subgraph agents ["4 specialized agents"]
-        direction TB
-        ANALYZER["Analyzer\nExplores code + recalls memory"]
-        EXECUTOR["Executor\nWrites code — no memory, 100% focused"]
-        VERIFIER["Verifier\nQA + checks known issues"]
-        MEMMGR["Mem-manager\nCaptures knowledge at close"]
-    end
-    TASK --> agents
-    MEMMGR -->|saves| MEMORY[("Memory")]
-    MEMORY -.->|"recalls"| ANALYZER
-    MEMORY -.->|"recalls"| VERIFIER
-    MEMORY -->|"next session"| DISCUSS
-```
-**The closed loop:** work produces knowledge, knowledge improves work.
+<p align="center">
+  <img src="assets/workflow.svg" alt="MangoBrain Workflow" width="800" />
+</p>
-| Phase | What happens |
-|-------|-------------|
-| `/discuss` | You describe what to build. Claude recalls past decisions, known bugs, patterns. Analyzers explore the codebase *and* memory. Output: `task.md`. |
-| `/task` | 4 agents execute with strict roles. The **analyzer** checks memory for gotchas. The **executor** writes code (no memory access — 100% focused). The **verifier** recalls known issues before shipping. |
-| `close` | The **mem-manager** runs automatically. Captures decisions, bugs found, patterns learned. Zero effort. |
-| `next session` | Memory surfaces relevant knowledge. The loop repeats — each cycle smarter than the last. |
+| | What happens | Codebase | Memory |
+|---|---|---|---|
+| **`/discuss`** | You explain the task. Claude explores code, recalls past decisions, brainstorms with full context. Output: `task.md`. | reads | reads |
+| **Analyzer** | Deep analysis of the areas involved. Surfaces gotchas from memory before any code is written. | reads | reads |
+| **Executor** | Writes code — 100% focused on implementation. No memory access by design. | writes | — |
+| **Verifier** | Runs tests, checks quality, recalls known issues from memory before shipping. | reads | reads |
+| **Mem-manager** | Captures decisions, bugs found, patterns learned. Zero effort from you. | — | **writes** |
+| **Next session** | `/discuss` starts with everything the last cycle learned. The loop repeats — each cycle smarter. | | reads |
 ---
@@ -111,7 +97,7 @@ flowchart LR
 ### 1. Install
 ```bash
-pip install mangobrain
+pip install mango-brain
 ```
 Lightweight install (~50MB). PyTorch and the embedding engine are installed in the next step, optimized for your hardware.
@@ -142,7 +128,7 @@ Open Claude Code in your project and paste:
 Install MangoBrain for this project.
 IMPORTANT: Use Python 3.11 or higher. Check available versions first (python --version,
 py -3.12 --version, python3.12 --version, etc.) and use the correct one for pip install.
-Run: pip install mangobrain  (using Python >= 3.11's pip)
+Run: pip install mango-brain  (using Python >= 3.11's pip)
 Then run: mangobrain install
 Then run: mangobrain serve --api (in background)
 Then tell me to open http://localhost:3101 and to restart Claude Code.

{mango_brain-3.2.1 → mango_brain-3.2.3}/README.md RENAMED Viewed

@@ -15,6 +15,7 @@
 <p align="center">
   <a href="https://pypi.org/project/mangobrain/"><img src="https://img.shields.io/pypi/v/mangobrain?style=flat-square&color=7c3aed&label=PyPI" /></a>
+  <img src="https://img.shields.io/badge/v3.2.2-7c3aed?style=flat-square" />
   <img src="https://img.shields.io/badge/python-3.11+-blue?style=flat-square&logo=python&logoColor=white" />
   <img src="https://img.shields.io/badge/MCP-compatible-8A2BE2?style=flat-square" />
   <img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" />
@@ -49,33 +50,18 @@ After 500+ memories across two real projects, session 50 is genuinely better tha
 ## How it works
-```mermaid
-flowchart LR
-    DISCUSS["/discuss\nPlan with memory"] -->|task.md| TASK["/task"]
-    subgraph agents ["4 specialized agents"]
-        direction TB
-        ANALYZER["Analyzer\nExplores code + recalls memory"]
-        EXECUTOR["Executor\nWrites code — no memory, 100% focused"]
-        VERIFIER["Verifier\nQA + checks known issues"]
-        MEMMGR["Mem-manager\nCaptures knowledge at close"]
-    end
-    TASK --> agents
-    MEMMGR -->|saves| MEMORY[("Memory")]
-    MEMORY -.->|"recalls"| ANALYZER
-    MEMORY -.->|"recalls"| VERIFIER
-    MEMORY -->|"next session"| DISCUSS
-```
-**The closed loop:** work produces knowledge, knowledge improves work.
+<p align="center">
+  <img src="assets/workflow.svg" alt="MangoBrain Workflow" width="800" />
+</p>
-| Phase | What happens |
-|-------|-------------|
-| `/discuss` | You describe what to build. Claude recalls past decisions, known bugs, patterns. Analyzers explore the codebase *and* memory. Output: `task.md`. |
-| `/task` | 4 agents execute with strict roles. The **analyzer** checks memory for gotchas. The **executor** writes code (no memory access — 100% focused). The **verifier** recalls known issues before shipping. |
-| `close` | The **mem-manager** runs automatically. Captures decisions, bugs found, patterns learned. Zero effort. |
-| `next session` | Memory surfaces relevant knowledge. The loop repeats — each cycle smarter than the last. |
+| | What happens | Codebase | Memory |
+|---|---|---|---|
+| **`/discuss`** | You explain the task. Claude explores code, recalls past decisions, brainstorms with full context. Output: `task.md`. | reads | reads |
+| **Analyzer** | Deep analysis of the areas involved. Surfaces gotchas from memory before any code is written. | reads | reads |
+| **Executor** | Writes code — 100% focused on implementation. No memory access by design. | writes | — |
+| **Verifier** | Runs tests, checks quality, recalls known issues from memory before shipping. | reads | reads |
+| **Mem-manager** | Captures decisions, bugs found, patterns learned. Zero effort from you. | — | **writes** |
+| **Next session** | `/discuss` starts with everything the last cycle learned. The loop repeats — each cycle smarter. | | reads |
 ---
@@ -84,7 +70,7 @@ flowchart LR
 ### 1. Install
 ```bash
-pip install mangobrain
+pip install mango-brain
 ```
 Lightweight install (~50MB). PyTorch and the embedding engine are installed in the next step, optimized for your hardware.
@@ -115,7 +101,7 @@ Open Claude Code in your project and paste:
 Install MangoBrain for this project.
 IMPORTANT: Use Python 3.11 or higher. Check available versions first (python --version,
 py -3.12 --version, python3.12 --version, etc.) and use the correct one for pip install.
-Run: pip install mangobrain  (using Python >= 3.11's pip)
+Run: pip install mango-brain  (using Python >= 3.11's pip)
 Then run: mangobrain install
 Then run: mangobrain serve --api (in background)
 Then tell me to open http://localhost:3101 and to restart Claude Code.

{mango_brain-3.2.1 → mango_brain-3.2.3}/mango_brain.egg-info/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mango-brain
-Version: 3.2.1
+Version: 3.2.3
 Summary: The learning layer for Claude Code — persistent associative memory + development workflow. Claude Code gets smarter the longer you use it.
 Author-email: Federico Anastasi <federico.anastasi@outlook.com>
 License: MIT
@@ -42,6 +42,7 @@ Requires-Dist: scipy>=1.12; extra == "dev"
 <p align="center">
   <a href="https://pypi.org/project/mangobrain/"><img src="https://img.shields.io/pypi/v/mangobrain?style=flat-square&color=7c3aed&label=PyPI" /></a>
+  <img src="https://img.shields.io/badge/v3.2.2-7c3aed?style=flat-square" />
   <img src="https://img.shields.io/badge/python-3.11+-blue?style=flat-square&logo=python&logoColor=white" />
   <img src="https://img.shields.io/badge/MCP-compatible-8A2BE2?style=flat-square" />
   <img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" />
@@ -76,33 +77,18 @@ After 500+ memories across two real projects, session 50 is genuinely better tha
 ## How it works
-```mermaid
-flowchart LR
-    DISCUSS["/discuss\nPlan with memory"] -->|task.md| TASK["/task"]
-    subgraph agents ["4 specialized agents"]
-        direction TB
-        ANALYZER["Analyzer\nExplores code + recalls memory"]
-        EXECUTOR["Executor\nWrites code — no memory, 100% focused"]
-        VERIFIER["Verifier\nQA + checks known issues"]
-        MEMMGR["Mem-manager\nCaptures knowledge at close"]
-    end
-    TASK --> agents
-    MEMMGR -->|saves| MEMORY[("Memory")]
-    MEMORY -.->|"recalls"| ANALYZER
-    MEMORY -.->|"recalls"| VERIFIER
-    MEMORY -->|"next session"| DISCUSS
-```
-**The closed loop:** work produces knowledge, knowledge improves work.
+<p align="center">
+  <img src="assets/workflow.svg" alt="MangoBrain Workflow" width="800" />
+</p>
-| Phase | What happens |
-|-------|-------------|
-| `/discuss` | You describe what to build. Claude recalls past decisions, known bugs, patterns. Analyzers explore the codebase *and* memory. Output: `task.md`. |
-| `/task` | 4 agents execute with strict roles. The **analyzer** checks memory for gotchas. The **executor** writes code (no memory access — 100% focused). The **verifier** recalls known issues before shipping. |
-| `close` | The **mem-manager** runs automatically. Captures decisions, bugs found, patterns learned. Zero effort. |
-| `next session` | Memory surfaces relevant knowledge. The loop repeats — each cycle smarter than the last. |
+| | What happens | Codebase | Memory |
+|---|---|---|---|
+| **`/discuss`** | You explain the task. Claude explores code, recalls past decisions, brainstorms with full context. Output: `task.md`. | reads | reads |
+| **Analyzer** | Deep analysis of the areas involved. Surfaces gotchas from memory before any code is written. | reads | reads |
+| **Executor** | Writes code — 100% focused on implementation. No memory access by design. | writes | — |
+| **Verifier** | Runs tests, checks quality, recalls known issues from memory before shipping. | reads | reads |
+| **Mem-manager** | Captures decisions, bugs found, patterns learned. Zero effort from you. | — | **writes** |
+| **Next session** | `/discuss` starts with everything the last cycle learned. The loop repeats — each cycle smarter. | | reads |
 ---
@@ -111,7 +97,7 @@ flowchart LR
 ### 1. Install
 ```bash
-pip install mangobrain
+pip install mango-brain
 ```
 Lightweight install (~50MB). PyTorch and the embedding engine are installed in the next step, optimized for your hardware.
@@ -142,7 +128,7 @@ Open Claude Code in your project and paste:
 Install MangoBrain for this project.
 IMPORTANT: Use Python 3.11 or higher. Check available versions first (python --version,
 py -3.12 --version, python3.12 --version, etc.) and use the correct one for pip install.
-Run: pip install mangobrain  (using Python >= 3.11's pip)
+Run: pip install mango-brain  (using Python >= 3.11's pip)
 Then run: mangobrain install
 Then run: mangobrain serve --api (in background)
 Then tell me to open http://localhost:3101 and to restart Claude Code.

{mango_brain-3.2.1 → mango_brain-3.2.3}/pyproject.toml RENAMED Viewed

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "mango-brain"
-version = "3.2.1"
+version = "3.2.3"
 description = "The learning layer for Claude Code — persistent associative memory + development workflow. Claude Code gets smarter the longer you use it."
 readme = "README.md"
 license = {text = "MIT"}

{mango_brain-3.2.1 → mango_brain-3.2.3}/server/cli.py RENAMED Viewed

@@ -445,10 +445,23 @@ def _setup_mcp_json(project_path: Path) -> None:
 def _patch_claude_md(project_path: Path, project_name: str) -> None:
-    """Add MangoBrain section to project's CLAUDE.md if not already present."""
-    claude_md = project_path / "CLAUDE.md"
+    """Add MangoBrain section to project's CLAUDE.md if not already present.
+    Searches for CLAUDE.md in order: .claude/CLAUDE.md, then root CLAUDE.md.
+    Patches whichever is found first. Creates root CLAUDE.md only if neither exists.
+    """
+    claude_md_inner = project_path / ".claude" / "CLAUDE.md"
+    claude_md_root = project_path / "CLAUDE.md"
     marker = "## MangoBrain"
+    # Find existing CLAUDE.md — prefer .claude/ over root
+    if claude_md_inner.exists():
+        claude_md = claude_md_inner
+    elif claude_md_root.exists():
+        claude_md = claude_md_root
+    else:
+        claude_md = claude_md_root  # Create at root if neither exists
     if claude_md.exists():
         content = claude_md.read_text(encoding="utf-8")
         if marker in content: